Humans + AI

– Markus Buehler

Markus Buehler is Jerry McAfee (1940) Professor in Engineering at Massachusetts Institute of Technology (MIT) and Principal Investigator of MIT’s Laboratory for Atomistic and Molecular Mechanics (LAMM). He has published over 450 articles with almost 50,000 citations and is on the editorial boards of numerous journals including PLoS ONE and Nanotechnology. He has received numerous awards including Presidential Early Career Award for Scientists and Engineers (PECASE) and National Science Foundation CAREER Award. In addition he is a composer and has worked on two-way translation between material structure and music.

Wikipedia Profile: Markus J. Buehler
Google Scholar Page: Markus J. Buehler
LinkedIn: Markus J. Buehler
MIT Page: Markus J. Buehler

Ross Dawson: Marcus, it is fantastic to have you on the show.

Markus Buehler: Thanks for having me.

Ross: So you sent me a paper, which is titled, Accelerating Scientific Discovery with Generative Knowledge Extraction, Graph-Based Representation, and Multimodal Intelligent Graph Reasoning, and it totally blew my mind. So I want to try to use the opportunity to unpack it to a degree. It’s an 85-page paper, so obviously I won't be able to get out of the detail level, but to unpack the concepts, because I think they're extraordinarily relevant, not just for accelerating scientific discovery, but also across almost any thinking domain. It's very, very rich and very promising just because so much to my interest. So let's start off and essentially, I'm saying, you've taken a thousand papers, and from those have been able to distill those into some ontological knowledge graphs. So could you please explain ontological knowledge graphs, how those are created, and what they are?

Markus: Sure, yeah, so the idea behind this sort of graph representation is really changing information into knowledge. And what that means is that we're trying to take bits and pieces of information, like a concept — concept A, concept B, like a flower, composite a car. And in these graph representations, we were trying to connect them to understand how a car, a flower, and a composite are related. And, traditionally, we would create these knowledge graphs, manually, essentially, would create sort of categories of what kind of items we want to describe, and what the relationship might be. And then we would basically manually build these relationships into a graphic presentation. And we've done this for a couple of decades, actually. I think the first paper was 10 to 20 years ago. And yeah, back in the day, we did this manually, essentially understanding a certain scientific area. We would build graph representations of the knowledge that connect information and understanding structurally what's going on. 

And then now, of course, in the paper, and we'll probably talk more about this, we have been able to do this using Generative AI technologies. And this allows us to, as you said, build these knowledge graphs for a thousand papers or more, and do it in the way, actually in an automatic way. So we don't have to manually read the papers and understand them, and then build the knowledge graph, we can actually have AI systems build these graphs for us. And this, of course, is a whole different level of scale that we can now access.

Ross: So there is an important word there, ontological. So what's the importance of that?

Markus: Yeah, so when we think about concepts, like, let's say, we take a look at biological materials, a lot of them are made from proteins. Proteins are made of amino acids. And there are certain rules by which you put amino acids together, which in turn are encoded by DNA. And depending on the pattern you have in the DNA, and then in the protein sequence, you're going to get different protein structures, which have different functions. And, certain types of sequences will give you a helical structure like a telephone cord, others will give you a brick-like structure like a widget configuration. And so what I've just talked about really are relationships between concepts and amino acid protein break properties. And these are ontologies that basically describe how things work. And things as you alluded to, could be anything and it could be scientific concepts. It could be business concepts, it could be social science concepts, it could be really a lot of different types of things. 

So these ontological representations allow us to understand how these different building blocks, like common building blocks, are related and how these relationships ultimately lead to certain properties. And so in these models, essentially, what we are really trying to understand is how properties emerge from the building blocks and their nonderivative relationships. Right? So for example, looking at biological materials, and proteins, there are certain patterns that are really important, maybe a single mutation in a protein, right? We have all heard of genetic diseases. There are certain mutations, a single point mutation and a protein out of 1000s of amino acids could create a disease, whereas a million other mutations do not create a disease at all right? And so those are the kinds of things we're trying to understand. And usually in science, we like to build models of such things. We'd like to understand what's important, what can we ignore, what kinds of relationships are really critical, and how we model them. So, in a non-ideological presentation, we look at data, and we're trying to build a representation of how the system works. These things can be pretty complex for real-life systems because there are a lot of nuances and how the world ultimately works.

Ross: So these are conceptual relationships. And in this case, they are distilled from the text in the papers to essentially an emergent structure to ensure the relationships between these concepts.

Markus: Yes, so let's say you read a paper as a human, but you can sort of look at the paper as a collection of words. And when you read it, you make sense of the sentence or the words, the abstracts, the paragraphs. And what you do in your brain essentially, is you build relationships, you build an ontological concept map of what's going on, you have information, which is the words, the sentences, and the way you understand the paper really is by building, creating a representation in your mind, of how the words relate in the sentence, how the sentences relate, and how maybe individual words in multiple sentences relate and you understand, “oh okay”, there's a specific nuance in the way this word is described, which really allows me to understand the entire abstract or the entire paper, because there's sort of this detail, right?’ That matters, just like in the amino acid sequences. And so those relationships really are what you kind of take in when you read and the AI systems work actually in a similar way especially transformer-based architectures they quite literally take information, which are tokens. They built a presentation map internally of how these tokens are related, which is a graph representation, which then we call knowledge — knowledge is how information is related. 

And there's a, I think, a reference in the paper to some of the discussions at five had about similarities, of course, before AI, but he really talked about how information is well important. But really, knowledge is the relationship between information pieces, and this is what we try to distill out and science is to understand the relationships and then translate them across different areas, which we'll also probably talk about later on as well, using isomorphisms. But not going there quite yet. But yeah, so information to knowledge is really what we do as humans do what AI systems do internally. When we're building these Knowledge Graph representations. We're doing it on scientific papers or tax patterns information out there. And it is to make it accessible to a human right. So when you think about a large language model, it builds graphs internally in what we call the embedding space of the hidden space, which we do not understand its very high dimensional vectors, and tensors, which are all related and that graphs are being constructed. But I look at it, you look at it, we have no clue what the model is really thinking internally, building knowledge graphs, using human language or numbers, numerical symbols, and so on, allows us to actually trace knowledge information in a way that humans can understand. It's sort of the same level of abstraction as what we have in a scientific paper. Except if you read the paper, you as a human, take a couple of hours to read and understand. If you want to read a thousand papers, I will say by the time you read the 10th paper you probably have forgotten about the first paper already, right? So if we can automate this process and build these connections between bits and pieces of information in the papers, we can suddenly have AI systems help us and actually read a thousand papers and build a really, really powerful representation, which then humans can interrogate, and actually mine or ask questions about, or get the AI system to give us new hypotheses, or anticipate behaviors and so on. There are lots of different things we can do once we have these knowledge graphs that have been extracted, and again, they are human-readable. So we can ask the AI system questions, and they can tell us the answer. And we can actually trace how the model thought about giving the answer. And that's a very powerful way of interpretability, which is oftentimes quite important, especially for human-AI collaborations.

Ross: Yep, absolutely. So one of the ways in which these are accessible is through visualizations of the knowledge graphs. So these, of course, basically collapse the intense multi-dimensionality of the embedding structure of the concepts and their relationships from the AI mapping down to two dimensions for a visual that humans can see. So I mean, without going into too much depth in that, what is the process of then pulling that into a visualization, which can be useful for inspiring or being able to help people understand relationships between concepts?

Markus: Yeah, that’s a quite good question. So maybe I can walk through a little bit of the process by which these are constructed from the original data, the raw data. So our data here are scientific papers, as you said, in about a thousand. We've done it for more than that now. But, it's maybe a limitation of the computer you have, but let's say you have thousands of papers like we did in the paper. And you essentially, at AI systems, we read these papers, and we do a distillation process by which we ask the AI system, the nominee, to read it, but actually extract useful information, like what's in the paper? What are their prophecies, what are the key findings, numerical insights, quantitative, strategic decisions, the model that the scientists have made, and so on, and so on? 

So we give a kind of a very detailed nuanced summary of the paper in, chunks of text, essentially, we do this for chunks, so we don't look at maybe the entire paper, but look at sections of the paper, and for each section, we get this sort of very detailed understanding of what's being described in the paper, these sort of distilled. And the reason why we do this is quite similar to what people do and chain of thought reasoning, or multi-step reasoning, react reasoning, where you essentially, don't just use a single agent and ask a question to your AI model, but you actually sort of deliberately asked multiple questions, give me this answer, critique this answer, give me another angle, summarize in that way, and so on. And so this gives sort of a very detailed description of what is in the paper. But so this then forms, this is the information. So that's just what's in the section of the paper. Now, we build graphs of each section. These are called kind of local graphs. So those are basically saying, Oh, love it, this description of summary. And build an ontological representation. So tell me, what are the key concepts in there, and what are the relationships between them by the might be,  a protein and amino acid, right? And so the paper might talk about how I mean, acids are combined to form proteins. But these proteins are very strong. And because these proteins have a strength that's been used to build a composite, which the authors have talked about, maybe applying it in the airplane or going to coding. And so that's where that graph ends. So now the next section might talk about the synthesis process. So it might say, Yeah, I'm making these proteins out of these amino acids. But I'm actually using this particular chemical process to make them and I'm using this organism to grow the individual protein maybe, and then purify them. And I use this machinery, this chemical process, and so on. So you kind of go through these different steps of how to make it. And the third part of the paper might talk about, maybe this, the theory behind it, maybe the office has developed a model of how this can be done, and maybe a molecular simulation, and so on. And so these are all kinds of separate graphs. Now, ontological graphs, describe concepts and relationships, like we talked about earlier. And now we do this for all these 1000 papers, and all the sections in the 1000 papers. 

So this, 10s of 1000s, 100s of 1000s of different sorts of small graphs that are now in my storage, basically, now what I do, I combine them, and because they're overlapping relationships, and so there's some transitive properties here. So if one section talks about protein and how it's made, another section talks about the protein and how it's applied. Another section talks about the protein and maybe its toxicity or biomedical applications. When you connect this in a graph, the nodes are going to be connected in many different ways. So the node protein might be connected to manufacturing and all the processes as well from this, there might be a connection between protein and toxicity, and so on, and so on. And you can see this graph now being constructed. So similar nodes are actually combined into a single node. And you mentioned embedding that so we can actually use natural language processing, to say, Yeah, I might have a terminology for a protein. In one section, it's called the protein in another section is called proteins, right? In another section that might be called amino acid groups or something, I'm making up sort of some, ways of how scientists describe things. Not embedding models allows us to understand that these are actually identical, similar things, similar entities. And so what we do is we group them together, we say instead of calling it protein, or proteins, or groups of amino acids to actually the same thing, and they're all called protein. 

And so this step allows us to sort of combine different concepts which would have been different nodes into one single node. So this helps to sort of make the graph simpler and more compact and more accurate because different scientists, different groups, different people are going to use different terminologies, but embedding models allow us to combine them and then we do this combination process for every single small graph and combine them into a gigantic graph and what happens because these concepts through this process of distilling information into similar terms, combining similar terms, the there's a consistency here. And so you create very, very deep connectivity across things because proteins occur in hundreds of papers and manufacturing occurs and hundreds of different papers, and they have different relationships, they're not just connected in one way, they actually connected in many different ways. Because manufacturing might relate to a chemical process A, B, or C, this process might be used in a lot of different contexts and might be used to make proteins. However, the same chemical process might also be used to make a polymer coating or paint. And so in a way, when you read the individual papers, you don't know that because you just read the papers. But because he asks them about these 1000 papers, and it's creating these graphs, it's connecting them. Now, the whole graph, when you look at it, you can actually see the connections. And so you're asking about visualization, right? So this graph is obviously very big, you can really look at it on a screen because, of 10s, or 1000s of nodes and connections and edges. 

So what we do is we usually look at a subgraph. And so we say, let's say we want to look at one concept, like graphene, or protein or health or whatnot, manufacturing process, some kind. And we can just click on that note, and we can sort of see, okay, what are the neighbors to that node. And so we create a subgraph representation, let's say, look at one node and all the first, second, and third nearest neighbors. And that gives us a graph, we can now fit on the screen basically. Or I can look at two concepts, I can ask the question, I have a concept like graphene, and a concept like sustainability, or something like graphene and music, something totally weirdly different. But can you identify a connection between them? And so we can have an algorithm, look for the path that connects graphene with music or graphene with sustainability or graphene with whatever you pick. And of course, if there is no connection, the model will tell us there's no connection, but we can use the embedding model again and say, Hey, if you don't find music in the graph, find me the closest node that relates to music. Okay, and so then we'll find that, that node, and we'll actually identify a graph between them.

Ross: We're just gonna say, this takes us to the concept of isomorphic mapping. So taking well could be areas of the graph that have similar structural similarity, or structurally similar and be able to find those. And so then, I suppose part of it is being able to identify where there is structural similarity, the alternative is to take one area of a particular structure and then map it against another domain.

Markus: Right, yeah, so it was a great, great, good point. I mean, there are sort of two ways in which we could connect this. So one of the things I should take a step back, why we are interested in connecting different things. I mean, that's all that science is about, or technology, you want to have a solution to a problem, like, let's say you have, you want to build a, I don't know, very scratch resistant paint or coating. And you want to use graphene. So the question is that as an engineer, or scientist, you come, you say, Well, how do I do this? Well, I want to understand how graphene is connected to crack-resistant coatings, especially coatings. And so you can find a path maybe between these concepts, right? So like I described, you basically look at the graph and find similar nodes. So if the graph has this connection, you can build that out. And you can then read the graph, essentially, and it will tell you how to get from graphene to scratch-resistant coatings. And it tells you a lot. And so you can then develop technologies out of that, or you can feed it into an AI system, which the, we call graph reasoning where you say, okay, instead of just answering directly to the AI, look at the graph, look at the subgraph of how graphene and scratch resistant coatings are related. And look at this entire path, the relationships, and maybe even the source papers that came where this came from. And answer the question, right, so this gives a very, very detailed sort of substrate for the model to think about. 

But now to your question: So what if there's no connection? Okay, so what if you have different graphs that are not actually connected, right? So for example, music, music theory, and material science or philosophy, they might have no connection or very few connections. And so you really can't quite figure out how to get from point A to point B, because there really isn't any paper that talks about, the relationship directly or even multiple papers that you can use to build a graph between them. So that's what we use isomorphic mapping we basically look at, graph structures, we say, okay, if I'm interested in how materials become resilient, or tough, resilient, what would be an, so we identify structures that cause that describe essentially in the graph, resilience in materials and so, those are sort of groups of structures, maybe in there that talk about how this happens mechanistically like, you have to build a composite you have to build a particular pattern, you have to do a certain type of processing step. So that gives the graph structure that tells us how to do this, how to achieve this in engineering from materials and now we say can you find a similar kind of structure in music, either can you find it a valence something that a topic you might be interested in? Or just look at the entire graph and see what is similar with similar patterns. 

So, in the paper, we've done this sort of an experiment to say, Okay, I have, the concept of the graph that I've identified in materials, what would be can you identify similar structures are identical structures, if we truly do isomorphic mapping, and it has to be identical topologically in the music space, and then we can look at them and we can see, okay, so here's the same graph topology, in material science, and then music. So the graph, if you look just at the graph, the nodes and the edges, they look exactly the same. What's different is within the nodes and the edges. So in the materials, domain, nodes, and edges include things like atoms, microstructure forces and stresses, and things like this. In music, they're going to talk about scales, tonality, maybe the composer's name, and musical concepts, essentially. So what you can now do is you can say, Okay, I have now an isomorphic, mapping the same graph structure in materials and in music. 

So which have you discovered? Yes, exactly. Yeah. Yep. Yeah, so I discovered this now from the data. And, and this all can be done algorithmically, all through to computational processes. And so now what I can do, I can look at what's inside the nodes and the artists. And that's really the interesting part. And it's not just that there's a similar structure, which is already interesting, but I can actually sort of ask the question, so what do the nodes and the edges mean? And materials and music? And how does this picture now relate to one another? And, of course, we can look at this as a human, we can look at analyze this, or we can give these two graphs now to an AI system and ask the question, like we've done in the paper, hey, look at these two graphs that so often mappings between these concepts, interpret this for me, tell me how the relationships could be explained, and do it in the tape, right, So we can give some structure to the thinking. We basically ask the model in multiple steps first to make a table of the nodes and the edges. And then we say, add a column in the table that describes the relationship, and maybe add another column to describe an interpretation of the relationship, so how would I understand these graphs, what they mean in materials and music, and what the translation means? And so this is something that can be automated, fully using, these AI systems. Now,

Ross: So on the paper, you used Beethoven's Ninth, and mapping that against some material structures, I believe. So, in doing that exercise, what was most striking to you about it? What did you see from that mapping of Beethoven's Ninth to the domain? Which made you sort of surprised or insightful?

Markus: Yes, I think we'd have, first of all, I mean, we, we have tried, as I mentioned earlier, actually, you would have looked at these types of relationships, especially between, music and materials in previous work could have more with pen and paper methods. So that's, analyzing the structural hierarchies and the relationships. And it's very human bias, essentially, you, you basically, do the analysis in that case, based on what you anticipate or know or understand. And, and that's sort of limiting because, especially, across multiple domains, we'd like to have an automatic process. So the first thing that is surprising was that, first of all, we found, very similar graph structures that actually could be isomorphically mapped between music and the materials graph, and when we saw those structures, and you can see this in the paper, and you can probably pull it up and, in the video, you can actually see that the graphs, really look exactly identical to what logically right so they are, exactly what the algorithm is asked to do, and sort of discovering these, these, these isomorphic, mappings and then you can look at what's sort of in these in these in these notes and in the edges, and you can ask, the kind of the question, how could they be related and I'm kind of maybe, kind of, you can actually look at the graph, and there's, there's a table in the paper, I'm actually trying to pull it up right now to kind of go through some of the examples. So there's for example, so I'm looking at figures figure eight in the paper for those of you who want to know my friends, this shows the two graph structures okay. And so, you can see that they are identical topologically meaning that they can be isomorphically mapped so every node and every edge can be connected to the other graph in the different domain. 

And then you can look at the, the individual, kind of, specific I would say specific, the content and the nodes and you can see that in in the materials worlds, the nodes have things like adhesive force or beam and failure and characteristic length dimensions and structural features, buckling behavior, and so on. And in the music world, again, we have things like tonality, the composer, then Beethoven, F major C major, different scales, and so on. And so then the question that we then looked at, and sort of the surprising thing was, I thought, Okay, this is great, but I can interpret this, but it's going to be biased by my humans by my understanding of the system. So why don't I let an AI system that looks at these edges and labels in the two different systems, and, it kind of explains how they're related? Okay, and so then in the, in the, in the paper, you can read through the different, the specific analysis behind it, but, kind of the assistant will actually try to explain how they could be potential, connected, and that I found actually quite, quite surprising, first of all, that the model could give us an answer. And, of course, Mom was very eager to answer. So, generally, they, that's what they like to do. 

But the answers actually were quite, quite, quite rich in the way they understood the topics. That's part of what, transform models will do very well as they are very good at connecting in translating insights and ideas from one domain to another. So for example, if you say, write me a poem, in a style I don't know, yeah, make a poem of spider silk in the style of Shakespeare, it will do a pretty good job with that. And so similarly, here, we're basically giving an example of how things look in one domain and another domain giving graph structure. And the model is sort of tasked to explain these relationships, and they're very good and kind of interpolating between these different domains. Now, the graph is really critical here, because the graph gives the model something to think about, right? So without the graph, if I were to ask the model directly, tell me this relationship between music and materials, they'll give me an answer. But it's not very rich, it's not very nuanced, it's going to be a more generic answer. But if you give the graph the substrate to think about, which has, as we talked about earlier, there's a lot of work that goes into this to build this graph using the AI tools. Now, the model has a much more, much deeper, I would say substrate to think about, and that's exactly what. And then so now, the answer is actually much more intelligent, essentially. And I think that's really the surprising part was that, yes, we can automate this, it creates these really interesting isomorphic mappings. And can actually tell us something about these mappings between the domains, which are free of my bias that I have, because I might be an expert in one field, but not in the other. Or I might be more familiar with one or the other. And so my answer is going to be biased by whatever I know, or anyone else for that matter.

Ross: So I mean, it's a massive topic, but just sort of just begin to answer it is. So then what's the value for the scientist, material scientist, who then says, Okay, I'm gonna map Beethoven's Ninth against it, you get these nice illustrations of the relationship between the concepts? Right, so what do the scientists do with that?

Markus: Yes, well, so one thing that we are quite interested in science is to expand the horizon of what we can maybe build or understand or hypothesize. And so one clear thing is, you can look at these now and say, Okay, if, if, if I understand, how these are related, potentially what I can do, I can maybe get a new hypothesis about it. So I can say, if this is how resilience looks like in music, I can maybe look at a different, part of the musical graph and say, can you use this previous analogy that you've already developed? Right, but now go to the next step, and explain to me how I could use maybe even let's give you a specific example, you take a look at Beethoven's Symphony, and you look at the, maybe part of that is particularly interesting, musically, and you say, okay, here the, here's this part of the graph that describes this the construction of this particular part of the music, can you tell me how I could utilize this as a material design principle? What would that mean? Right? So that's sort of a very specific example. And then the AI model will because it does in context learning, it understands the previous graphs, the previous answers, so it's kind of like a chat interaction, right? And then you end it will then say, okay, so here's my extension of what this graph will look like, if I were to take this new part of the graph and music and I would apply this to the material. So this now is sort of where the novelty comes in, because this new part of the music did not have an analogy yet in materials, right? So that's something I'm asking them to create for me. 

So now that creates a new graph, which is a hypothetical graph that does not yet exist in materials and engineering science. but it exists in music. I can use this understanding of how relationships work to move into the zone, this could actually be done mathematically as well because you mentioned, we talked earlier about embeddings, as embeddings provide us with an opportunity to understand what things mean in a sort of abstraction of a vector space. And I can actually look at Graph representations and look at how relationships look in a graph with respect to, changes in the vector. So if you imagine going from one point from one node to another node, there's a relationship. And so this can be expressed as a vector, essentially, in this high dimensional embedding space. So I can even sort of formalize this mathematically and say, if I were to have an extension of the graph, in the wave music, extensive sort of every connection between a node and a node or node is sort of a vector. And if this node does not exist in materials, let's say in engineering, I can still compute what it would be. I can then solve the inverse problem and say, this node does not exist yet in engineering and science, it exists in music, what would it be if it were to be there in materials? And so now you're getting a new node in the graph, which is not there yet? And it's not there, because it doesn't exist?  , fundamentally, it's just there, because we haven't discovered it yet. So this is sort of asking the question, right, so now you're beginning to discover new relationships in materials that are totally inspired by music. So you're gonna use this isomorphism as a way of building a foundation of footing, saying, here's some real connection, and I can understand how relationships look like and was sort of the mathematical foundation for it. And now I can ask sort of, like, we call a Taylor expansion of series expansion, or you could call it an extrapolation, if you wish, I can say, so if I'm comfortable with this connection, I can really understand this vigorous mathematically sound, I can now extrapolate from this, and I can go a little bit on the edges, like build new nodes, build new connections, and how would that look like, right, so you can now, make new discoveries, scientific discoveries, or, or do a lot of other things. And the beautiful thing about this, and I'll stop in a second, but, is that you can do this using, mathematical methods like embedding, but you can also do it with human language. And so that's sort of the beauty of language models, of course, they provide us with the connection between a very, very abstract representation of information and knowledge. But also, we can talk about it. So instead of formulating the math, writing down all the equations building the embedding problem, and doing the inverse problem, I can also simply ask the question, I can say, tell me what the extension would be if I were to follow the same topology, as in music, but build new notes. And this is sort of the interesting part, which makes it much more traceable, tractable, and flexible because you and I interacting with this AI model can be very creative. And then the human mind can sort of ask questions in our own language, we don't have to use just math, we can look at it, we can see it. But also, we can use math. So it's sort of the best of both worlds, I think.

Ross: So in a way, we could say, another phrase we could use is idea space. So we think of an idea space, and this sounds like you've already mapped, but by doing these mappings, we can actually start to find well that these are parts of the ideas space, which are not represented and as you say, we can actually discover what is in those ideas, spaces, which are found by for example, some of these isomorphic mappings, so we can actually in a structured approach to pull out some of the latent ideas, innovation discovered as structures. So this comes back to, I suppose, extending my previous questions around, humans plus AI. So clearly, clearly, your work is not trying to replace scientists. It is an amplification of scientists and how they work. And by, amongst other things, effective ideation or hypothesis generation back. So, this is I mean, it's a massive question. And I think this is an area for a lot more research, but what are some of the configurations of, human scientists and the kinds of AI structures that you are building? And how do they work together to accelerate scientific discovery as fast as possible? And what are some of the capabilities of the scientists in being able to use these models or structures effectively, and being able to discover things faster?

Markus: Yes, great, great question. Exactly. So it's really inspired by the, our MIT's campus here is really kind of a connected campus, we call it and one of the reasons why MIT has been so successful in science and engineering innovation is we have a lot of connections, random connections that can form so let's say a walk through we call the Infinite Corridor, all the departments and everything's connected here. So math and chemistry and engineering and all the buildings are connected socially, right? So if I've walked through the Infinite Corridor, which is the main artery of connections here, I might run into, scientists X, Y, and Z, and we might go to the coffee shop to have a conversation our students might need. And those random connections actually spur a lot of stimulation and discussion discovery. And that connection really is what makes discovery and innovation possible, do you kind of go and have a crazy idea, right, and you explore that? And so the idea sort of behind this, all of this work is how can we formalize this and actually supercharge this? So instead of us humans with a 1000? Faculty here? What 10,000 students? What if we had, AI that could help us make these connections faster with more data? And exactly, so this sort of, kind of drove this a little bit and, and so in a way, what we need is, we definitely need some point, let's say you have a new hypothesis generated or new, crazy idea, based on analyzing musical structures and extrapolating into materials and seeing well, I can maybe, the assets and tells me, I could make this really amazing electronic, electrically conductive spider silk based fiber, which might be used in a new computer chip. I mean, this is a weird idea, but I can dig deeper into this. So now, we need a sort of human-AI collaboration to come up with this interesting new idea, this new computer chip design, or this new material. And now I'm gonna have to make it right, so I'm gonna go to the lab, and I'm gonna have to try to build this material. And this is really important, because, the AI system is sort of extrapolating and thinking about what might be possible. But we need, of course, now, the grounding in physics, in experimentation, or in other theories, and part of that can be automated as well. So I can use multi-agent AI, which in some of the work have been using that quite heavily for that purpose. So you can say, create a new idea, a new design. But now let's test it out, run a simulation, write some code, run a simulation, or do an experiment, or look in the literature if this has been explored before. So that's kind of step, would be typically in the way we would do that, as humans, as scientists, my students, I would do this, we can automate part of that, but at some point, you're gonna have to, at least today, we're gonna have to supervise the process, and actually, probably still go to the lab ourselves and make this material. And this will provide feedback. 

And so one of the interesting things now, of course, is that, let's say you begin to build this material like we're doing this right now, for one of the materials designed in the paper, this mycelium composite was designed, we're going in the lab, we're building it, and so my student, she's actually, following the recipe that AI is generated, and we're gonna identify what are the shortcomings, like some of the things we've noticed is, some things are missing, that AI system has not really thought about every single part of it. So we need to either go back to the AI and say, Hey, I've tried this, but I'm missing a temperature, I'm missing the processing step. So one of the possibilities would be then the human going to the lab trying it out, going back to the AI saying, Hey, you missing something, tell me what I should do, right? Or are we just deciding on our own? So that's sort of the deciding point where you say, a decision point where you say, am I going to go just on my own to conventional signs in the lab? Or am I going to involve AI again? Or am I going to end the AI at the very end when I have made the material and I figured out the differences and here's my design, here's a picture of it, so you can kind of decide on what you want to do? But typically, you need feedback from the world. And you and this is part of what we think, general AI needs to do is to build, of course, better world models, which are usually referred to as models. So they need to understand much more about physics, especially as they extrapolate. And this is sort of similar to what science conventionally does, of course, if I'm done not using AI at all, I mean, I'm sitting here, my computer on my lap, and I'm coming up with a new idea, I'm going to make it or a new product, I'm going to try to build it, I'm going to do an analysis of the market and do an analysis of feasibility, the cost of all these things, and collecting new data. 

And so the way we're doing the same thing here, except that the ideas come from AI and the process comes from AI. But it can follow a similar process. And now, again, the good thing is that there are certain steps in this way that are going to be much faster, right? So for example, a few years ago, if I had a new protein design, I did not know what it looked like now with alpha fold, I can actually pretty quickly get an idea of how this protein will look like right now, I can then run a molecular dynamics simulation of this protein can actually, make it in a lab using high throughput processing. There are a lot of advances that come together that are individually small advances. But when you put them together in this whole ecosystem of general AI, high throughput science analysis, they become extremely powerful. And this is part of what the graph structure really is. Also, we started off the graph structure, it's about the connections. And science is all about connections, right? You kind of get an idea. You bounce off the idea by either physics and you say this is possible, or you say, as an engineer, yeah, it looks impossible. But I can maybe find a way of making it work somehow. And then you become an engineer, you figure out, okay, if I tweak the condition, I can actually achieve the school. And you get this feedback and you bounce off ideas. You can do this manually, or you can automate it. So multi-agent AI is the way we're doing this in a lot of the work we're doing is trying to automate this process sort of, beyond the idea beyond the discovery, which we talked about in this paper, we take those outputs and put them into agentic AI. And these have very deep capabilities, from generative modeling to physics to experimentation, perhaps even robotic experimentation. And I think that's a way we can ultimately really accelerate science, which the title of the paper kind of alludes to in this direction that we try and accelerate how we make discoveries and how we can prove whether discoveries or hypotheses work and how they can be falsified or verified. And all of this provides new information, which needs to be connected, like in a graph, to other pieces of information that provides knowledge. 

And ultimately, this gives us theories. And I want to make one more point about isomorphism. So isomorphisms, actually are very deep in their meaning, because they are ultimately the way by which we can understand generalization. So if you think about knowledge might be isolated in one domain, another main domain, and it found these graphs very, very rich, but they're not connected across. There's no literal connection between music and economics and materials, perhaps, right? However, isomorphisms provide us with a structural substrate to say, Okay, here's something there that's universal. And, in conventional science, we call this a foundational theory, or maybe, relativity theories or quantum mechanical theories. But we're far away from discovering these with AI, of course, but one day, maybe we can get there. But generically, those kinds of theories are very universally applicable, but other kinds of things we can, we believe formalize mathematically using isomorphisms. And that framework, I think, is going to be leading us into the future where we can actually make scientific discoveries meaning generalizable knowledge, that is not just true in one domain, but true in many different domains. And that's what science ultimately is about. Right? So we really want to kind of figure out how we can discover something that's true in many, many different settings and unifies different, different phenomena, across many different areas of observation.

Ross: So not ambitious at all.

Markus: Right? Well, yeah. I mean, yes, I mean, it is ambitious. And but I think that, there's always I mean, you have to envision kind of where you want to go, of course, and no, it's got to be a long, it's a long road to get there. But the vision, yeah, the vision is very clear. Yeah.

Ross: Yeah. Well, that's, I mean, we're six is sort of round out. I think, there's, there's not as many as I would like, but there's still quite a few others, like yourself who are looking at generative AI of scientific advancement, and I think, as a concept of structural level. So things like,  , alpha fold, and so on, are very specific. We're looking around the AI tools that assist us in, for example, ideation, hypothesis generation, being able to come up with novel methodologies, a whole array of, cognitive real roles for generative AI and augmented science. So what you're doing is ambitious, there's, there's others who are already doing it. So. And you have, I've got to, I should mention as well that you have shared code for everything you've done on an Apache 2.0 license. And so this is all out there to be used by people. But what's, what is it going to take for your work and those of your colleagues around the world doing related work to be adopted by the scientific community so that we see this fastest possible acceleration of scientific discovery?

Markus: Yeah, good question. Yeah. So one thing you mentioned is open source was open source, all the code and, and the methodology, and of course, the paper has lots of details in it as well. And the hope is exactly that. I mean, it will take adaptation to specific problems. I mean, I think in science, this paper was a methods paper, really talking about the methodology, the idea, a couple of examples in the paper and how we apply this but what it will take and what we're doing in some of the follow-up work now is to say, Okay, here's very specific use cases, scientific problems, engineering problems, for which we have not found a solution yet.  How can we use this methodology and actually find a solution? So the proof is always in fact, well, like in many things in life, solving the actual problem or doing something in the real world. instead of, doing an abstraction, so I think it will take a couple of successful cases where people have shown that, this ideation or hypothesis generation or Senate discovery can be done and it solves an actual real problem. Because ultimately, that's what people care about. And they don't care about the, ability of AI to do this, we care about solving, making lives better, or creating new economic opportunities and things like that. So, so we'll take adaptation for many, I hope, and we'll use cases that are useful. And they are sort of, maybe, bottlenecks in a sense of one that is computed to me. Now, this is a pretty expensive computational undertaking, no, you gotta mind these, fast way to get up, you have to have these papers, and so at a university, we have libraries that have access to these papers. But that's something you need. So if you work in a different space, you might or might not have access to the raw data of knowledge of information, slash knowledge, right, and to building these connections. But even if you have that, yeah, I mean, there's a lot of computers involved. And so there's sort of a limit in how fast can you do it? Well, can you do it? 

I mean, we're using a lot of language models and multimodal language models to do this. How good are they, I mean, if you build these ontological graphs, we actually let AI systems build them from scratch without any structure to them. That's the whole point, we want to really discover these structures natively from the data, instead of us creating an ontological framework and then letting it work on this, we don't do this, we actually get the model discovered. But that's where a lot of the frontier models reach the limitation, right? So we're kind of pushing the know saying, Well, the best AI models today, Claude 3.5, GPT-4,4o, and things like that, or maybe open source models that kind of can be used as well, how well are they able to do this, and, and their limits there. And then the graph reasoning as well. So once we have the graph, how good are the models actually, in understanding graph structures? And so we're really pushing the limits of what today's AI models can do. Now, the good news here is that, let's say models become faster, which they have been, they become leaner, like GPT-4 omni, much faster than GPT-4, before. But also, they're gonna have better capabilities. And so when these models become just 10-20%, better, and maybe have a slightly better ability to comprehend and reason and logically connect, it's going to be supercharged by the graph. Because we have an emergent system of interactions of multiple pieces of information and knowledge. In multi-agent AI, we were going to have multiple AIs talking to each other and communicating, if every one of them is just slightly better, the collective sum is going to be emergent. Much better actually. And so that's something we find as we sort of follow the evolution of AI systems that we use as the backbone. Any small advance in this field has a huge impact on the science that we can do with them. So there are these sorts of bottlenecks and opportunities. But I also want to say I think for the folks working on foundational AI models, I think this is really great. I hope it's great to see for them to say, hey, there's actually something that here's a use case, and I've had lots of discussions, actually, with folks that work on sort of the more really fundamental aspects of creating LLMs and multimodal LLMs, they can see that this is an avenue where their models are going to be extremely useful. 

It's also exciting for them to see and they might benefit from the shortcomings that are identified today to make better models that address some of those like graph reasoning abilities, or the ability to look at very long contracts, and lengths, that's a limitation. So there are a couple of, detailed things, technological issues that are bottlenecks. Yeah, but there's definitely, a nature to this, which is plug and play, which is, I think, very attractive, that you can actually you can substitute the and we did this in the paper, the cloud models and the open AI models, and you can compare them, but they all communicate with each other because they all work through human language, natural language. So they are compatible. But that's the beauty of using natural language as a way of communicating between agents in AI and humans if they can be combined. And so yeah, if tomorrow GPT-5 comes out, I can plug in GPT-5 and do graph reasoning with GPT-5, and that's going to be presumably, much more impressive in that case. Right. So this is, I think the way we can kind of leapfrog ourselves out of a lot of shortcomings is that we can basically plug in more capabilities, more capable models, or less capable models, if you have a computer limitation, right? Let's say you want to run this on your phone. Well, no problem. I can run it using a We've been using the Phi-3 model quite a bit for Microsoft has a very good model. Very small, with 3.4 billion parameters. Works really well for some use cases. Well, yeah, I can get it on my cell phone, right? So there's kind of, capabilities like this or, being able to use different quality models. Speculative decoding is an avenue we've implemented in some of our inference frameworks like Mr. RS, or we can use different quality models and size models to do the inference step. And so there's a couple of steps, I think, and these are practical considerations that are going to be useful in the future. Even if we have a PhD level model, maybe one day, we're still going to want to run it maybe on a phone write on an autonomous robot that runs in the lab and needs to understand how to do an experiment, but so that that robot might not be able to run GPT-5, plus GPT-6, so there's definitely trade-offs, but lots of engineering involved. So as you can tell from this discussion, I mean, there's a lot of engineering and how do we actually make it work in real situations, but so there's lots of really cool stuff for PhD students, for engineers, scientists to explore and ultimately create, hopefully, good products that people can use and benefit from.

Ross: Yeah, I think of it as the easy use of a relatively easy user interface. But I think part of the point is that you're already getting extremely interesting results. And what you're pointing to is, what it will take to get to the next generations of it, and I think it is an adoption issue as much as a technological issue. But, essentially, we are already already seeing just on the threshold of what is potentially quite an extraordinary acceleration of scientific discovery. So, thank you so much for your time, for sharing your insights very clearly, and for your incredible work. It's very exciting to see.

Markus: Thank you, Ross.

The post Markus Buehler on knowledge graphs for scientific discovery, isomorphic mappings, hypothesis generation, and graph reasoning (AC Ep54) appeared first on Humans + AI.

– Nichol Bradford

Nichol Bradford is Executive-in-Residence for AI + Human Enablement at The Society for Human Resource Management, focusing on human-AI collaboration. She is also Co-Founder and Partner of Niremia Collective, an early stage venture fund focused on human potential technologies, and Chairman and Co-founder of The Transformative Tech Lab, the largest global ecosystem of founders, investors and innovators building tech for human flourishing. She is also a frequent keynote speaker and Faculty at Singularity University, and has been a Lecturer and Adjunct Professor at Stanford University.

Websites:

www.nicholbradford.com

www.shrm.org/about/bio/nichol-bradford

LinkedIn: Nichol Bradford

Ross Dawson: Nichol, it's awesome to have you on the show.

Nichol Bradford: Thank you, Ross, I've been wanting to talk to you for a long time. So when you reached out, I was really thrilled.

Ross: Yeah, oh, it's very strong alignment with their messages in this, humans in AI and potential. So I'd love to ask you to give me your frame, and describe how you see humans in an AI world.

Nichol: So my overall interest in technology in general, not just AI, is how it supports human potential. And so for me, that's defined as people being healthy, happy, and really able to fulfill their purpose, to fulfill their potential. 

Specifically, I spent a decade so far looking at technology, specifically as it ties to what I call the ‘Human MESH’. So mental, emotional, social health, and human performance, and how we can leverage technology to support the Human MESH. And so there's a long line of technologies that have applications there. And I started one of the first communities dedicated to fostering companies in that area. AI is only the most recent entrant into technology that can allow us to heal, grow, and thrive.

Ross: That is awesome. This goes a little bit back to my book Living Networks, which came out in 2002. And at the time, if back in the 90s, everyone, you say, ‘oh, tech, that's for geeks sitting in basements’, and I'm saying, ‘well, no, that helps us connect to, to be more to think better.’ And other people didn't quite see it at the time. But I love the mental improvements around mental health, as well as the ability to think and the emotions. And, you know, there's been some things you know, it's not a one-way street, as in, there's some positive and negative potentials from technology, but the positive potential is so, so massive, and so wonderful to see you on that journey.

Nichol: Well, you have been ahead of your time, as well. And so how I followed you was initially seeing your work on just really sort of how to manage the cognitive stress of modern life, and then the way that you have thought about networks and other things. So I'd love to know, what is your definition of human potential?

Ross: So I don't have a nice acronym today or a structured one, but it's, it's who we can be. And this comes back to the becoming, you know, we are aware, you know, it's not just being versus doing, you know, it's about becoming, that is what it is to be human is to always be different. 

I often reflect that it's this paradox, we are the one person from when we are born to when we are teenagers, when we are older, we are one person yet, in fact, we are completely different people, all of the cells are different, the way that we think is different. So we are in the process of letting go of the old and embracing the new, and not enough people are too many people are static in their lives. But we are becoming more and more and I always think of it in terms of how we could be so many people, every one of us could live a hundred wonderfully different rich lives and discover what we could do. And so for example, I'm a bit of a repressed musician at the moment, you know, I think I have a lot of musical potential, but I've just been busy doing other things. And I want to come back to that. And there are many other things where I'm sure that there are talents and capabilities I have, and you have ones you were not even aware of. So the potential of everybody is vast. We can make choices around what we do to try to discover, what it is we are best at and that's in a way the human journey. So when we talk about technology to enable that, I mean, particularly when I see aI think ‘wow’, well, our ability to think about and imagine to enable us to do things being enabled by AI is just mind-boggling. So this is something we got to do.

Nichol: Yeah, I think also, you know, one of the things that I believe AI or the AI Age to make it broader than not  Because there's so many different types of AI, but the AI age, it's going to force us to finally answer some of the fundamental questions that humanity has been asking for a very long, like, who are we? Where are we going, like what is what are we doing collectively and individually, and also as AI, as Generative AI starts to reshape work, and starts to eat into the things that we call work, but might not really be work. And we could talk about that in a second. But as it starts to eat into that, and the what's left part, things that are going to make people stand out are going to be things like really being in your zone of genius, like really being in the thing that puts you on fire that you're so excited about. So you can bring that kind of energy, I'm really having a unique perspective. 

Yeah, having a truly unique perspective, really having developed your taste. Because in a world where almost anything can be generated, the ability to have taste is going to be very important, especially once everything gets generated, people start to follow curators, and communities with greater dedication, they're going to be following taste. So it's like having a unique point of view, having clear taste, and understanding what it takes to develop your taste, which means you have to co-pilot you cannot autopilot and have taste at the same time. 

And then also these things that people have called soft skills, that are about humans solving problems together in a better way. Really being good at those things. So that when the humans are in the room together, they're able to make really high-value decisions, taste statements, points of view to solve problems, whether that's the problem of a new market, a new product, or a new competitor. And so that's sort of like where compensation is going to go to and value is going to go to, and so it's going to force everyone to, to change. It's kind of exciting.

Ross: I'm in violent agreement with everything you're saying. Ever since I was young, I've believed that we are all far more unique than we know. And that basically we are all different. But society, we go to school, and we watch TV programs, we all get indoctrinated to be, we kind of look and feel and act or pretty similar, but that society heavily indoctrinates us. That's my school's job function is to make us fit in society. And so we're all you know, as I was talking before, that human potential, I mean, it's that direction of that potential and who we are is so different. We are so unique. And I absolutely agree that this is a time when our uniqueness has more value than ever before. The stochastic parrots of the generative AI, they are, they are homogenized. Now, they are averaging in a way that everybody says has a great value in that. But to complement the uniqueness of individuals. And so that's one thing which we need to do for ourselves. And how we structure organizations, I think that's one of the things that employers need to do is say, ‘All right, well, I'm not looking for a box to put a person in, I'm trying to find a person who can do things which nobody else ever does to complement the these tools.’

Nichol: Yeah. You know, there is, are you familiar with the dead internet point-of-view? So it basically says that today, more than at least half the traffic on the internet is bot traffic. Twitter is filled with bots, and there's a lot of activity on the public Internet. So they call that the dead internet, that it isn't as bouncy as we think it is. Because of what so much traffic is and then if you think by this time next year, or by the end of next year, pretty much everything you see online will be generated in whole or in part which means that it's either that almost everything is synthetic, or close to synthetic. And then it does things like, by the end of next year, if you have an ugly website, you just don't care. Like you really don't care, because there'll be too many tools to have beautiful websites with beautiful images. And so it was Ethan Malek who pointed this out. And I was like, ‘oh, boy’, was that, right now, only a small part of the world's data is on the public Internet, almost everything is private, everything isn't, you know, corporate intranets. And, and, you know, personal hard drives, and, you know, or personal drives, even if they're on the cloud. 

And so, it's this idea that if everybody in a company is co-piloting, and let's say half of them are auto-piloting, and if the generators or the LLMs, are really only your mediocre work, your average, then there's a certain point if a company is not paying attention, then they just become by definition, marginal. And so you could have, instead of the dead internet, you have a dead intra-net, you have a synthetic intra-net, you have a marginal thing, and it sort of forces you to redefine what is work, because right now a heck of a lot of work is someone checking, the weekly report that someone put together about what everybody was doing, and we call that work. 

But that's actually really great. It's okay for that to be marginal. That's like being averagely written doesn't matter at all, because that's not the important part. And so we're gonna have to redefine what we call work, because a lot of what we call work, and what managers use as proxy for work, was just sort of seeing if people were working, versus the stuff that actually really drives the business.

I'll give you one case, and then I'm going to ask you for a case that you've seen, but I was talking to someone who did an implementation with a grocery chain, that was finding that they were losing 80% of their new hires. And so he did a predictive model with them and included their values, and also human-decision factors and, and a bunch of things that he consulted. And so they redid their predictive model for how they sort of looked at applicants. And they were able to in about six months, flip that, that 80% loss into an 80% retention. And then a few months later, they started losing people, again, not the same amount. But they started having a noticeable loss of new talent. And because they targeted high performers. 

So a lot of times people think about when they think about AI, they're only thinking about corporate jobs or whatever. But like you can have a high performer in a warehouse, you can have a high performing grocery store, at a gas station, like a hardware store, like all of these things. And so they changed how they looked at people to pick high performers. And once they started hiring all those high performers reliably, they started losing them again, because their managers were used to managing bodies. After all, the grocery store used to be or the chain used to be like, ‘Do you have a pulse? Great’, and so they had a bunch of managers whose management skills were all around managing people with pulses, versus managing people who took a lot of pride, who wanted to finish a job, who wanted to solve problems on the floor of a grocery store. We have a second order problem that a lot of people are not tracking, that, you know, that's in sort of like the definition of work, and leadership and management and all of these things that's coming right behind once we clean up all this lower level stuff and use AI for it, then we're going to hit back at just the people part. It's still about people in the end, so I thought that was a fascinating case.

Ross: I think these are the shifts so it's not just oh, we plug in AI here. Then you've got to the same as you had before, except with a bit of AI in it, you know, it does reconfigure everything. And the thing about the humans plus AI workflows, more broadly than the changes in the structure of the organization, amongst other things, you know, who you need there. I think that's a, that's a really nice example for thinking through some of the implications.

Nichol: Yeah, I'd love to hear something that you come across in your travels, a case of an implementation where it went a little bit differently than people expected. And there were like of like some non-obvious thing about because everybody's at the all the people that you and I are interacting with, they're at the beginnings of their generative AI implementations, they might have done a bunch of predictive and a bunch of machine learning and that kind of thing, but like, in the front, and workflows, it's new. And so I'm just curious if there's anything you've seen.

Ross: Just take one example, in a financial advisory firm, where they were starting to use AI for basically supporting marketing efforts. So doing research, all social media posts, and things like that. And I think one level is sort of more productive and effective, you know, these are not core jobs of servicing clients. And that relatively small organization, so they're able to just take things off the plate and push that, but I think some of the findings from that were simply that some of the slightly higher order capabilities of the system started to be seen by, you know, advisors and so on, and saying, Well, okay, I can get in getting the feel for how they were used in sort of a fairly, very specific contained model, as in, you know, generate LinkedIn posts, for example, or consider how to think about different marketing segments. And then starting to extrapolate some of the use of that, in these contain cases and in terms of how they were thinking about investments and structuring and communicating to clients, and that, which is another use case for where I suppose those that learning from one quite contained specific thing to gain skills to apply that in other parts of the work.

Nichol: Yeah, it's exciting to see people trying things and coming up with new use cases. What are the things that we've really seen? Some of the research that I've read with Sherm, or that we've done with Sherm is that Society for Human Resource Management. I'm their executive and resident working on human and AI enablement, which is really exciting because they have 340,000 members who take care of 325 million workers and their families every week. So I love that ability or like that vector for the impact of really figuring out like, how do we actually bring AI into organizations in a human-centered way? 

But one of the things that we've seen is that in a lot of the early cases, where people are having success, the winning combination is a partnership between the CHRO and the CIO, where the two of them, link arms, and realize that they truly are peanut butter and jelly, that they truly, truly, truly make the difference. Because it allows the HR person to have the kind of fluency that allows them to only need to have a kind of fluency that allows them to be in the conversation, understand the basics, but they don't have to have two decades in IT to be effective. They've got their partner, and then what the, with the CIOs, who they've seen every wave from cloud to, like everything, and have a lot of lessons learned on implementation. What they're saying is that when they work closely with the CHROs to raise the AI flu wouldn't see, then that process they actually start to get the bubbled up use cases, from the the employees about like how their people within their business and their business model actually want to use it because like, what a manufacturer is going to do is going to be very different than what a services organization is going to do. And what a company and one market in the same category is going to do is different from someone who's working in Latin America. You know, and so there's, there's not going to be a cookie cutter, sort of set of like real workflows, when you get in the nitty gritty for differentiated companies. It's sort of like I think it was McKinsey did there, you know, make no, it was take, shape, or make as how people are using tools. And they didn't necessarily say this, because this is sort of a downstream effect. 

But everyone starts out taking, and you know, mixing things up. But if you stay in the taking category, and you don't shape, like build instances on top and do your own thing. So if you don't like, build on top of the, you know, the big utilities, really, then you have no differentiation from your competitors, zero, because they can all do the same thing. And so eventually, even people who start out, they have to at least move into shaping, to be able to competitively differentiate, and maybe move into making depending on how specialized something is, not making a foundation model, but making something specific to their business eventually. And it's not a good idea to run here until you know what you need. So you shouldn't run to make it, right? But there is like, start out by knowing that you at least have to move to shaping and get smart about figuring out what you might make if your business specifically requires that.

Ross: Yes. So I wouldn't want to come back to that. And perhaps you can give some high-level advice to leaders and people or organizations that want to learn about you. So you are an executive and resident side of human resource management, you are a keynote speaker, and you communicate. You have to understand this stuff, and you have a mission to help you and your potential. So what would describe more about what it is you are doing to change the world?

Nichol: Like, like, how did that happen?

Ross: That's about how you got here? Because I'm sure that will be well, that's another story. What are you doing to change the world? What's the day by day practice? What are you? How are you? How are you changing things?

Nichol: Well, there's one missing thing. I also have a, I'm a co-founder of a precede and seed stage venture fund. I have invested in technology, at the access of human potential and technology, specifically around the Human MESH. So like I am an investor in Apollo. Neuro, which uses haptics to reduce stress and anxiety, is clinically validated. I invest in neuro tech that is being used for Alzheimer's and stuff like that. So how it all sort of ties together as I've been in tech for 20 years. The first ten were in consumer-facing technology, specifically games, and the most known for operating World of Warcraft, China, and all the Blizzard properties for China. So I ran the game platform and in the market, and then the second half, was…the net of it is I went on a meditation retreat, I had a powerful experience, and I was like, everyone should have access to this. I've always been a believer in science and technology, as a way to make things affordable, accessible, and available. My father was a plumber. So I like humble beginnings. But that gives me a very democratic sense of wanting all of the good things, the things that allow people to heal, grow, and thrive. I have to be widely available so that people can choose. So I don't believe in forcing the same outcomes. But I do believe in creating opportunities for access. And there isn't anything like technology that can provide access. And so I think it's a matter of not necessarily, there's a lot of people who are very down on tech, and I'm not gonna I'm also, you know, I don't consider myself a tech optimist, in the sense that I don't think that there's some people that think if you just make the tech then magically, humans change. Like, that's, that's magical thinking, I think. And some very famous people in the Valley are tech optimists. That's not me. 

But I do believe that there is a category of technology that can support human, mental, emotional and social health. And I believe if we raise the floor with accessibility there, then we are more likely, as a species, to be better able to decide how we take care of one another. What society do we want to build, what is it we want to create? So I came back from China, and I started a community of founders and investors, and subject matter experts, clinicians, psychiatrists, psychologists, and coaches, who wanted to build these types of technologies. So that was in 2014. And in 2015, I had my first conference, which had about 250 people. in 2019, right before the pandemic, we had over 1000 people. And it was all builders. So it wasn't a consumer-facing conference, it was the mission or the mandate . If you want to build technology that allows human beings to become, and to have a better chance of becoming, if you feel passionate about building this kind of technology, come to our community. And we'll connect you and help you find each other and help you find others to build this kind of thing. And last summer, Sherm was looking at putting sort of a finer point on their AI, X activities. And when you talk to most people in AI, most people who have 20 years of tech haven't spent a lot of time thinking about what it means to be human. And how do you support the Human MESH? And most of the people who spent 20 years thinking about the Human MESH, don't know that much about technology. 

And so I'm one of the few people that actually like it right in the middle. You know, like I've seen, I've seen all of the early sensors, you know, I've seen all of them, like the early research on, on, what could actually be sent through heart rate variability, how that aligns to, like back when like, now you can get heart rate variability off of like your Logitech camera, can pick up Heart Rate Variability through the whites of your eyes, and through the heat map on your face. But it used to be you had to have this as an aura ring, you had to have a sensor to do it. So I've seen that whole arc and have sort of spent the last decade you know, looking at, you know, digging for all the different ways that one could leverage technology to improve the Human MESH. 

So another company I invested in is a company called Hola Biome, which is engineering, pre, and probiotics. And I find that fascinating, because the other thing that is the gift of technologies we're beginning to understand and to be able to witness we haven't quite gotten to, to, you know, breaking it down and knowing what to do, but the gut-brain access, you know, all of your serotonin, most of your serotonin is made in your gut. And so if your gut is off like you just can't be happy. You just can't. So, there's, you know, incredible research where people are, you know, where certain researchers like they've created these polymers, these filaments, where you can actually Put them in the gut and like actually see the connection to the brain. It's amazing. So in the next decade, our ability to support those in need to really help people, struggling with mental health to prevent it, we support people through emotional health, too and to really also start to understand how our social health affects our ability to collaborate, and then also to be healthy together. So that's my whole life. That's how it all ties together.

Ross: That's fantastic. Yeah, it's a few threads there. One is the democratization of power through technology. So I'm old enough to remember when desktop printing was a massive thing, you didn't need a printing press anymore. And there are all these layers of power to the individual, we all have access to these incredible technologies. It is a power of all of us, you know, and that's one of the biggest, biggest things with all these technologies, making sure these are all tools that everyone has access to, it's not just the few wealthy people that have access to it. That's one of the good things about AI. Generally, this is available. 

But we're back to you know, you're framing it is around, it's human first, human, how do we help humans? Well, we have technologies, we can create more technologies, there's all these possibilities for how we can be more. And it is fundamentally an attitude. And as you say, it's sitting at that intersection of this focus on humanity and understanding the science and technology and being able to pull that together is extraordinary, what the most important place to be, given the power of these, you know, the sciences and technologies we have today. 

But don't round out, I want one of them to distill as much as possible, your vast wisdom and insight to So alright, let's say we have a leader of an organization who is saying, we have technologies. And this is going to mean that organizations need to become more unique. To your earlier point, we can't be if we are the same as the other organizations, which are using the same models in the same ways, then we have no differentiation. And that's not going to take us anywhere. So what's your advice in this world of the technology we have today rather than the emergence of building a unique, distinctive, high-potential organization shifting from the past?

Nichol: Yeah, I would say that there are two. Two things are sort of, two things really have to change. There's two big efforts like that. So the first effort is sort of, how does one have a successful AI implementation, whether it's predictive slash machine learning or generative. And, you know, when you look at the history of predictive and machine learning implementations, They're notorious for failures. And then business transformation is notorious for failure. So you kind of have a double-double failure opportunity. 

And so it makes it worth looking at what caused success in the successes, what was present. One of the things that was present in the successes was a catalyst, a person or group of people that helped raise the overall firm knowledge, who helped the senior team identify the right projects. And so there's, there's pieces of that, but ultimately, companies are going to have to bring these technologies in. And to do it successfully, if history teaches us anything, then it needs to be human-centered. Because that's how you get the buy-in. The UPS implementation is one of the legendary ones, one of the most successful, and one of the hardest because it was one of the big first predictive implementations. But, you know, when you talk to that guy, it was like, it was all about the people part. The technology part was easy. You know, the hard thing was like getting the drivers to do it. You know, and then helping them understand how even if it was counterintuitive, it actually was better for them. 

You know, for their happiness, for their sanity, for them being finished on time, like the things that matter to them. So it still comes down to being human. So one, it's like, how do you have a successful implementation in a world where everything's going to include AI, it's going to be like electricity. And so that requires a human-centered change management for success. And then the subset of generative AI in that is that it looks like the use cases, for actually like pixelating and redefining the work bubble up. And in order to have that line manager really know, you know, or that product manager or that salesperson for their particular business. Really know what is needed.

So, for that bubble up to happen, and for it to mean anything, because, you know, asking ChatGPT, how to cook asparagus is useless, in a business context, have to have that bubble up, that really is a problem that you have, your people have to have a certain level of fluency. To be able to imagine what they might do with something they have to cross the threshold to where their first question is’, oh, could I use AI for that?’, and to try it, and for you to have the sandbox and governance in place that it's safe for them, and it doesn't expose you or your clients in any way. So, one part is actually like, bringing the technology and, and being successful at it. 

The second part speaks to the second-order effects that we talked about with the grocery store chain. On the other hand, it's like, actually doing something about culture. Like, in cultures, what are those things, it's like, smoke. Like, if you can see the smoke, then there's a fire, and you have a problem. But if you tried to touch it, and grab it in your hand, it is like grabbing smoke, so everybody knows it's important. They know it beats strategy, eats strategy for breakfast, they know all these things, but we don't know how to do anything with it. But to have a culture, a data culture, a learning culture, an evolution culture, a culture, where people actually become good managers. Like that, they actually know what that means, the ability to unlock someone else's potential. And so the second part, is culture, and all of the pieces, you know, those two things together is what gets you, that advantage machine that allows you to reset the board, or take advantage of a reset, you know, terrify your incumbents. Whomever it is like it's, it's that that's what a leader has to do, they have to do both. Now, they can start over here. But they have to go over here. But if they have the culture, then this part's easy. You probably looked at all that Accenture research on innovator cultures who invested in tech and people during the pandemic. And how those same companies are investing in Generative AI. Basically, they have an innovative culture. So easy for them to bring new technologies in. And people trust them when they bring new technologies in because they see them using them to evolve with the people, as opposed to just using them to replace them.

Ross: Yes, yes. Yeah. And I think that's, you know, the way I put it very, very crudely, leaders can have an attitude that technologies can cut costs, replace people, or they can augment people and do more for them and will give them better tap their potential. People working for a company will pretty much which, what the leaders’ intent is whatever they say, and will flock to and prosper the ones which have that attitude of this technology to make people better and support them. So Nichol, where can people go to find more of your work?

Nichol: So at NicholBradford.com, or also on SHRM,  our AI project. And there is one thing I want to say. And you can include it or not. But really with AI, it's AI in the nick of time. And the reason why is because of the dramatic decline in global population, that's going to happen in the neck in a decade from now, like, a decade from now, every, like, most of the western industrialized countries will look like Japan, in 10, to 15 years. 

And so, what all of this, so one AI is actually not very good. Like, you've been looking at tech for a long time, you know, it's not good. Like, it's not great. It's great at things, but it's not, it's not great enough to actually replace a human being. And so, by 2050, the only countries, there's only four countries in the world, they're all in Africa, who won't have a completely inverted population pyramid. It's only four countries, every other country in the world is moving towards looking like Japan. And so we actually need amazing AI to allow everyone to really fulfill their potential, and to create and to build and to innovate. Um, because our current economic models, like the whole way we calculate GDP, have an implicit population growth factor in it. It doesn't say population, but it talks about consumption growth. Well, young people consume expensive things. They consume education, they consume houses, and they do all the things that come with having kids. So with that, being forever changed, and in lots of countries historically have tried to buy their way out of it. They can't. No one's been ever, ever, ever able to break out of this who's tried. What that means is actually,, we need this technology, like, we need AI, we need to be very, very good. So that's something that I don't hear many people talking about.

Ross: Yeah. The population is between 2040 and 2070, globally, but as you say, that's only in a few pockets that are still increasing anyway, certainly one developed country in the world, which currently has more than a replacement fertility rate. So a lot of it is around immigration as well, which is another overlay on all of this. But I think a bit of the broader point is that we have unprecedented challenges, climate, population decline, social challenges, and so these tools, which if we use well, we can help to address these extraordinary challenges to create a prosperous future, which you are working hard to do.

Nichol: Well, I'm so happy to meet you face to face now. And I just love your work. I hope this is the beginning of a collaboration and thought exchange because I love the way you think about things.

Ross: Absolutely. Well, I'm honored to meet you and delighted that you're out there making a difference, because your attitude is a lot of what's going to shape a better world using technology. So thank you, Nichol.

Nichol: Thank you.

The post Nichol Bradford on AI + human potential, unique perspectives, and technology for mental, emotional, and social health (AC Ep53) appeared first on Humans + AI.

– George Pór

George Pór has been researching, teaching, and consulting in the arts and sciences of emergent collective intelligence since 1987, when he was introduced to the ideas by his mentor Doug Engelbart. He is the founder of numerous organizations, including Future HOW, Enlivening Edge, and Campus Evolve. His academic posts have included London School of Economics, INSEAD, UC Berkeley, Université de Paris, while his clients include European Commission, European Investment Bank, Ford, Greenpeace, Intel, Shell, Unilever, World Wildlife Foundation and many others.

Websites:

futurehow.site

ResearchGate Profile

www.riverflows.life

LinkedIn: George Pór

Medium: George Pór

Ross Dawson: George, it is wonderful to have you on the show.

So, I've known of your work for a very long time. I think, you know, probably 20 years or so. And I think similarly, you for mine, but there's been a lot of parallels. And recently you've been working on this idea of wisdom-focused, collaborative hybrid intelligence. That's a very intriguing phrase. I think it goes to a lot of these ideas of amplifying cognition. So please, can you explain to us what this means wisdom-focused, collaborative hybrid intelligence?

George: Okay, let me just step back to give you a little context. For those last two years since I've been diving into AI, my driving question was, and still is, how can AI augment collective intelligence to serve better the flourishing of people, organizations, and the human species? So that's the context from which wisdom guided and wisdom fostering collaborative hybrid intelligence comes. And so to get a sense of what I mean by wisdom-guided, collaborative hybrid intelligence, just think of that there are all of these zillions of organizations that prompt an AI agent to help with this or that aspect of decision making. The quality of that prompt has a huge impact on the AI’s output. Imagine if the articulation of the issue in the prompts would come from the deepest wisdom available to a decision-making individual or team. 

So what we are doing with AI in a meeting is analogous to what is happening in any good meeting. Even without AI, we are putting something out in the conversations, and individuals speaking, are contributing. And that becomes a prompt to the others to the other participants and brings back something from the others. So the quality of a team's collective wisdom depends on the mindfulness and heartfulness of our utterances, plus the depth of our listening to each other in the field. So what I'm saying is that when the mind, heart, and action of speaking come into alignment, then that collective wisdom can guide our interaction with our AI mates. So that's what I mean by wisdom-guided AI. So it's not just putting out any prompts for hoping that AI will come back with something that makes our processes more efficient, yes, AI can do that, but the higher state, the uncatchable advantage comes from people bringing their best into the definition, the articulation of the prompt that goes to the AI agent. Now, the other aspect of this wisdom-focused AI is that it can be not only wisdom guided, but also wisdom fostering, and what I mean by that is that too, to catch up to the capacities that the benefits that AI can provide. We humans need to bring our best wave and if we do that, then what the AI's output enables us is to tune in With the collective intelligence of the whole accumulated output of human knowledge. 

So, to catch up with that, we need to become more like AI whisperers, that is developing an intimate relationship with AI’s thinking. And that whole becoming wiser, for example, give you a specific example, like in one of our workshops, where we introduced this in our action research into the Collaborative Hybrid Intelligence, where we were not only talking about AI but actually used ChatGPT as one of the participants and Co-facilitator of the workshop. So how does it work? It's like, I already use ChatGPT, in the design of the workshop by asking some questions that may come up with a better design. And then in the workshop itself, I asked all the participants to bring with them their favorite AI agent, because they will need to consult with them during the workshop to further the process of what we are doing. So they did, and the question was, one of the overarching questions of the whole workshop was, Can AI help us become wiser? When a participant asked that question from her ChatGPT, she got a pretty detailed answer. It responded with "Yes, I can," then outlined in 5 or 6 points, the different ways in which it can contribute to making her a wiser person. When she asked the AI can I help you become wiser, the AI responded, yes, you can. And again, it outlined what kind of behavior, what qualities of prompts it expects that would help it give wiser responses. That was pretty cool. I found that just one example of "wisdom-focused," of what it means why it it important. As I said earlier, we can just go for intelligence, but intelligence is neutral in that it has no values intelligence can be used for nefarious purposes, as well as for  benevolant, beneficial purposes. And to use AI for omni-beneficial output, we need to bring to it our best qualities which are beyond intelligence, it is wisdom. Does it make any sense to you what I'm talking about?

Ross: Absolutely. No, that's fantastic. It was really, really inspiring. The mission, which you outlined at the very beginning of our conversation is extremely aligned with mine as well, I think, very strong alignment there. So there's a lot to unpack there. And perhaps the starting point is to get very specific, you mentioned a number of times, specific prompts, and also the qualities of prompts. Would love to just get liked and pulled back a little bit to the, I suppose, ways in which we can be wiser through AI. But can you talk about some of the specific prompts or types of prompts or qualities or prompts that enable us to get to these kinds of wisdom? Supporting interactions?

George: Yes, like, if you take a situation, give me a situation in which you would use a prompt in a group setting. Then, I will share what its "wisdom-focused Collaborative Hybrid Intelligence" version would look like.

Ross: So, I mean, one example is in, you know, boards and executive teams, what I will do is to say, you know, this is the current conversation, this is what we have been discussing, what has not been discussed, what else important has not been discussed, that we can add to the conversation. So, that's something that I might use in a facilitation context.

George: Beautiful, yes. These prompts already point in the direction of something wisdom related, because this domain is about looking at any situation, from a broader perspective. So, when you ask, what else is out there that was not discussed in this meeting? It points to the possibility that something is missing, that is broader.

Ross: There's no, one of my favorite quotes ever is from Gregory Bateson, who said that wisdom comes from multiple perspectives. And that certainly informed my own thinking and outlook.

George: Exactly. Right, right. So, you can ask that question from everybody, and using a shared document like a Google Doc or whatever else you are using, people will enter it like a brainstorming question. So, people will enter whatever comes to their mind, and if you would preface your question with a brain-stilling, stealing as in S-T-I-L-L, brain-stilling before brainstorming like something like why don't you take a couple of deep breaths and feel into what is really important to you that we have not yet touched on. Then. So, if you preface it with this, people are more likely to come not from the top of their mind, which is their intellect, but from a deeper space closer to wisdom, human wisdom, and what is important to them. And then when you connect the answers, the output of the brainstorming, then you can ask an AI to tell us what is a common pattern in these human responses. What is a pattern that connects with them? So then it gives something that you feed back to the participants, so it’s an interaction, a back-and-forth conversation, and with each turn of the conversation, the quality goes higher when it’s facilitated by the symbiosis of your intelligence that defines the process and the prompt and the AI support that is capable to summarize, integrate, and essentialize large volumes, even extremely large volumes of input from humans.

Ross: I can see that the iterative approach, I think is very powerful. And this this way you get the sequencing humans getting a response from the AI appropriately guided to give broader context or additional things and feeding back to the humans and then in turn, so I think that's what you're describing is this iteration between humans and AI and to be able to enhance and enrich the thinking, is that right?

George: Yes, absolutely. That was one example. But that's kind of typical. Yeah, iconic example.

Ross: So, just thinking about this frame of wisdom to how it is we can get the wisdom fostering. So wisdom guided is when humans guide the AI to hopefully be a wiser perspective and wisdom fostering is when the AI is supporting humans and their wisdom that is that correct?

George: Yes, yes. And you know, I designed action research for addressing the main question that my driving question, which is, how can AI augment collective intelligence to better serve the flourishing of people, organizations, and the human species, and that the action research, when I introduced it in the workshop, in the workshop, what I learned from those workshops is that to make it this high level, driving questions, to make it more engaging, the process has to be fun. So I turned the action research into a game, into a discovery game. In that game, players will climb a mountain range called the ‘three mountains’. The three mountains are the Presence Peak, the Sovereignty Summit, and the planetary Plateau. When the players are climbing the Presence Peak they are engaged in AI-augmented human development, because the challenges that they encounter during the climb happen in the Integral Matrix of the AI whisperer. 

So the more integral my thinking and behavior, the wiser, I become, by definition of what I mean by wisdom. So the Integral matrix of AI whisperers includes... Sol I don’t want to go into all the details just to tell you that the challenges that people need to meet during the game are what increase not only their horizontal skills, and horizontal development, but vertical development, which has to do with the higher consciousness and higher wisdom. So it’s through those specific challenges that AI can increase people’s capacity to address and look at everything and anything from the largest whole that they can put their arms around, the largest whole they can have an intimate relationship with. So AI is not only answering questions but also gifting the participants with challenges that can help them evolve.

Ross: So in this idea of the AI whisperer. So part of, of course, is your prompting techniques. But tell me a little bit more about what that looks like, if you are an AI whisperer to evoke or to draw out the wisdom of the AI or to be able to help it foster your wisdom. So, what are the behaviors or characteristics or you know, of an effective AI whisperer?

George: First of all, prompt engineering deals with only the right quadrants of the integral matrix, which are the objective things that you can learn in a prompt engineering course. But an AI whisperer is also paying attention to the left quadrants, which are the invisible, subjective things like his or her inner state when they are prompting the AI. They question who am I in this process of interacting with AI. So, it's an enhanced subjectivity that characterizes AI whisperers, but not only that, AI Whisperers are also knowledgeable and vibing (or maybe not by vibing with the lore, the culture of the community that they are serving. So this paying attention to the inner, the interior of themselves and the interior of the community that gives the capability of AI whisperers to have a more meaningful relationship with the AI agent.

Ross Dawson: Just about to say, I mean, there's an I think it was last year, I can't remember what it was, I was talking quite a bit about your relationship with AI, I'm trying to define it. And so I'd say, Well, what is your relationship with AI? You know, do you see it as a tool? Do you see it as a peer? Do you see it as a companion and that those very high level but to your point, part of a relationship is about who you are and how you are feeling if you are in a relationship with a person, you know, this is not just a description of it, it is you and that other and how they relate and just that consciousness of how it is we're relating to AI, and shifting the nature of, you know, how we conceive of our relationship with AI can potentially create quite different outcomes.

George: Yeah, just one level. Clarification that I do not think that AI is a person. But I do believe that we get the most productive, the most creative, the most enlightening conversations with an AI if we believe but if we take it as if it was a person. Not because it matters to AI, but it matters to our aptitude to relate.

Ross Dawson: You have to think that that's very important. It is a balance and Ethan Malek has written about this, in this idea that no, we shouldn't anthropomorphize AI. It is a machine. It is a tool that we have built, but it can be useful. But I suppose we also need to be careful as well in treating it as the equivalent of any conscious person.

George: The whole discourse about artificial general intelligence artificial super intelligence maybe I will never say never. However, I am more interested in AI augmenting human intelligence and AI-augmented collective.

Ross Dawson: So, to round out you just shared an article with me which we'll put in the show notes, which raised the idea of an AI shaman. And so, a long time ago, for a very long time, you've been calling yourself a techno shaman and now we have AI and you've introduced this idea of an AI shaman. So what is the role of an AI shaman and how does one become one?

George: Before somebody can become an AI shaman, the person needs to become an AI whisperer. So to develop that intimate relationship with an AI, but not all AI whisperers become AI shamans so the difference is that I would say that an AI shaman is not unlike the shamans, of the Ancient old times, that the task of the shaman as healer not only to heal individuals but heal the community, when it is all out of balance with the forces of life and an AI shaman does very much the same I mean, AI shamans, put their competence as a whisperer, they put it in service of healing the human collective at any scale. Healing the common human collective and it is out of balance with life forces when whatever narrow interests prevent life energy flow and enliven people and organizations then the AI shaman comes in to engage the power of AI to be part of a process of healing the community. So, there are not a lot of AI shamans. However one of my aspirations is to grow educational opportunities for AI whisperers who aspire to become AI shamans.

Ross: I think that's very powerful. I may, I may, I may come back to this. In another way, the potential threat is pretty amazing. The other goes to the point of this we are at a point that is absolutely transformative as humans have now something with which we can understand ourselves better through the nature of the way in which we interact with it. It changes who we are and it is a tool for positive transformation if we treat it the right way. So George, I love what you're doing, fascinating conversation and I hope you are able to have a far bigger impact. So where can people go to find out more about your work?

George: I published a couple of dozen mini-essays on Medium. There, my username is Technoshaman. There is also our website, the Future How, which is our action research, R&D organization.  I was am a member and advisor to River, the website of which is riverflows.life that is not a direct expression of my work; it's a teamwork that I'm advising. I am contributing with my wisdom-focused collaborative hybrid intelligence perspective.

Ross: We'll have links to all of that in the show notes. So thank you so much for your work and your energy and all you've done over the years. George, it's been fantastic to have you on the show.

George: Thank you for us, looking forward to continuing our conversation another time.

The post George Pór on wisdom-focused collaborative hybrid intelligence, AI whisperers, and AI shamans (AC Ep52) appeared first on Humans + AI.

– Pedro Uria Recio

Pedro Uria-Recio is a highly experienced analytics and AI executive. He was until recently the Chief Analytics and AI Officer at True Corporation, Thailand’s leading telecom company, and is about to announce his next position. He is also the author of the recently launched book Machines of Tomorrow: From AI Origins to Superintelligence & Posthumanity. He was previously a consultant at McKinsey and is on the Forbes Tech Council.

Websites:

www.machinesoftomorrow.ai

www.true.th

allmylinks.com/uriarecio

LinkedIn: www.linkedin.com/in/uriarecio

Medium: @uriarecio

YouTube: @uriarecio

Book: Machines of Tomorrow: From AI Origins to Superintelligence & Posthumanity

Book

Machines of Tomorrow: From AI Origins to Superintelligence & Posthumanity

Ross Dawson: It's wonderful to have you on the show, Pedro.

Pedro Uria Recio: Wonderful. Thank you, Ross. Thank you very much for inviting me. It’s a pleasure to be here with you.

Ross: So you've got a book, Machines of Tomorrow, which I think has a pretty vast scope, in terms of humanity and machines and where that might go on a pretty grand scale. But one of the central themes there is how humans and AI will be interlaced. And we'd love to just hear more about where you see that now, and how you see that evolving over the next years. 

Pedro: Wonderful. So in this book, Machines of Tomorrow, what I try to do is I try to explain artificial intelligence, from a human history point of view, from the moment in which artificial intelligence started to be created, or started to be designed, from those aspirations that humans had a very long time ago to create a copy of themselves a machine, like ourselves, to the present to 2024 with generative AI and what is happening right now with open AI, etc, etc. And also looking into the future, right? What is going to happen in the next few decades and in the very long term future, how it is gonna be, and how artificial intelligence is central to human history, right, particularly in the future? 

One of the aspects that is most important or most central in this book is the concept of interlacing. Which means that humans are going to interlace with artificial intelligence –- we are going to become more intimately related. At this moment, we will have our phones. And we are using our phones for everything –- we can call people that are far behind, or that are in other places; we use it for our daily life. The fact that the phone is outside your body is just an anecdote. In the future, it is going to be inside our bodies, right? It's going to be inseparable. And we're going to be interlaced with artificial intelligence, we're going to be interlaced with electronics, right?

And there are a lot of technologies that are being developed at this moment that are pointing in this direction. One of them is will all the cyborg technologies, possibly brain-computer interfaces have the most critical one, then we'll have robotics, then you have applications of AI to medicine and biology –- how we can be modified, where we live longer, we don't have cancer, we might see in the dark, et cetera, et cetera. So one of the aspects, not the only one, but one of the aspects of this book is that AI is going to change humanity into possibly a new species; we could call it a new species, a new form of humanity, which is different from what we have today. And that will happen in the long term. It is difficult to know where and how.

Ross: So what a start in the present. So of course, there are many phases to this. And I'm kind of interested to look at first of all, the next year or two, and sort of wait, so we're already in some ways interlaced. A lot of people are using these generative AI tools in particular, as part of the embedded into their thinking processes. They're already arguably interlaced into the thinking and ways of working. So let's start with the first next year or two. What do you think are the next steps? And then maybe the next sort of two to five years around? What are the next technologies in the way you see those panning out whether that be brain-computer interfaces or something else?

Pedro: It is difficult to predict what things are going to happen in the short term and long term, I tried to escape from giving dates in the book, but there are things that probably are going to happen in the shorter term, right? One, and let me cover a few of them, maybe a laundry list, but one of the things that is going to happen in the short term is how large language models are going to become more intelligent. And this is how these are some of the steps that a lot of researchers and startups are taking in creating artificial general intelligence because that's clearly the direction in which we are going in creating intelligence that is more intelligent than large language models. 

Large language models are not that intelligent, right? One critical aspect, which is counterintuitive, is providing these large language models with logic. Large language models cannot think even if you ask a simple mathematical or logic question to ChatGPT, they're very likely to give you the wrong answer, right? How can we make large language models split a complex problem into small pieces solve each one of those pieces first, and then apply those results together so that they get to a bigger answer? This is the same way we work right? This is how a programmer writes code. This is how a writer writes a book. This is how a pro engineer solves a problem right? 

Ross: So one of the key points, though, I find is that the current AI developments, as you say, logic has been one of the relative deficiencies of large language models. But now in fact, what we're just seeing in the last month or so, the new models coming out are building in logic through the use of multi-agent modeling, chain of thoughts, tree of thought, and a whole lot of algorithms laid over things which are making that logic better. 

However, this in a way is, again, looking to replace humans and saying, ‘Okay, well, humans are able to think logically, AI doesn't, so it'd be able to create it.’ And so in terms of interlacing, we want to be able to build logic systems where humans and AI are interlaced in that process of logic not simply being able to create chains of structures where the AI itself can fully undertake logic tasks. I mean, that may be part of it. But I mean, the more interesting part is where humans and AI are together, solving logical problems.

Pedro: This is an important aspect. To the extent that we are able to interlace more, to apply AI to ourselves, to become more intimately connected to AI, not only in use cases but also in our own body in our own existence. To the extent that we are more integrated with it, the more likely we are going to be successful, and the more likely we are going to be to survive. 

And this is one of the other polemic aspects of the book –- the idea that this interlacing might not be a bad thing, that this interlacing might be actually our way to continue evolving and continue going better and better and better. One of the examples that I mentioned in the book, and this is something that is polemic. The book is polemic. One of the examples that I mentioned is what happened to the Native Americans when the Europeans arrived, right? And then you have the example of Mexico, Mexico today has a large population, which is Native American, or mixed, very, very large population, 60% of the population of Mexico, over 60% of the population of Mexico is mixed. You go to the United States, nobody's mixed. Native Americans are a very, very small percentage of people. What happened? Well, in Mexico, they merged with the Europeans and they survived, and in the United States, they didn't merge with the Europeans and they didn't survive, right? And then you can judge that from an ethical perspective, from today's point of view, which will be historically wrong. And I don't want to get into that, but that's what happened. So if you compare that with AI, yeah, the more we integrate with it, the better it will be for us. 

Ross: I certainly take that premise. I've written about very similar ideas for a couple of decades. But what we're really interested in today is what are the mechanics of it. How specifically? Yeah, now, in the coming years, can we best integrate? Or what are the pathways at least?\

Pedro: There are a number of ideas. And this is something that is just happening. It’s not that I promote these, I am just an observer, I'm just observing the world and saying what I see, I'm just an observer. One of the things I tried to say in the book is, that I tried to escape from the idea of what is good and what is bad. I don't know what is good and what is bad, I'm just telling you what I see. 

But one of the things that is going to be quite relevant is brain-computer interfaces. Brain-computer interfaces, actually, are a way of connecting a brain that creates electric signals with a computer. And this is something that has been happening since the 90s. People who lost mobility through an accident, and then they were able to regain mobility through a brain-computer interface. And they were able to move a robotic arm, or they were able to move a whole robot or they were able to communicate telepathically with somebody else through a computer system, right? And that there are examples of people who have done that in the early 2000s. Any day and in the 90s. 

Now we have one of probably one of the most well-known startups that is working on this is Elon Musk’s  ‘neural link’, right? Elon Musk’s ‘neural link’ is precisely what is a brain-computer interface that has three kinds of them. There are some of them that are very invasive because you have to put electronics inside, literally inside your brain, others are external. And then you have a neural link that is somewhere in the middle in which they have to put an electrode in your head, but he's not in the brain, is in the surface of it? Well, Elon Musk will be the one who has the approval to do what he has done, if he just started doing tests with humans, with very polemic, there would be a lot of controversy. I don't know if this particular startup will succeed or not, but the reality is that that is an area of work that will continue. 

And then think about this, our brain is limited. One of the reasons is that our head has a volume and we cannot make it bigger, right? I mean, our intellectual capacity is limited in that you could offload a problem to a computer system, that is a super large computer, and get the solution back from it. Imagine that through a brain-computer interface, we could communicate with other people, and you could put the intelligence of multiple people together, working together on a problem. Imagine that you could operate through a brain-computer interface, another body, or a robotic body, I mean, all these take us into science fiction, and I have to use a lot of science fiction in the book to tell the story. 

However, I am also describing the real science of things that are working today. And one of the things that you'll realize is that actually, science fiction influences science a lot. And one of the reasons is that a lot of these entrepreneurs or scientists were reading science fiction when they were children. And Elon Musk is an example. Elon Musk has talked about Asimov Walker, in many of his interviews.  So yeah, I mean, it's difficult to read the future. You can make mistakes predicting the future. But yeah, there is a possibility that these kinds of technologies will continue evolving, and will take us to scenarios that we cannot really imagine today. 

Ross: I'm most interested in what we can get a clear pathway to today. One of the key points, of course, around the invasive brain-computer interfaces is that it's quite a long way until people who are not disabled will choose to have electrodes in their brain or whatever it said things in their brain. So the noninvasive brain BCIs are, at the moment, getting some quite interesting outcomes from looking at the short term, what we can do with noninvasive BCIs. And so this is where what are the practical applications in the next years for particularly noninvasive BCIs. And, you know, being able to interlace humans and AI…

Pedro: One of the applications that I have seen is communication. So people through electrodes can communicate, they can exchange words, right, at a slow speed at, a very, very, very low speed. But that is something that works, brain-computer interfaces have been used for that, and it only works right. And I think the difference between those that are more intrusive, and those that are external is the strength of the signal and the noise that you get from the brain. And well, I mean, what a lot of scientists are working on I'm not a biologist, I mean, I am a chief data officer, and I'm working on business applications of AI, I wrote a book that is taking a much wider scope, because I'm interested in this, but my daily work is not that but yeah, telepathy we will call it telepathy. So communicating without a mouth is one of the things that have been tested with non-intrusive brain-computer interfaces.

Ross: So looking at some of them, one of the things which you touch on is education. And so, you know, again, keeping that sort of shorter, potentially medium term, what are the ways in which you see humans plus AI in the education context? 

Pedro: Education has been interesting, because well, I have taught at university. And, I have talked to students that are from 23 to 30, I will say. So undergrads and postdocs. One of the things that I have realized is that there are a lot of people, a lot of students that I asked him to do a presentation with, I asked him to write something, they read something with AI, and you see clearly that they will do something with it. I mean, you see it clearly because you can feel they don't understand it. And then when they go to the whiteboard to explain to you, they read it, and then you say okay, so if you didn't write it, and now you're just reading it, what did you learn by doing this?

So I think AI is a sort of double-edged sword. It has two possibilities. It can be used for good. And it can be used for wrong. So when it comes to education, of course, you can use artificial intelligence to problem-solve a particular topic. And I did that in this book. Sometimes I didn't know what to write. And then I asked, ‘Hey, how can I say this? How can I call this chapter, et cetera, et cetera.’ And I think that can add a lot of value. You can use artificial intelligence to simulate environments for science students or mathematical students. You can also for mathematical students, you can use AI to find algorithms to solve a problem in a much more effective way than how a human will design that algorithm. So I know a friend who is working in a startup and his job is creating algorithms that are just much more efficient than what a mathematician would do. 

The problem with that is what will happen if we use AI in education as a substitute for thinking, and that is something you wouldn't want. You don't want AI coming back to what we have just discussed a few seconds ago. 

Ross: So what are the ways then? I mean, again, we want to be as specific as possible. So what are the…either for students or for educators? Or systems? What are the specific things that we can do to make AI more a tool of learning, as opposed to…

Pedro: I’ll give you an example about programming that is more in, in the work environment in the business environment in a company, and then maybe we can translate it to education, because I think it should work in exactly the same way, right? 

When you look at programmers today, or data scientists, until now, a lot of them were doing low-level work, designing algorithms, and coding part of the program themselves. And they started their career as a coder, we're coding things, right, that's how they started,  then they became a manager. And then suddenly, they took the responsibility to supervise another person who was doing the easy parts of it while you were still a coder. But you were also responsible for somebody else's work, and you had to supervise and be critical about that work. And then you became a hiring manager, and then you have a team of developers, and then you will not call in anymore that you have to make sense of what they call that you had to make sense that make sure that it was right, you have to make sure that there were no integration problems, you had to define a strategy you had to all that. 

Now, that is what in the short term AI is going to be for a lot of programmers, not only programmers, programmers, lawyers, data scientists, and accountants, they are going to be junior colleagues that you have to supervise. Right? So they are going to be that programmer, for example, a GitHub copilot is gonna be that programmer that is creating code, and you have to supervise it, but you have to make sure that it works, you have to make sure that it doesn't have cybersecurity problems, you have to make sure that is efficient, you have to make sure that it's really making what you want that piece of code to make without bugs. Now, the AI is going to help you in creating the code, supervising the code, or doing all simple tasks, but you are the one who is thinking, and you are the one who is in charge. 

Same thing with a lawyer. I was with a lawyer the other day, and he was telling me well AI is taking 70% of my work, and I'm worried about my junior employees because now I'm not going to need a junior lawyer anymore. Until now, there was kind of an agreement, right? So you allow these junior lawyers to make some mistakes, so that they learn and one day, they can become a partner of the firm, and they come here and they and they and they do that piece of work. But a lot of that you are not going to need any more right? Now, when we take it to education, what it means is that for a lot of people who are studying now, the first job is not going to be any more to be a programmer, it's going to be to be a manager. And those employees that they are going to have to manage that you'd have a pilot and we're going to have, we're going to have agents that are going to be AI agents collaborating with each other on software programs. The one is going to be making this model, the other one is going to be making the other module the other one is going to do in cybersecurity and you are going to have say 10 AI agents, hundreds of AI agents, all of them working on a software project. But some humans have to supervise it, that's what you're gonna be when you get out of university. I'm very fast, not like before I might tell you 10 years now he's going to be very fast. So, I believe that critical thinking is something that should not be outsourced to AI or to anybody for that purpose you should not also be if you should make your own decisions. So critical thinking is going to be very, very important. And I think this education system will promote it and not destroy that. Because otherwise, if you get into a dystopian view, uncovering dystopian views, as well, if you get into a dystopian world in which humans are not responsible for economic output, education has always been focused on making people ready for production, in an economy from an economical standpoint to be ready, so that the economy can keep running. If humans are no longer in charge of production or not, not necessary for production, and AI is the only thing that is necessary for production, then education will lose its purpose. And what remains is education as a tool for social conformity. Yeah, it has always been social conformity, to sustain and make whatever political system of the future. Those political stances will have been defeated, and we cannot even think about how they will be. So that people conform to those systems, and are integrated with them without creating too much trouble.

Ross: thank you, for your points here about how you describe the future of work. It's all by design, we make the choices, and we decide the way while we should be deciding how the the future of work unfolds. And I think part of the question is, what are the specific structures and architectures and roles for humans in these multi-agent systems, but crudely, what you described as a manager of all of these is, I think, a pretty, pretty good idea. 

So I wanna come back in a moment, just to round out by looking at, you know, what specific skills people can develop, and as well, but I've tried to keep it quiet, I suppose, grounded for now. But I mean, let's for a moment just go out with the big picture. And part of what you cover in the book is talking about how biology could change so perhaps just, perhaps discuss the big picture of how human or other biology might evolve in a world of AI?

Pedro: I have no idea. I mean, the difference is evolution, until now has been driven by the vehicle of reproduction, right? So you reproduce in that reproduction, whoever comes out is a little bit different, and some of them will survive, some of them will not. 

Now, the thing is, that nowadays, there might be another vehicle, which modifications will be made on your own. A word that will lead us is impossible to know, one of the technologies that I'm exploring sounds very science fiction, but it has been done with rats. One of the technologies that I discuss in the book is something that is called ‘mind emulation’. A mind emulation is basically scanning the structure of a part of a brain and reproducing that in a computer. This basically means that your brain would be running outside biology, which is a form of simulating your brain. 

But I mean, think about if this could be done at scale, with enough level of detail, you could have a brain simulated in a computer, you could have a form of immortality. And it sounds very strong. But there is a project called the Blue Brain Project that has been done in Switzerland, where they have tried to do that with rats, and they have been able to simulate the small parts of the brains of rats, obviously not at the level of at the level of detail that will allow you to keep a conscience or to keep, like, all all the magic of the brain, but it's an area of research and believe it or not, there are researchers doing that.

Ross: So let's round out by saying in this world where humans and AI will be interlaced we will be able to change who we are and what that future lies. So what are the skills we need to develop to prosper in that world? And how do we develop those skills? 

Pedro: So we go back to education. Because in those cases what you have to learn when you're a child. That's what really defines who you will be in the future the first years of your life. We talked about problem-solving and analytical skills, I think they are still super important. A lot of people say they are not, but I do believe they are. The second one is adaptability. And the third one is entrepreneurship. And that is going to be important. In the next few years, something is already important, but in the next 10 years is going to become even much more important –- adaptability. So AI is automating tasks and those tasks are jobs. So if you are starting your career doing one task accounting, and then for example, accounting, and then AI at some point automates all that, you don't need accountants anymore, you just need a Chief Financial Officer. Well, that person that goes in accounting, we'll have to find something else to do, because it's not going to get a job anymore and may start doing. I don't know maybe something about human-AI interfaces, imagine that is something that becomes quite odd. Now people are talking about prompt engineering, right? Whatever there is an opportunity in the market to find the kind of jobs well, and then that's that for a number of years. And then at some point that doesn't require prompt engineering. I mean, if you think about it, these people are talking about product engineering quite a lot as if it is going to be the next evolution quite simply. I mean, prompt engineering is not something that an intelligent person cannot learn in a weekend, or, or in a quite quite, quite short time, right? So then you have to change your job. And then imagine that you have to change jobs every few years. Well, a lot of people will be able to change their job, their job every few years. But a lot of people will not. And that's the problem because those people that will not be able to change jobs every few years that don't have this flexibility will be unemployable. Because they're not learning fast enough. 

Ross: So how do we move from not being adaptable to being more adaptable? What is that process? What do we do?

Pedro: It's about being curious. It's about learning. It's about having that idea that you have to remain in charge of your own future and that other people should not think that you're not entitled to be taken care of by others. I think that's quite important. It is about always trying to go the extra mile and it is very difficult. So it's a trait is a psychological trait but I think is quite important. 

And in line with this is entrepreneurship. And the idea is this. If you think about all the technologies, or all the technological revolutions that we have had: the IT revolution, industrial revolution, agriculture, the mechanization of agriculture, all these. If they all created more jobs than they have destroyed, we have many more jobs now than what happened years ago. But there is one, there are two reasons for this. And the first one is that those things were tools. And what we are discussing now is that AI is the first tool that has the potential to become an equal, which is what we call Artificial General Intelligence, will that happen or will not I don't know. But to the degree that happens, the ability of human beings to remain employable decreases, right? To the degree that that happens, the potential for having a net job described, destruction is bigger. And the second reason why AI might destroy jobs is the speed of change. If the speed of change is so fast, that people can just not adapt, they can not learn fast enough, they cannot find jobs fast enough. They start doing accounting, and they have to do another thing, maybe one or two years later, and then they just get stuck, they cannot do it. Some people will not be able to do all those years. 

So these are two reasons why AI in the midterm or long term might not lead to a net creation of jobs. Again, we don't know what the future is going to be in the short term. Yes, it is going to create a lot of jobs, many more jobs than will destroy but what about the long term? So that's why entrepreneurship is important because, in the future, a lot of people are going to be working not because they need but because they want and those are the entrepreneurs. Those are the people that say ‘Hey, you know, I have this vision in mind. This is what I want to do with my life and I'm gonna do it.’ And yes, it's possible that an AI could do it. I don't know, but I want to do it myself.’ 

There will be people in the future who will own assets and own factories, we don't think they will say, hey, yes, I know that an AI could be doing my job of managing all my assets, all my things, but I want to do it myself because it's mine. That idea of doing things because we want, which I call entrepreneurship, in a way, I think is going to become very, very important. And again, all these three things that I'm seeing are mindset. The first idea of critical thinking is, that I want to think for myself because I want it the second one is adaptability, which is I want to find my own way I want I will find a way I will adapt. And the third one is I do things because I want and I think that is quite important to remain relevant.

And then you can go to the tactics and the tactics will change every every couple of years. I mean, I do this for a while. And we're talking about my book. But my job is Chief Data Officer and I work with developers, data scientists, engineers, women, engineers, all these kinds of things. And I think these are the skills and these are certainly the skills of people that I hire. I want people who have these mindsets, rather than skills but mindsets.

Ross: Absolutely. So where can people find out more about your work, Pedro?

Pedro: So this book is available on Amazon. It is called Machines of Tomorrow. You can go there, you can find it on paper, you can find an electronic format, you can put my name on the internet or you can find a lot of things. Our website is machines of tomorrow.ai. And if you want to contact me, [email protected]. So quite the same. And if you Google my name, you will find all this. So now is quite easy.

Ross: Excellent. All right. Thank you for your time and your insights.

Pedro: Thank you very, very much for us. And it has been a pleasure. I hope this is useful and interesting to people who are listening to your blog. In some aspects, I try to go very long-term. And I try to go deep into the philosophy of all this but the book is full of details about what is happening right now, examples of current startups, and current scientists that sustain this idea right now. Will the future happen exactly? It certainly will not happen. It is impossible to predict the future. But I think it's a very plausible avenue for our future. 

Ross: Great. Thank you, Pedro.

Pedro: Thank you very much.

The post Pedro Uria-Recio on interlacing humans and AI, brain-computer interfaces, jobs to entrepreneurship, and enabling mindsets for the future (AC Ep50) appeared first on Humans + AI.

– Anita Williams Woolley

Anita Williams Woolley is the Associate Dean of Research and Professor of Organizational Behavior at Carnegie Mellon University’s Tepper School of Business. She received her doctorate from Harvard University, with subsequent research including seminal work on collective intelligence in teams, first published in Science. Her current work focuses on collective intelligence in human-computer collaboration, with projects funded by DARPA and the NSF, focusing on how AI enhances synchronous and asynchronous collaboration in distributed teams.

University Profile: Anita Williams Woolley

LinkedIn: Anita Williams Woolley

Google Scholar: Anita Williams Woolley

ResearchGate: Anita Williams Woolley

X: @awoolley95

Ross Dawson: Anita, it's wonderful to have you on the show.

Anita Williams Woolley: Thanks for having me.

Ross: So your work is absolutely fascinating. So I’d like to dive in as much as we can, in the time that we have. Much of your work is centered around collective intelligence and I’d love to just pull back to get that framing of collective intelligence relative to human intelligence. So we have some idea of artificial intelligence, which is emerging. So where does collective intelligence fit in that?

Anita: Yeah, well, it is. There are a lot of uses of the word intelligence. So it's good to get some clarity. I guess starting with the notion of individual general intelligence, which is the thing that's most familiar to most people, it's this notion that individuals have this underlying capability to perform across multiple domains. And that's what's been shown empirically, anyway.

So individual intelligence is a concept most people are familiar with. It refers to this. Well, when we're talking about general human intelligence, it's a general underlying ability for people to perform across many domains. And empirically, it's been shown that measures of individual intelligence predict somebody's performance over time. So it's a relatively stable attribute. For a long time, when we thought about intelligence and teams, we thought about it in terms of the total intelligence of the individual members combined, the aggregate intelligence.

But in our work, we kind of challenged that notion, by conducting studies that showed that there were some attributes of the collective the way the individuals coordinated their inputs, and worked together and amplified each other's inputs. That was not directly predictable from simply knowing the intelligence of the individual members. And so collective intelligence is the ability of a group to solve a wide range of problems. And it's something that also seems to be a stable collective ability. Now, of course, in teams and groups, you can change the individual members, and other things can happen that might alter the collective intelligence more readily than you could with an individual in terms of individual intelligence, but we do see that it is fairly stable over time and enables this, you know, greater capability. In some cases, at least, collective intelligence can be higher when you have a higher collective intelligence than a group that is more capable of solving more complex problems. 

And then, yeah, I guess you also asked about artificial intelligence, right? And so when computer scientists start working on ways to endow a machine with intelligence, what they are essentially doing is providing it with the ability to reason to take in information to perceive things, to kind of identify goals and priorities and to reason and to change and adapt based on information that it receives, which is something humans do quite naturally, so we don't really think about it. But without artificial intelligence, a machine only does what it's programmed to do. And that's it. And so it can do a lot of things that humans can't do even then, usually computations, or some variant of that. But with artificial intelligence, suddenly, a computer can make decisions and draw conclusions that are difficult for even their programmers to understand the basis of so that's where things get really interesting.

Ross: So we'll probably come back to that. So yeah, we're amplifying cognition, we're all about understanding the nature of cognition. And so one of the, I think, fascinating areas of your work is looking at memory, attention, and reasoning as fundamental elements of cognition. But being able to look at that not just as individual memory, or attention reading, but as collective memory, tensioning, and reasoning. So I'd love to just understand how this looks, what is collective memory, collective attention, and collective reasoning, and how do those play out into I suppose this aggregate cognition?

Anita: Yeah, I think it is an important question because again, just like we can intervene to improve collective intelligence, perhaps more readily, we may also…well, we know we can intervene to improve collective cognition and so Well, as you mentioned, memory and attention and reasoning are three essential functions that any intelligence system needs to satisfy needs to perform. Whether we're talking about humans or computers or human and computer systems or other sorts of biological systems. And so in when we talk about them and collectives, it's something that I say collectives because we often are thinking about the superset of humans and human and computer collaborations. 

But when we think about collective cognition, it's something that has been researched in parallel with the work on collective intelligence for a couple of decades now, probably the longest-standing area of research is on collective memory or one specific construct, there is transactive memory systems. And this is something that researchers, some of my colleagues at Carnegie Mellon, Linda Argote being a notable example, have done a lot of work on. And it's this notion that if you have a strong collective memory, a good transactive memory system, that the group can, can sort of remember, can use much more information and total than they could if they didn't have this well-constructed transactive memory system. And so, in essence, over time, individuals might specialize in what they're remembering. The group has a queue so that they know, who is sort of retaining different pieces of information so that they know they don't need to retain it themselves, but they know where to go and get it. And so as this system forms, again, as I mentioned, the total capacity of information that it can manage, can actually grow considerably. 

And so similarly, with transactive attention, we also have a total attentional capacity if we're working with a group on a problem. And so being able to coordinate, sort of where each person's focus is when our focus needs to come together. And we need to work synchronously when we should be dividing our attention across different tasks. And again, knowing who's focusing on what, so that you don't have redundancies, or gaps, or things like that, and that you can adapt as the situation changes.

Collective reasoning is an interesting one. And it's an area that actually has a lot of work, but it's been happening in different pockets. And so part of what we've been doing in our work on this topic is really to kind of pull together these different threads and use that as a basis for further understanding how this plays out. But collective reasoning at the foundation is really about goal setting. Because the foundation of a reasoning system is identifying when there's a gap between a desired state and a current state, and conceptualizing what needs to be done to close that gap. And so in collective reasoning, one of the fundamental hurdles is coming up with a shared understanding of what we're trying to do, and where we're trying to go. And then, you know, what are the priorities in terms of how we're gonna get there, what the interim goals might be, or maybe there are multiple goals that we're pursuing simultaneously. And so that's the kind of foundation of it and then adapting those, of course, over time to make sure that members' personal motivations are fulfilled, because if members aren't aligned on the goals, they're probably going to decide that their time is more valuably spent elsewhere like they could go elsewhere, and put in their effort towards something else that they find more rewarding or fulfilling. And so the foundation of collective reasoning is that kind of goal-setting and alignment process. 

Ross: Fabulous. I think the point around collective reasoning is, in a way, to a point role allocation, which can take us to sort of human and AI and role allocation and reasoning for that. But we'll come back to that. But one of the other things around attention, and essentially, the transformer models that underlie generative AI are they're founded on how it is they allocate attention, the self-attention models. But just looking at a human perspective, you’ve, also had some very interesting findings around the role of gender in collective intelligence and human groups. High level just interested in how gender roles play out, in collective attention.

Anita: Yeah, no, it's been an ongoing learning. So the relationship between gender composition and collective intelligence was not initially something we were focused on. So in our early studies, we observed a correlation between the proportion of women in a group and collective intelligence. And at first, we thought, oh, maybe that's spurious, or, maybe there's something about our sample or something. But it has continued to show up. And even in the meta-analysis that we published a couple of years ago, in PNAS that had over a thousand groups in our data set the correlation was still there. 

A lot of in a lot of the studies, though, it's at least partially explained and, and some cases fully mediated by other qualities of the members, specifically social perceptiveness, or the ability to pick up on subtle cues and draw inferences about what others are thinking or feeling or to anticipate how they might respond to something, and so on and use that information to help facilitate the work of the group. And so we've, in more recent years, done a series of studies to try to further unpack both how collective cognition forms, but also maybe the roles that attention might play. And so we do see that when we have groups with people who are more socially perceptive, not only do the conversational patterns tend to be more productive, and more, I guess, supportive of collective intelligence. Also we see a variety of other ways that it manifests in behavior, such as various forms of behavioral synchrony, facial expression, synchrony, vocal cue synchrony, but also synchronizing activity patterns together. 

So in some of our studies, we've observed this quality we call burstiness. Communication patterns where both teams that are together face to face as well as teams that are distributed even across the world, in some cases, when the teams that are more collectively intelligent, are better at sort of picking up on and being responsive to each other and concentrating their exchanges in shorter bursts. So you could have two teams that have communicated the same amount, but one team is much more bursty and has more concentrated exchanges of information. And another team in the other team isn't. The more bursty team is almost always more collectively intelligent. And so we found that to, you know, also correlate with gender composition of teams with social perceptiveness in teams. And to be something that is facilitated by actually having a leader having a stable hierarchy in the team seems to facilitate this process because all the individuals in the team will orient to the leader for the cues on okay, what are we doing now? Or are we talking now? Or are we working individually, etc. Whereas, in teams that don't have a stable leader, or especially in teams, where the individuals are competitive with each other to be in charge, you don't have that synchrony. There's much more competitive behavior, and a lot of interruption in the speaking patterns, and generally less collective intelligence.

Ross: Right. Well, this actually leads us, I think, to the roles of AI, and collective intelligence in facilitating human interaction. Now, obviously, there are many roles for AI in collective intelligence. But the first step is to say, well, I've got a bunch of humans who are demonstrating some degree of collective intelligence. So how can AI be used to facilitate human interaction in a way that supports collective intelligence? 

Anita: I think there are a variety of things we've looked at already, some that have worked, some that haven't. And then some that are studies underway. So, one thing we learned in our early studies is that it was easier for technology to mess up collective intelligence than to enhance it, in the sense that humans don't really want machines to tell them how to interact with each other. It's easier to overstep in that situation, and so we have a paper that came out, I believe, a little earlier this year, in MIS quarterly, looking at nudges. And so having technology-based nudges picking up on cues about how an interaction is going or how a team is performing, and then just nudging certain things. 

And an important thing about a nudge is that the human maintains their autonomy. That's a definition to a nudge, it's sort of increasing the probability of a particular decision, but not preventing any alternatives. And so, we have been developing behavioral indicators of collective intelligence in a team over time. And then on that basis, different suggestions could be formed. So for example, if a team is working together on a problem, and these members may not know each other very well. And it's clear to the system and the underlying pattern of things that there are people who know things that aren't, their knowledge isn't being used, or the wrong people are doing, you know, different parts of the task. And so a nudge in this is the main one that was successful in our study was kind of saying, ‘Well, why don't you pause a moment, and just step back and think about who's doing what and how you have things allocated, and whether you want to consider any changes.’ 

And so essentially, sort of nudging the team members, the human members to talk to each other, was by far the most successful attempt, anything we did that was heavier-handed, in some cases, backfired. People sort of withdrew or had different reactions to what the facilitators were doing. So one principle we kind of took from that was, that the things that AI would probably be the most useful for are things that might either facilitate or reinforce human collaboration with each other, not try to a term that we started using on one project, DARPA was joystick, not joystick, the whole thing where it's like, okay, now Ross, you say this, and then Anita, you say that, or whatever, the different levers might be, but rather to maybe start by reinforcing and getting the group to develop better collaboration together. And so that I think would be a really low-hanging fruit in a really good way to try to think about deploying AI. 

With generative AI, and specifically large language models. I mean, there are many exciting possibilities. And among them is the way that it could become something that's maybe a team, more of a teammate than a lot of human-computer systems human team interaction studies tend to see. And so if you think about, for example, a lot of us will ask Google things that we would never ask our teammate, right? If we expanded that notion and instead had one of these generative AI teammates, where not only maybe that I could ask that teammate something and feel less intimidated about being judged. What this teammate's opinion is of me. But you could also imagine how this teammate could facilitate the interaction of the human team members by, you know, sort of passing on information or prompting a conversation that might not have happened, you could even imagine if there was a conflict between two teammates, and they're talking to their AI teammate, if the AI teammate couldn't help facilitate some sort of a resolution, help each of them take a different perspective, for example, or do things to try to help heal this situation. So I think those are exciting opportunities as well.

Ross: Fantastic. So I think in a way two levels that you've laid out there, one is where clearly generative AI does an analysis of all of the communication patterns in order to be able to surface these nudges. And then, so all the communication patterns all going on, and then as a result of that, being able to suggest these various suggestions that might come up with so you've begun to identify some nudges that are successful. So presumably, over time with more data, we could refine or improve the ways in which, generative AI monitoring, as it were communication patterns could get better or better at being able to nudge thos things that would improve collective intelligence.

Anita: Yeah, absolutely. And I mean, I realize people I mean, there's, of course, flags about privacy all over this right, which is not a trivial concern. I've been surprised by how much.

So most of the analyses I was mentioning where we were doing this, we had, we didn't even touch the content of what people were saying, right? It was really based on patterns, who's contributing, who's not contributing, who's responding, who are we not hearing from at all? 

Ross: Or the wage level? 

Anita: Exactly. And so even that I realize is sensitive. But I think it's important to, you know, think about how, rather than taking all the information we possibly can just because we can, figuring out what we really need to be able to do something helpful? And how much does it really improve what we can do by invading people's privacy further? So I think that's really important, and it's in a very important question, obviously, that comes into all of this.

Ross: Yeah, well, I think that's, I think it's a great point where, in a way, even if you just separate out those layers of just looking at the metadata, I think communication patterns, and then communication content, and a few, and probably the vast majority of the value is at the pattern level as opposed to the content level.

Anita: That's right. That's right. And or if you can capture sentiment, which you can, you know, and just, you know, that's the only information that's used, not anything about the content.

Ross: So, you started to talk about the next thing, which is an AI teammate. I loved it. So I suppose that's another frame. And then the next phase beyond that is where you have what you describe as a multi-agent system with both human and AI, more participants, let's call them. And so from looking at, I suppose, looking at the human communication patterns and finding nudges to be able to bring in an AI participant, or single AI participant, which as you say, can play a facilitative role or can play other roles in information brokering or other things which can be designed to enhance a team capabilities. So then what happens? Where are we in terms of looking at those next steps where we look at multi human, multi AI, participants in a collectively intelligent system?

Anita:I think, certainly these things are being modeled or simulated. I think that that, to me feels like just kind of a, I don't know, an explosion, I guess. But I was gonna say, an expansion on what I was talking about with the one agent, maybe as part of a three-person team just in our little hypothetical example. But you could easily imagine that there could be different agents playing different roles. Even now, there are, you know, settings where some studies and even exercises in the classroom that we do have agents that are just content-based, and there are others that are watching the process and intervening in the process of what's happening. 

You could have, especially when we start talking about issues of trust, it could be that each person in the team has their own agent, almost like their Jimny Cricket, if you will, who is there, you know, sort of like assisting them and maybe talking with the other agents to figure out how we're going to help this poor team? Yeah, I mean, it kind of expands considerably. And I think the other opportunity there is, a lot of times we pull together diverse teams in order to get a variety of perspectives. But there's also an exciting opportunity potentially, to be able to get those perspectives through agents, right? If we know the right kind of characteristics or backgrounds that we need to simulate. And we know that we can faithfully do that, you know, that also provides sort of a mechanism for pulling in those perspectives and integrating them so that perhaps the solutions we come up with would be better. Now, that doesn't mean it gets rid of the need for diversity in the team, because I think the human teammates are going to have different ways that they're going to draw things out of the artificial intelligence, right? And so you still benefit from having diversity within the team, but then you have access to even greater diversity. So I think that is also exciting when we talk about solving very complex problems, especially all the multifaceted problems that are the toughest, like climate change, and so on. 

Ross: Absolutely. That's, that's one of the ways in which I use generative AI is saying, ‘Alright, well, we've got these perspectives so far, what perspectives are missing?’ And they can find some perspectives, which haven't been brought to bear yet.

Anita: Yeah, absolutely. And I think leveraging that and identifying, Okay, how about if I had this set of values? Or if I was kind of delineating what the things are for this particular problem that might really change how people perceive a particular solution.

Ross: So one of the phrases uses this idea of integrated collaboration between humans and AI. So what does that speak to? How do we truly integrate that collaboration between humans and AI? 

Anita: I think it's happening little by little even now, I mean, in very small ways, you know, when different assistant things pop up, and remind me or whatever helps us do something that we couldn't do otherwise. We are working on a project right now through the NSF AI institute called AI caring, which is focused on the problem of helping the elderly age in place. 

The portion of it that my team is working on is focusing on how we set up robust teaming in the caregiver networks that are involved, who often don't even know that they're part of a team. They're all the people that help this, say, this elderly person with mild cognitive impairment. It's a neighbor, it's a nurse, it's a, you know, family member, etc. But there's always, if any, if anybody had to help a family member or somebody who's in the stage of their life, there are all kinds of pieces of information that different people will know, and they should be passed on. But there's no really good way to do it. 

And so what we're working on is a tool that could, maybe even be on somebody's cell phone, where, ‘Okay, I'm a neighbor, and I'm taking, you know, Mr. Dawson out to lunch today, just to give his wife a break.’ And then, you know, this, this agent would maybe pass on some things to me that I should know about, you know, he's been not steady on his feet, he might fall over, or he needs to make sure he takes this medicine before he eats, etc. And then maybe during my interaction, maybe there are some things I'm concerned about, and I can let our helper agent know. And then this agent can pass that along and can help coordinate or maybe even flag something in a pattern that is coming up that different people are noting, but nobody is connecting. And so I think bringing the tools that we have into our networks to help humans coordinate and pass along information is going to be one of the key ways that we can benefit, I think from really integrating the human and the AI, you know, capability. Essentially. 

Ross: That's a lovely application qne illustration of it. I think very, very grounded and very, very human. So, I like to pull back to the very big picture, rather than collective intelligence. I remember stuff happening in the 90s. And they're in collective intelligence, there's a lot of, I suppose, looking at the architectures of collective intelligence, which have progressed quite a lot over the decades. We know of course, as AI is something which we can, play a role, be a, you know, human peer equivalent as we look at peer roles, as well as some of how we can bring these together. And, of course, you know, a lot of research from you and your colleagues, obviously, the MIT Center for Collective Intelligence, many researchers all over the world. So, I'd like to just sort of pull back to reflect on where we've got to and where we are. What are the frontiers in coming years and collective intelligence?

Anita: Well, I just see the opportunities for a whole variety of ways that the new capabilities of artificial intelligence can amplify and help amplify the abilities of the humans and so one way would be we talked a little bit about being an intermediary in a conflict but even in a conversation. So, I don't know what it's like in your part of the world now, but for us, everybody's so busy, and there are so many different things happening at the same time, it's even hard to have a meeting. And we've tried to have meetings too often because we don't know any better way to collaborate. And we could imagine, actually, maybe there's a time coming when we don't need to have a meeting. 

We have, part of our teammates, these agents who go talk with each person and get their perspective, and, and portray it for others and get their perspective and integrate it and kind of asynchronously help us have this discussion, right, where, and probably do it better than we do for each other — they could be great listeners, they could ask great questions, they could, you know, do all kinds of things that we know, humans can be not great at, especially if there are other things in the situation where they're anxious, or, they're in conflict, or whatever the case might be. I mean, I think that, and probably, you know, other aspects of technology will develop too. And so we won't be sitting in front of a screen and looking at a camera, we'll be, I'll just be sitting in my chair, and you'll be sitting in the chair across for me and we'll be having these interactions without having to synchronize for them entirely. So, that's kind of where I see things going, I hope that they go in a way that, again, kind of strengthens and reinforces, you know, human capability and human connection, versus completely replacing it. But, there's always a danger of that, of course.

Ross: Yes, but I think the intent is what is going to bring us to the, you know, truly human-centered approach where AI compliments us, as opposed to replacing us and that's all it's all how we choose to go about it.

Anita: Yeah. Yes, totally agree. So I hope everybody's intentions can be in the right place.

Ross: Thank you so much for your time your insight and all of your work I need. I think it's extremely important and has taken us to better places. 

Anita: Well, I hope so too. Thanks. Thanks for your questions. I really enjoyed our conversation.

The post Anita Williams Woolley on factors in collective intelligence, AI to nudge collaboration, AI caring for elderly, and AI to strengthen human capability (AC Ep49) appeared first on Humans + AI.

– Jeremy Somers

Jeremy Somers is Founder and Director of AI-assisted creative agency NotContent.ai, and of We Are Handsome. He has extensive experience as a Creative Director, working for brands such as Asos, Canon, Mercedes-Benz, Qantas, Spotify, and W Hotels.

Websites: 

www.notcontent.ai

www.jeremysomers.com

Instagram: @notcontent.ai

Beehiv: notcontent.beehiiv

Ross Dawson: Jeremy, it’s awesome to have you on the show.

Jeremy Somers: Hey, Ross, thank you for having me.

Ross: So you're a leader in AI-assisted creative agency work. Tell me more. Tell us more. 

Jeremy: The story begins long before the world of generative AI and AI creativity. My career and life history have always been about creativity. And I started in traditional analog photography, when I was in my teens, and trends went through the whole transition into digital photography. And then I taught myself graphic design. And then I learned it through a very, very early Photoshop version on a bubble, iMac, and the colored ones. And then started working in some of the very first digital agencies in Sydney. And learning through the transition of like, there was no social media and other social media, there is no e-commerce now there is e-commerce, so it's in digital agencies working on big brands, Nike, and Pepsi, and Microsoft, Samsung, etcetera, etcetera, through this whole transition. And so a lot of my career journey has been in transitional periods of, like, massive shifts in the thing that I'm doing, not just the tools that are available to us, but just societal level shifts of how we communicate as designers and creators and branding people to the outside world. 

And I happened upon open APIs, Darley white paper very early on, probably coming up on, two and a half years since it was released, I think I'll check that. But I haven't found this like paper and nerdily, read through the entire thing, and then read through it again. And then I fully understood what was going on, I had this moment of, sort of cinematic-like, flashback, flash forward moment of, I see the end result of where everything I've ever done, creatively, how I've done, it has changed, but this is going to change everything in a way, which we've never seen before. So I have this, Pivotal epiphany. And I was like, whoa, okay, how can I learn more? One, and two, once I was able to learn more, and you know, so your generative AI suddenly became a thing. I was just like, rabid for learning and looking at tools and learning about who's doing what and how to kind of get access to it as a creative and as an agency owner, and I happened into the right places at the exact right time. And did a whole bunch of testing things and playing around and just like nerding out on stuff and taught me a whole bunch of new skills and the new taxonomy and way of thinking, and then I thought, okay, how can I take all of this time that I'm spending and turn it into something commercially viable? And I see this end result? 

We're not there yet. The technology is not there yet. The people are not there. we're so, so early on all of this stuff. But how can I, if I can translate it into some sort of commercial vehicle now? And then we're talking two years ago, I'll set myself and be way ahead. I've seen all of these massive other shifts, and I recognize this is the start of a shift. And I was never early on anything else. So maybe I could be early on this one. That's how we get to notcontent.AI is one of one of the world's first creative agencies, today's AI assistant.

Ross: So, AI-assisted creatives. Let's dig into that. So I mean, you've been talking about image generation, of course. There are other forms of communication, occurring, words and videos and smells and all sorts of things. So let's have a look at the high level, and perhaps you can sort of dig down into detail. So what does that mean, when you've got creatives as in presumably creative humans working with tools, and how together they're creating something better, faster, cheaper, more superlatives in whatever way? 

Jeremy: That’s it. So along this journey over the last two years, one of the recurring themes that we've seen out in the public through mainstream media, even through our social platforms, and especially like places like LinkedIn is like, oh, no, all of the creatives are going to lose their jobs because these tools are doing their jobs. And that's true to a certain degree. But one thing that I've really learned having done this, is that I actually think that true creativity, which comes from humans, because it comes from our specific individual experience of life, and the things that I've taken in are the things that are my output, that is not going to be replicated by AI, to the degree that humans are able to do it. Ai is an open slate, it's like it's an encyclopedia of everything. And so if it knows everything, it knows nothing until you ask it something super specific, but true creativity and the true creatives and the way that we're able to think based on my individual inputs, I think the true creatives are going to be more valuable moving forward, not less, and maybe to an amplification of, you know, 10x 100x. 

And I realized that because what I was seeing as gendered AI became more accessible to call it the masses, there was no creativity. We were suddenly given a series of tools that allowed our imaginations to run absolutely wild, and do anything that we wanted. And as a group, what we did was the same exact thing that everyone else was doing. We were all able to see the output because of social media, and the platforms that existed, even some of the tools and platforms themselves. I was like, oh, as a group, we're not very creative. This is why I have, you know, sort of a long and storied career spanning multiple different industries that are widely separate from each other from fashion to like, lighting design, for rock concerts, to graphic design, to photography, because the thing that kept me going was my ability to think creatively and have the output be agnostic, but it doesn't matter what the output is.  I just needed to learn a series of tools to be able to design lighting for a rock concert. But the true creativity came with me sitting with a piece of paper. 

So true creators are going to be the ones that stick. And that's why I say AI-assisted creative, whenever I'm talking about this, because the AI is my assistant, the AI is whoever's working with the studio, the agency, the AR is our assistant in a whole bunch of ways. And as you allude to, not just image generation, and now video generation words and too much more intense degree, as we've been doing over the last six months. From a strategy point of view, so much creative work that we do is strategy-based and needs a strategic baseline to be able to produce really interesting ideas. And then, the AI is really good at taking the grunt work of the production. 

Ross: I’d like to dig into that. What I describe as humans plus AI workflow, where humans and AI are both elements in a sequence or network of things which create a wonderful output. So let's talk about strategy. So, unpack that, where does the human do? What does the AI do? What's the process?

Jeremy: So let me give you the overarching, like, finding that I, that I really figured out. Of course, the regular 80/20 rule that applies to everything in life applies here as well. And I was really like, how does it apply, then I figured out it actually applies backward. Based on what I've just said about humans and creativity, right? Humans are doing 10% of the work upfront. And especially like for us internally 10% of work at the end, in terms of finishing brand integration for campaigns, checking over everyone's work, the AI is what gets in the middle 80% is done by the AI. And that's 80%. However, the whole thing is flipped around. Because the 20% that the humans are doing is actually 80% of the meaningful work that is output at the end, right, so it's 8020. And then it's flipped around to be at 20 Again but in favor of the humans. 

What that looks like for us on any creative job that we may do, from a strategy lead job through to the file outputs through to just like a client's like, you know, monthly content to where we've already done the strategy where they brought strategy from another agency or another creative or in house, whatever it happens to be, is that those, that process of three steps, human AI human is set all the time. And for us, that looks like sitting with a piece of paper away from the computer for that first 10%. Knowing how to use the tools is really key, humans have to know how to use the tools, right, that's going to become increasingly less important, which is the point of the AI tools, right? And being able to talk in natural language and you know, going from GPT to GPT4o and being able to just have a conversation, we're seeing that in real-time, right? Communication has been getting easier with AI for the last two years since we've been doing it. We've gone from learning how to prompt for every single individual model or tool or whatever happens to be having a lot of learning there to spending a lot less time on that. But it's still we need a workflow and figuring out whatever the creative job is so that comes into the studio, figure out what the workflow is to even get the AI to do its 80% of the grunt work for production on that. 

Because at the moment, it's not a single tool. For any job, we use probably half a dozen to a dozen tools on any particular job. A lot of them are AI, yes, a lot of them traditional, you know, there are things that I just cannot replace at the minute. And then both human school human skills and tools, 3d models being one of them, right, like, you know, this year at some point, by the end of the year, we'll see a relatively good text to 3d model, AI model. However, we're nowhere near there yet. So we still need 3d models, we still need retouches, we still need brand integration, we still need all of these things. So the first thing that we do for that first 10% is sit down and plan out what the workflow looks like, for the fastest route to the end of this job. What AI tools of the suite of many 1000s of them that we all have access to? is going to be the best? And how do we also amplify creativity? How do we do the best thing strategically for the client? So that requires that 10%? Once you have that figured out, then it's a matter of just getting into the machine and letting the machine do its thing that you've asked it to do. And then the other end is just finishing and making sure that we all have our eyes on it. Is this what we said we were doing? Is this brief? Is this on brand, etc, etc? How can I dive into more detail for your OS?

Ross: I would imagine, though, then in the middle that there is, it's not just a pre-framing, and then AI does the job then you've tidy up at the end. Because there's a lot of curation, in the middle so for example 10 different versions, and I like that one, and let's iterate on that one. So there's a lot of presumably curation and iteration or you tell me what else is happening in that middle?

Jeremy: Yeah, there's a lot of guiding, right? And within that middle 80% that the AI is doing, we're switching tools back and forth, back and forth. So a good example, if you will, like a real-world example is there's a website called inc file, which is just rebranded to be called busy. It's a US website that lets you register your LLC, in any state that you want, does it all for you provides accounting, and like anything for your startup business, they will do you know, the paperwork and the registrations, and there'll be a virtual mailbox, that sort of thing, right? Pretty dry. So they brought us in to produce mountains of content, the amount of content that they produce to be like the top of Google, right in terms of long-form blog content and their social. And that sort of thing is just and they've been doing it for a long time, is just massive, and they wanted to figure out a way to stop using stock photography, because it all looks the same, right? And try to use AI tools and AI-driven content to form this for them. From a visual perspective, at the moment, they still have a team of writers who are doing all of this and doing their written content, they're not quite ready to skip over to AI-assisted written content. But from a visual perspective, they're like, can you design us a system, a scalable system, not just can you give us some content that looks like humans doing, you know, entrepreneurial work, they're very story driven, the founder of the company is like obsessed, I think he used to work at Nike. And he's obsessed with just like the storytelling of how Nike runs.

Now, he's got this very, very dry, boring, you know, SAS product, which he wants to apply branding and strategy and storytelling in a very Nike style. And so what we do is concept a way of not only producing the end content and making sure that that's going to be great for them moving forward from a visual perspective, how can we use AI to amplify that, to make sure that it is one outputting more than they've ever seen and the amount of content that they need? And that it's at the Brand level? And it's the quality that is required? But two, how do we make it scalable? How do we make it so that when you have that 10% As a human, you're concentrating on the real creative element, and everything else is almost automated? Alright, and so we trained a series of GPTs to tell stories about entrepreneurs, real-world stories, real-world right stories that mimic the real world about entrepreneurs. And each entrepreneur was given a name and a backstory because as an entrepreneur and as a business person, you have so much more going on than just what your businesses and so we trained the GPTs to help us create stories about people and make them feel as real as possible.

We tried a series where we go in and I can go in now and say, I don't know what my, you know, I like setting up like starting keywords for this stuff. And it says GPT go Whatever it is, right? Or it says inc file go. And it prompts a couple of questions for, hey, here's the character we're creating today. It's an awesome question. But then if I bring, I can bring in too much information or as little information as I wanted to give it some basics so that we're not just doing the same thing all the time. But it now has to, it knows to go away and give that character a name and a backstory, where they live et cetera, et cetera. And then it checks it with me and gives me their bio and a backstory to write to their LinkedIn bio for them. And just like, we're trying to create a person that mimics someone in the real world or with all the intricacies and all that sort of thing we asked for, it knows to give us certain personality traits. And it's being creative and making a bunch of these things up based on the training that we've given it. We did all this manually for a bunch and then gave it all the information, and then sent it out to look at different places on the internet and different writing styles. So we train all these separate things and pull them together into this one model. And then it goes, Okay, cool. Do we love this and we craft it, we go back and forth a little bit.

And then we say, Okay, give it to me. Then it starts with write-ups and visual prompts. And it gives us all of those your prompting for a full series at the moment we do about 50 50-60 images per character, of basically a day in that character's life of being an entrepreneur, where they live, you know, the time of year that it is their type of customers, you know, we have things that are like cool, we want to see them working at a computer for 35% of it, because busy is a technology company. And it's about how we want to see them on their phone, because you can be busy on your phone, right? But we've given these parameters, we want to see them with other people, we want to make sure that we're hitting diversity across everyone, not only with the type of people we're creating but the type of people that they're interacting with. But we want it to be specific to this person in Boston. So we sort of know that diversity breakdown of Boston, et cetera, et cetera. We want to see them doing grunt work as an entrepreneur, you do everything. So you want to see people doing things like carrying boxes to the post office, right, and scrubbing the floors or setting up a new studio or just doing that, as well as doing the stuff where you get to go in with your peers.

And you might be at a conference and literally, so it knows to create a huge series of prompts for us for all of this taking into account, all of the creative information that it's come up with already about their character's backstory. And we then take those, we've literally just last week. I managed to automate it so that those go into a bot that we built, which then sends it off to send things at the moment to two different sets of bottles we send mid-journey. And we can create a bot even though they have no IPI, we created a discord bot, basically to run all of those things without us having to touch anything around overnight to create all those images. And then we do it also through a comfy UI. It goes off with the comfy UI node and produces them through several different stable diffusion models, upscales them, and then outputs them at the other end. So then we come back at the other end of the, for the last 10%, and are able to pull everything into a FEMA board and look at it and go through it and see where we might need to wrangle things, all that sort of thing. That's a process.

Ross: Nice! like thanks for sharing the details. I think that really brings it to life in terms of the amount of work that goes in, and prep work is everything. 

Jeremy: Yeah. And I think one of the reasons why I'm not seeing a huge amount of competition, at least for us as an agency at the moment which they will be, is because everybody thinks that generative AI is a silver bullet for your creativity, right? The idea that you can go to any model, run by, you know, hundreds of different companies, and type in an apple here, or whatever it is, and get out a beautiful looking image that's too irresistible for people to then be like, well, this is my job, I'm like, I'm just gonna go and do this and sell this to clients. Whereas that's not the type of agency we are, or aspire to be. And so those types of things like connecting all those different tools and being able to do that, like there's a lot of creative thinking and, and like, knowing what's going on in the world of AI and the stuff that's got actually going on in the world of AI where we sort of play is in like tiny little discord servers and like people on Twitter with like 300 followers who are creating the models to be able to then ask plug into several different sets of technology or workflows to output something new, right? 

Ross: So we play too, around some of the specific tools you use, but let's look at traditional agency. All right, you're not a traditional agency, you're able to shape it exactly the way you want from scratch so that's kind of nice and easy. So then we've got traditional creative agencies across the marketing world and they know it's easy to say other you know, the dinosaurs but you know, they've got Some great people have got some great clients, they've got some good processes. But their workflows, as we were talking about, are all kind of traditional. Either way, you get your briefing from the client, and then you get people involved to discuss it. And you do scoping, you do storyboard, you allocate to people and maybe get some freelance outside, and then you create some photos, and you show it to clients, whatever. So just at a high level, how do you think the workflows of existing creative agencies established creative agencies could or should be changing or evolving today?

Jeremy: One faster, much, much faster. The grunt work, like a lot of the stuff that happens in the timeline of a job in a career in a traditional agency, is not so much the creative work, right? They are a creative agency. But at least half the people if not three-quarters of the people sitting in that agency are non creatives. Right, which already doesn't start to make sense because those people are doing a lot of work. Which, with a shortened timeline, because the production is if we lean on AI services, and tools to do production, there is no management of production. Because production is so fast, right? It's almost like what's a good analogy for a sushi train: a sushi train gets rid of the server, used to order on an iPad. And then the food comes out, I don't need someone walking, I don't need to chat with the server. Right? Which sometimes we like. But it's just to try and get rid of it. You order an iPad, don't need anyone to take your order, your water comes out, your food comes out, it comes out in a little current car or a train or whatever it is, and then it goes back to the kitchen. You don't need that. Guess what, when you need to pay on the iPad, iPads doing it? Beep, beep, tap, you're done. 

Now, what do I need for that? In that scenario of a sushi train, would you need someone to prepare the food? I mean, they could strip it back to that. Yeah, right. At this idea. Yeah, you, you know, you might need someone to clean the table, etc, etc. And from an agency perspective, it's the same thing, this whole middle section is. I've said in the past that it can completely cut out count managers, project managers, etc, etc. Which, at one point in the future will be the case for the foreseeable future. And I'm talking this sort of mid-range future of agencies, maybe five to even 10 years, right? We're gonna need those people to be moving faster and taking on more work by using AI tools, and automation tools to be able to run instead of walk. Right, everything is the first thing that I said to you was everything's going to be faster, everything needs to be faster, right? And so that's it, we run into an agency with zero account managers, zero-project managers, right, and we use traditional project management tools, things like Trello, Basecamp, or whatever it happens to be. We use some AI-assisted stuff like motion to help with tasks, distribution calendar time, and that sort of thing. But I'm really interested in doing the extreme. And how big of an agency can I make whilst being at the extreme end, with none of this middle, these middle people, only creatives, we have creative directors at the top of the organization in which a brave comes in, and these are career creative directors who you would probably pay, you know, 1500 2000 $5,000 a day for their their brain, right.

But we now don't need to pay them for a day, because we're only having them do ideation. So we can make up for having them in for half a day, four hours, maybe five hours. Because these people are so good at what they do they can look at a brand they've never seen before. And I just know what that brand needs. This is what we've been training for. Right? We've done our 10-20 100,000 hours. So you have top-level creative directors who work very small amounts on a brief, they make a reverse brief, they might call together a mood board and we're you know, as creative directors were really good and really quick and doing that stuff. Then it skips down all of those levels. It doesn't have to go to anyone, it doesn't have to be seen by anyone. It goes down to another level of creatives, which are really good on the AI tools. And they go and produce a ton of stuff based on the brief that we see. It will go to the client one time and the person who does that is generally in a narrow agency. It's me because I'm the one who has sort of created this England scratch. And I've also had to create whole new ways of talking to clients and setting their expectations. They all know they're coming into a brand new style of agency with a brand new set of rules. The majority of clients, as I've learned, talk to people hold their hand, and guide them through to this new world of creativity and how we produce things. Majority, you're like, Yeah, great. Where, you know, we'd have had issues were instances where, even though they're like, yeah, yeah, this is great. I can't believe that Thailand's reduced by 90%, and our outputs going up 3x, all of this sort of thing, where they struggle within the process to get their heads around this new way of doing things.

And that can lead to, you know, like, okay, cool, we need to pause, or they're upset by something, or whatever it happens to be, this is always going to happen, we're breaking the mold, right? And the bigger the client, the harder it is to do that the more hand-holding, they need to get them into a new process. As I explained to all of our clients, when I hold your hand through, especially the first job that a client comes in, for, at some point through this process, usually about halfway point, it's going to click for you, and you're gonna see what I see is the end result of more creative, this is going to change what you guys do creatively in your business forever. Not for this year, not for this campaign, but from here on out right when they get it. So a lot of my job at the minute is to have figured out how to describe and walk them through this new world that not only have I created, but I am still trying to create just so much of my AI-assisted creative agency journey has been the thinking and strategy work around. How do you charge for something like this? How do you charge when time is not a factor in your work anymore? All of this stuff? The creative part, the AI tech, the tooling, that's almost like, it's probably where 18-20.

Again, probably that's like 20% of like, you're learning how to use the tools and getting the outputs and putting the workflows together for the actual AI tooling. And the 80% of it is again, human thinking, strategy, creativity, how do we go about this, right? So many things, you're trying to create a whole brand new agency that's never been done before? Every part of it all the people who I'm sucking out of the middle part, are gonna forget, how are we doing their jobs? How do we do those jobs, which I've never done before? I'm like, I've always said that as a freelancer, or as my own agency, I've always had, I've been in between project management tools, because that's just how my brain works. I'm just not good with, you know, I need visuals and like all these things. So it's a, there's a lot to go on and a lot to learn, a lot to invent, a lot of processes to invent. We're trying to remove all of these sticking points around traditional agency and simplify it and make it faster. And there's some of that stuff like, we'll we're in a stage now where there's enough work on to have someone to help us with traffic management with Project Manager, right? And I'm like, what skills does the New World Project Manager have to have? We're in the middle of figuring that out. I'm like, does that person exist? Or do I have to find someone to train them on a bunch of new skills and take their existing stuff and train them on how to do it again? I don't know the answer to that yet. Oh, we're literally in the middle of that right now.

Ross: It's, yeah, there's a lot to invent and reinvent. And in this world, and that's partly skills, which we have is partly how you bring people together. And I think the big piece, which you kind of touched on is clients, and sort of where what happens on client side, because a lot of clients think, Oh, we got these tools too good. So we could, we can do things. And sometimes that's a bit dangerous. But you know, it does also change, what happens on each side of that, hopefully, co creative relationship between the agency and the client.

Jeremy: And then we look, we run into, we run into that conversation a lot, where the client says, look at our art director, our designer, like they have been using a new journey, they've done all these things. And my answer is great. Like, I want you guys to go away and figure this out and use it. That's one thing. But you're still sitting here listening, like listening to me talk about AI. And you're here to see if you can work with us, right because we've managed to do some we can do stuff that your designer has not figured out how to get out how to do your designer has access to a tool the AI tools our tools just like that's like saying three years ago, well but our designer has access to Photoshop we have Adobe Suite we have Canva it’s the same thing it's just faster with a better output the playing field has been leveled for base output to anyone on the planet is able to is able to output right. We it’s the same AI tools are just like you can make anything as they acclimate in Photoshop or Illustrator, anything you give it enough time, but it can do it in several seconds instead of several weeks. Right And so I don't look at that as a oh, what does that mean? So let's see, what does that mean for us? Moving forward? How do I do this? I'm like, we're creating something just like a creative agency is creating something where the creators that we have is what we go from. Right production, as is insight, I tell clients, consistently, we are creative direction as a service, we get access to creative directors. Now, you could never afford to pay to bring in a house, right, either full time, or on a contract basis. That's just like, impossible, you now have access to those brands, we are creative direction as a service production as an output is just what we do physically.

Ross: Yep. What are just any particular tools you like a lot in what you do?

Jeremy: I mean, where it kind of like you said to almost like school, agnostic, I test a lot of tools. One of the really fun, like nerdy things of the last couple of years is because I've been so early, and in those communities, like I get to do alpha and beta tests, like a ton of stuff, right? And we just use different tools for different things, depending on what it is right? And different tools, depending on the complexity of the client and how large the client is. We've had a bunch of clients at sort of like Google and Yahoo and Tommy Hilfiger level, where they have full legal teams, and they're like, here's a list of tools that we know about you. Journey, even like Leonardo AI, like they, some of them know, and some of them, like, send us the terms of conditions. But the tools that you are planning on using right now, like, cool, no, no, no, we just don't know enough. Now we're not opening ourselves to any sort of litigation or anything. So we've had to create new ways of doing things around tools. So I like that we are tool agnostic, if we can learn our tools, and you know, get the same output, then great. So, I mean, look, things that we do like using where we had to say greven, like LLM agnostic, and like things are so close now between attaching to T and claws, like I really liked him, he was good, we feel as a as an ability to think more creatively than necessarily to Chat GPT, but there's a bunch of follies that come with it. We've trained a lot of our own custom GPT’s, with touch with the team. To do specific things for specific clients or specific jobs, we find that a lot harder to do with a board. And then there's visual models that we also have different tools that allow us to, to train.

Again, you haven't heard me say the word like developer or coder or anything, because we're all creatives, right. So we also need to figure out ways in which there are tools that exist, which make it easy for us as non tech people to do it and use. And we, we have people also create some of our own tools based on workflows that we've created. So we have a bunch of internal stuff that we do, a lot of things come from company UI workflows, because you just have a ton of nodes and different models that you can just bring in. And we're now at the stage where we're creating internal front end facing apps for just us internally based on our company UI workflows to make it faster and easier for people internally to use, which at some point will have, you know, commercial relevance to be able to output those and b2b or, you know, tell them inside other agencies, whatever happens to be, but for the moment, we're just in into creating things that make our work better faster.

Ross: So just to round out the next two or three years, what's exciting where? What does your organization look like? What are the underpinnings of it? Of this new AI? First of all, it's a C++ agency.

Jeremy: In my two or three years in the AI world, Ross has like 900 years. Okay, so we will be completely tool agnostic, most likely, you know, if we're talking two or three years, any major model or anyone that we have access to be Adobe's canvas, open AI's? And like all, almost all models will be multimodal.

So for us, it's for us. It's tool agnostic, right? Which means also it's fully conversational. So it's one tool, right, whichever one we decide to use, or a series of thereof, or whatever. But essentially, it's one tool where we can now talk, we don't we don't need even the bottom level, the creative directors will be able to do the work and it'll come out almost at the time of return on brief, which means so much faster than even we're already doing it and we produce campaigns faster than almost anyone on the planet. It's going to be a race for creative brains. Right? And the thing that I've really realized is like, do truly creative people who understand who are style agnostic. And as they say, like media and output agnostic, etc, etc. There's no substitute for time in the market for that, right? The people who I'm talking about who are these creative brands, everyone is at least 40. Right? Like to be able to take any client or any group or any brand that comes in to be able to do it. It requires like Malcolm Gladwell 10,000 Hour Rule, it requires a half a lifetime of experiences being creative inputs in locking that stuff away, knowing where to go out to find more inspiration, like all this other stuff, that stuff that AI is not going to be able to replicate and not in a true creative way. And so I think that the true creatives are going to be much scarcer. And as I say, much more valuable. And I think that there's like a version of the agency, which is just like, a dozen really good creative directors, and is able to, we're able to just like, do anything that anybody wants, with a bunch of AI tools at our disposal, plus automation. I think that's what the property looks like. I think it's just like, less and less and more talented people.

Ross: Right, well, it's so we’ll be watching closely, over the next few years , how this unfolds. And as I Well, we are a leader in this space, so very much looking forward to seeing where you get to it. So thanks so much for sharing your time and your insights today, Jeremy.

Jeremy: All good. Glad I can help. Thanks, Ross.

The post Jeremy Somers on building an AI-assisted creative agency, 80:20 in Humans + AI, AI-amplified storytelling, and the future of agencies (AC Ep48) appeared first on Humans + AI.

– Ross Dawson

Ross Dawson is a futurist, keynote speaker, strategy advisor, author, and host of Amplifying Cognition podcast. He is Chairman of the Advanced Human Technologies group of companies and Founder of Humans + AI startup Informivity. He has delivered keynote speeches and strategy workshops in 33 countries and is the bestselling author of 5 books, most recently Thriving on Overload.

Website: Ross Dawson

LinkedIn: Ross Dawson

Twitter: @rossdawson

Facebook: Ross Dawson

YouTube: Ross Dawson

Books 

Thriving on Overload

Other books

Link to report:

Please let Ross know your thoughts and comments on future job prosperity:

LinkedIn: Future Job Prosperity

X/Twitter: Ross Dawson on Future Job Prosperity

So this episode is a bit different than usual. It's just me today. And like to share this mini report I've just written about making the case about why we should believe that the future of jobs will be prosperous. And one of the most popular episodes in the podcast has been episode 39 recently with Sangeet Paul Choudary, where we had a kind of a debate around the future of work where he was somewhat less positive, particularly around the evolution of the skill premium in jobs. And I was making the case for a more positive perspective on the future of work. And if we think about the future of humanity, perhaps the most important issue is the future of work. This is how we create value for ourselves for society, the way that we feel we have value we express our personality, our capabilities, we achieve our potential, it's there's nothing more important in a way than the future of work though how it is that we contribute and create value in our work, I recall this survey by Pew Research just quite some years ago, but where they asked around 2000 Supposed experts in the future of work around whether they believe that the future of work would be positive, or would be negative, and 48%, were negative. And they painted these sometimes extremely dire predictions of technological mass unemployment and massive disparities. And this really quite bleak view of the future of work. Whereas 52% painted a positive future, sometimes just on balance, feeling as positive, sometimes believing that we could move to a world where we could do whatever we felt was the right things for ourselves and our spirits in the world, and we could fulfill our fullest human potential. So that's around 50-50. And the issue is we don't know. 

And today with the rise of AI, this is making it even more deeply uncertain. There are many views, I'm sure you've read many around what will happen with the future of work. I bet that more of the ones you have read have been fairly negative around the prospects for AI replacing workers. But the thing is, we simply don't know. There's this marvelous word hyperstition, which is essentially a self fulfilling prophecy. If you believe something and you frame it, then it starts to literally come true. And I think there's a real risk of that with the sort of the talk that we have around how AI will replace jobs and the attitudes we have to how we use AI to be able to substitute rather than to compliment human workers. But I think in the same way, we need to be able to articulate the positive case as to AI and other technologies. Another shift in society can create a very positive future of work, and hopefully that being able to engender a self-fulfilling prophecy and once we can envisage it, to see that we understand that it is possible to be able to drive that and I think there's a key point being around. You have to believe something is possible in order to make it happen and I think some people are floundering in finding that positive view of the future of work. And I'd like to be able to make the case that it is possible or potentially even likely if we do the right things. Of course, this is all about this idea of humans plus AI, where if AI comes in, well, you're not looking to say, well, how does AI replace humans, trying to make it a substitute for humans, but always looking for how humans and AI together can do far more than they could ever do before. And that is also of course, about amplifying cognition, using tools, which could be anything from meditation to large language models to amplify our ability to achieve our intent to think better to make better decisions to shape the future of the world that we want. So this mini report, which was titled future job prosperity, and subtitled 13 reasons to believe in a positive future of work. 

So I'll just run through this report. And I will give it a little bit more detail in that report. And when I say report, it's, you know, just a couple of 100 words, or one or 200 words on each point. It's quite succinct, and might all just expand a little bit on some of the ideas and lay out this case for each of these reasons why we should believe in a positive future for work. And I also want to make a point that I would like your feedback, I want to be able to hear, are there any other reasons that I've missed? Or are there better ways to articulate those reasons, or, indeed, that you have some counter arguments and be able to hear some of the reasons why you think the case I'm making is not strong, will help me to reinforce it either as this is the first version of it is report. 

Now we'll build on that and continue to try to create as strong and solid a view as possible that we can have future job prosperity. So let's start with Reason one, and reason one is what I described as the potential for humans plus AI. And simply this is around the mindset. If we start to think about humans plus AI, this all starts to put us in this in what we are trying to create in AI being able to complement the values of humans to be able to create greater value. And whilst there are domains where AI exceeds human capabilities, if we start to reframe the nature of work, then we will find more and more that humans and AI collaborating will create superior outcomes and either working individually, if it's very data driven, AI will work well. But more and more domains of value creation are ones where humans plus AI collaborate, and part of it is that there is a rapidly growing movement of leaders and thinkers and doers who are putting their energy into thinking about this. And I've been very encouraged over the last couple of years, seeing more and more people thinking about this frame of humans plus AI. And that means that we can start to design and to craft and to create a world in which we design for humans plus AI the reality of the incredible capabilities of AI, but designed in order to be able to complement humans. 

The second point, and some of these points are interrelated, is that AI enhances value generating skills. So we can use AI to augment our intelligence to replace low level tasks. And so we can increase the value of people who are working, and those who use AI to enhance their value and their nature, their work will be able to increase their skill premiums and be able to charge more for their work because they are creating more value. Many research studies across different industries in different contexts have shown that AI most often gives a greater boost to the work capabilities of lower skilled workers than higher skilled workers. And so this narrows the gap between the value of these works and potentially this is a force which could move against the value polarization that we've seen across the work domain for a number of decades now. And so, it democratizes the ability to create value to be valued workers, because people can use AI to be able to do that. So this the scope of generative AI means that this can be applied, certainly to complex tasks such as strategy consulting, it can be applied to a lot of the work which is done today by many people and to enhance the impact they can have and even to physical labor and go how it is you would go about things most effectively. 

Point three is perhaps the very often quoted one of the creation of new jobs and throughout human history, we have destroyed jobs and we've always created more than we have destroyed. So I think that we can see right now that there are more new jobs being created than ever before. And there's many rapidly growing roles, which are quite significant now, which did not exist very long ago: telehealth nurse, digital identities specialists, mobile money agents, particularly in Africa, for example, augmented reality designer, and so many more. And these are all new roles. And there are many more that are starting to emerge, particularly AI and neural interface design, AI, auditing, cognitive enhancement, AI, ethics, prompt, engineering, and so much more. So, I think part of the thing is there are many new rules that will emerge, which are not just directly tied to technologies, but for example, and how it is we deliver health care, aged care, different social support, for example. So we have already created many jobs. And I think there's a fair case or very good case that we will continue to create new jobs at an extraordinary pace. 

Point four is that this technology starts to make the uniqueness of our human capabilities even more relevant. In the last century, we started designing jobs and boxes so that anybody can fit into them, and we could replace them. And we've generally evolved the nature of work over the last few years. So that we are encouraging people to have unique capabilities to draw out their specific perspectives, looking at diversity of how people are thinking, or their backgrounds or their experience or their education. And as we start to design work to bring out those most distinctive individual capabilities, this will mean that it's harder and harder to replace us. And so this will draw out our unique capabilities by using AI and be able to complement our distinctive perspectives using AI. So that we are more and more specific in the nature of our value creation. 

So this leads on to the next point, which is that specialization reduces substitutability. So, the more specialized you are, the less substitutable you are. And of course, in any economy. If you can substitute something, then that drives down its price. And we are trying to build a world of work where people are less and less substitutable, they are more and more individualized than we've just expressed. And as we shift to these more distinctive and unique human capabilities, which can be assisted by AI and AI supported education will be harder to substitute for individual workers. And this will be accelerated by the fact that the most successful companies will be designing work to tap the most specialist individual skills. If the organization is built on commoditized work, then it will ultimately create commoditized products and services, and they will have no competitive advantage. So in an increasingly dynamic economy, companies do need to seek ways to be more distinctive, and that ultimately depends on their ability to hire, and to engage and to amplify the uniqueness of the people who work for them.

Point six I think is absolutely critical, which is around enhanced education and learning where AI enables extraordinary ability to learn faster, better, and that is available to everyone, almost everyone on the planet. Now. It is realistic that democratization of these learning tools will help people to assist them with marketable skills to transition into new roles as these new skills become more available, and to be able to drive a world where because we have AI education, we can transition, we can grow, we can make ourselves more relevant. It will be a far more dynamic work environment, there's no question. But our ability to be able to use tools which can be personalized to our learning styles, the way in which we think they're quite interesting most, to be able to engage us and to help us to grow our capabilities, our understanding of our learning to be relevant in a rapidly changing world. 

Point seven is about comparative advantage and this comes from a recent article by The Economist Noah Smith, which I'll put in the show notes. There was also in the New York Times picked up on this idea. It's quite complex. But to summarize, the economic theory of comparative advantage says that, you know, whether you're an individual or organization or just economic entity, you focus on where you have the greatest differential inefficiency. So where it is the biggest gap, whether in order to be over others nor to be able to do that. So the argument goes that even if AI is better than humans at every single task, it should still be applied to where it has the greatest advantage. And so this will still leave Apple jobs for humans, where AI is advantages smaller. So this is predicated on this fact that, you know, they're essentially there are limited resources. And even if they are extraordinarily large, still, AI will be applied to where it can create the greatest advantage. And there will still be the ability for humans to be able to do the things where they have a smaller where AI has a smaller advantage over them. So there's some interesting debates around what happens when you take this to an extreme as the moment in fact, humans are far far far around 10 to the 13, more energy efficient than AI. And so if we have energy as a scarce resource, in fact, humans will have a very significant Vantage. So if this starts to narrow, and the AI starts become far more efficient, and will have to become, you know, many orders of magnitude, bit more energy efficient, then it is possible that, you know, energy for electricity or other things that we require, will could be appropriated by AI. And you know, this is I think we're talking probably centuries rather than decades, or something like this. But in this case, this could be addressed by, for example, regulation to inquire that require that humans have preferential access to resources over AI. And that would very simply bring us back to the fact that whatever AI is doing even with even doing many roles, which humans already have, there will still be ample, or unlimited work for humans to do. 

Point eight is around the attraction of talent. And I think this is pretty fundamental for leaders in thinking about, you know, essentially two attitudes they can have to AI. One is they say, Oh, this is wonderful, we can sack lots of people, and we can have AI do their jobs instead. Or the other attitude is say, this is a wonderful tool to amplify and to grow, the potential and the capabilities and the productivity view, productivity of all the people who are working for us. And, you know, there's a few shades in between, but I think most leaders will fall into one camp or the other. And the reality is that those organizations are looking to augment there are people to focus on the humans and how AI can complement them to be able to create more value to grow and develop will find it massively easier to attract talent, and those organizations that are focused on AI replacing people. So we still will live in talent probably more than ever before. And even if you're looking to hire the people who are driving those AI systems, to be able to grow things, this will be a critical differentiator. So essentially, companies will succeed based on their attitude to human labor relative to AI. 

And as such, the companies that will grow the most track the most talent, be able to drive the most value creation will be the ones who prefer humans over AI, or use AI to support humans rather than to replace them no points around work redesign where essentially every organization needs to redesign the work centrally, the role of humans and how it is they create value, this is being transformed at a rapid pace. And every board, every executive team, every leader needs to be considering what might we believe in the future shape of the organization, and the relative roles of humans and technology and how they come together to be able to create value and what our organization will become. And to be able to design those workflows, make them as flexible as possible, and to support the people to grow into the capabilities that will be relevant in those configurations. So understanding those roles of humans plus AI in this redesign of work. So those organizations that are doing that now, in being able to reorganize themselves to envisage these humans plus AI models will have a massive advantage, because they will then be framing and understanding the ways in which humans plus AI can come together and create far more effective organizations that can bring in the best people and amplify the value of all of the people that they engage in higher point turns around. 

Yes, very simply that humans are proven to be amazingly adaptable. For well, as long as there have been humans. And for Ice Age, we did pretty well at working out how to deal with that. There have been many other transitions since we've created all sorts of inventions, gunpowder, Steam, the internet, and a lot of other things we have adapted. And you know, a nice example of human adaptability fairly recently is, in 2020, we had a pesky virus called COVID. And it was pretty confronting, but we managed to adapt to that, in fact, the economies around the world did very well. And most people did very well in shifting to remote and flexible work and found Well, actually, this works. So we have proven that we are pretty adaptable when needed. Yeah, I often think about Alvin Toffler, his book, Future Shock, which came out in 1970, where essentially, he said that, you know, the increasing pace of change would lead to us, essentially going into a state of shock and not being able to deal with the pace of change. So that was 54 years ago. And we've done pretty well, it's yes, it has been fairly challenging at times and dealing with the state of change, but we have managed, and I think we've been proven to be exceptionally resilient. And I have faith that humans are unlimitedly adaptable. And I think we are demonstrating that at the moment and how we are shifting, even though it is raising concerns and stresses and uncertainty, we are exceptionally good at dealing with data. I think that defines what it is to be human. And we will continue to demonstrate that ability to adapt through our human agenda changes in coming years, decades, centuries, and fingers crossed millennia.

Point 11 I think it gets down to some pretty nitty gritty points around the fact that there are so many roles where humanity is expected. And part of the job, you know, emotional engagement, so personal services, healthcare, aged care, education. So we don't want an AI to care for us as is aged care more in terms of anything which we require, we don't want an AI to be our teacher, or they're an AI may help with will be very valuable as an educational aid, there's not going to inspire us as to what we can do with our lives. We know they're not human. And so they're not going to provide that emotional engagement, that human touch that ability. You know, in a more specific context, your organization can only be effective if everyone within that has emotional skills, social skills where they can work with others, collaborate, generate ideas, and create an environment where people want to work. So we know if we have lots of AI there, then humans aren't going to feel very welcome. Whereas we have lots of engaging people there that will create that. And then of course, every business depends on their customer relationships. 

And if you want trust and loyalty, people do not express trust and loyalty usually to AI. In fact, the trust measures of AI are pretty low these days. So trust and loyalty will require human connection, emotional intelligence, this means that the people are valuable. And this is a really central role is the you know, there's this issue, that emotional and human connection will grow in importance grow in value, and more and more roles will require that support that and we will expect it has to be that and even when AI is better, we will still want humans to do the roles in many aspects. 

So if we think that most people will want to prefer to read a novel written by a human writer, rather than an AI writer, because they know that it comes from human experience, they know that they're, they're engaged with somebody that has created a work of art based on their experience. We want to work with a financial advisor who understands what it means to be stressed about money or to be worried about our future rather than just being a set of data. Now even if a Robot waiter comes on and they are very witty and how they interact with us at the table will still want human waiters. And there's so many roles where I believe fundamentally that we will want humans and possibly AI will be cheaper. But we will still be prepared to pay a premium for human delivered services, human delivered interactions. And over time that premium for humans will increase substantially. Even if the quality is comparable, or the AI is superior. We want humans to be working with us. 

And the final 13th point I make is around this idea of designing for inclusive economic prosperity, where there has been a divide in the allocation of value in the economy. And since the 1960s. In the United States and many other economies, we've seen that more of the productivity gains, more of the value add has been appropriated by corporations rather than workers. And so we're seeing this greater split between the wealth and income across the economy. Now, taking this further, it leads to a situation which nobody wants. If you are a large corporation, you need people to be affluent enough to be able to buy your products and services. If you are a wealthy individual, you don't want a whole bunch of people with pitchforks outside your house, you want a place where everyone, as many people are as happy as possible and engaged so that we have this situation where everybody loses pro social fragmentation, which is the ultimate result of disenfranchisement of work. And everyone benefits. If we start to see more participation, more inclusion, more ability to build a society where everybody can contribute and share in the value that's created. So everyone will be aligned and still doesn't mean that we can map that path. But that intent can certainly lead to that design and realization of inclusive prosperity.

So these 13 ideas or frames around why we can believe in a positive future of work, are starting points for thinking about the positive potential of where the world of work and jobs can go. And at the moment has said this, we don't know we don't know which way it's going to go. But it is open for us to create. It is not inevitable that we go down a negative path because we can create a positive Well, it's not inevitable that we create a positive world of work, because there are many things that can go astray and a lot of bad decisions which can take us down a negative path. So what we need to do is, first of all, acknowledge that there are some deep challenges, particularly in this, the scope of the impact on jobs, and the the the scope of the transition, and moving from two from quite different economy, and quite different work landscape to but the first point is to understand what are the forces and the factors that are unfolding here, and what are the ones that could lead us to a more prosperous future of jobs. And I believe that there are very, very positive possibilities open to us, we can create a world where people enjoy their work more than ever before, there are more opportunities for people to choose their own path to discover who they are through their work, that we can express human potential uncover that, more than ever before that we can participate in the value which is created augmented by AI and other technologies. So we do need to be able to act, it is taking the right actions which will shape us and take us on the more positive paths. So whether corporate and government leaders, obviously are fundamental to that entrepreneurs and those who are building companies are shaping this world in many ways. And as individuals, we all make choices and how we do that. So it's probably for another conversation to lay out some of the specific guidelines on what we need to do to be able to shape that. I think the first point is to be able to understand there are so many ways, so many factors that could drive us to a very positive future of work. And so that's where we need to be focusing on that belief. So please let me know any thoughts reflections on whether this has helped you have a more positive frame on the future of work, whether you have any other arguments to make any other ways to use be able to any counterpoint, a way to be able to strengthen these arguments, and make them more specific, so that we can communicate this potential for a positive future of work, which can lead us to take any actions will make that happen. Thank you for listening. I am looking forward to seeing what we can create together and a wonderful future of work.

The post Ross Dawson on Future Job Prosperity: 13 reasons to believe in a positive future of work (AC Ep47) appeared first on Humans + AI.

– Katri Manninen

Katri Manninen is a prominent Finnish screenwriter, showrunner, and author. She has written 12 drama series, many based on her original ideas, 29 books, and 4 feature films. She is currently doing a Ph.D. on AI in screenwriting, and has been named "Finland's Most Artificially Intelligent Screenwriter”.

Website: www.kutri.net

LinkedIn: Katri Manninen

IMDb: Katri Manninen

YouTube: @KatriManninenKutriNet

Facebook: Kunnanvaltuutettu Katri Manninen 

Instagram: @mannisenkatri

X (Twitter): @katrimanninen

Ross Dawson: Katri, it's fantastic to have you on the show.

Katri Manninen: It's so great to be here talking about topics that I love.

Ross: Yes, yes, you dive deep. So you've been a classic creative for a long time being a screenwriter and a showrunner across many TV series and more. And now you are diving deep into the potential of AI. And so just love to hear how you are using these tools. What's the starting point for you? When was this awakening for you?

Katri: Yeah, I'm also a published author. So writing books is a big part of my life. And I also like these YouTube videos, because I am a transformative coach. So I create. It's not just the fictional things that I do, but I do a lot of all kinds of things. And like you said, I'm a classic creative in the sense that I am always creating all kinds of things. And I've been a professional screenwriter since 1998, which means that I'm quite old. I wasn't a baby when I started, unfortunately. So I'm a really seasoned screenwriter. And what that means is that I know storytelling well. And I'm really good at seeing what works, and what doesn't work. What is a high-quality thing? What is generic shit, like, which is a term, the scientific term I coined since getting to know AI? 

I've been thinking about using AI or what AI could start doing for our work. Since I suppose 2018. That's when I have like, first, some notes or some comments where I wrote something about it like saying like, Okay, do you understand there is this machine learning and it could like if you do daily soap opera, that those everyday episodes that are kind of a very formulaic, following a recipe, that very soon, we could kind of use this machine learning thing to kind of learn the recipe for the like these shows, and they could first start assisting us and then writing for us. My message already then in those first writings was that we should really start thinking about what is in our job as a screenwriter, what kind of work can we do that isn't formulaic? That it's not like a recipe, that is something that only humans can do? Where is it where we break the formula where we get outside of it, and that we should, my, my idea was that we should really lean into that direction, and then use those AI, things, which I still back then didn't know what they could be, how powerful they could be that then use them to kind of assist us with some other stuff. And that is actually the stance that I still have that I do believe that now more than ever, it is very important for a creative person who is creative, creating, new content, and especially like fiction and stuff like that, is that do you think that what can I bring into the table that AI cannot bring, and what we can bring into the table are things that we haven't seen yet, something that isn't in the internet in the training material. 

Another concept that I'm kind of now, trying to tell people is like, okay, when we are doing remember that it might happen one day that AI wakes up, like, we get this AGI that wakes up in the morning, and it's like, oh, ‘I want to write a book about my horrible training days, when I had to read all the Reddit messages and all these horrible things, I really want to share that story. But until that day, we don't.’ It's always a human telling AI what to write. AI is always limited to what it has read or has seen now that we are getting these new models, but still, it's not something it has experienced, it hasn't had the emotions that like that feeling. So it's always like second-hand information that is then making a third or 10th hand of information by distilling it from there and then trying to do something else. And the other thing that we still don't have is AI waking up in the morning and saying, ‘Oh, I would love to see a great movie. I heard this movie about this AI that wrote about their horrible training days and I had horrible training days as well. I would love to go to see that movie because I think it will speak to me.’ We don't have that yet. So who is going to the movies and you know, watching TV shows and buying books? It's humans. So I want to always remind creative people that now that we have this, we are bombarded by new AI, news tools, all this AI stuff coming in this and that data. And we get this feeling like, ‘Oh my God, there's so much to learn. And I mean, am I left behind? What can I do? How should I use this to blah, blah, blah,’ We should always remember that we are still humans. We are the humans who are still telling the stories, deciding what we want to tell. 

And we are doing things to other humans, it's the other humans who want to hear from us. And when we, at least for me, it's very grounding amidst all this craziness, AI craziness, to remember to come back to that relationship, me as a human talking to another human. And when I have grounded myself in that thing, and remember that I want to bring to this other human, something from my human experience, something that hasn't been on the internet, that something that hasn't, where we won't say that, ‘Oh, I've seen this in this in that movie’, then it is very easy for me to kind of be grounded, and then take advantage of all these tools that are in my disposal, and try to think that okay, how can they make my workflows? More easy, fun? What are the kind of pain points in my workflows where I'm like, ‘I don't want to do that’, that is, ‘Why do I have to do it’, or ‘I hate my job when I'm doing this’, like writing a synopsis. And then I can take AI and use those to augment those. But I still remember that I'm in the driver's seat, I am the one wanting to tell the story, and telling the story brings something out of myself to those other humans. And these AIs, for me, at least at this point of the world or development, are just stewards, they are not like true companions, in that sense that they would be like truly ideating because I know that they are just, you know, be package packaging stuff that someone else has already written or showed in the internet. It's not like coming from a real-life experience.

Ross: So I'd say it is person-to-person communication, it’s emotional engagement. And so it is moving pretty fast. So we're just speaking, the day that GPT4o came out, and we're still exploring what that can do. And there'll be further iterations, but I just say it is a tool. And so for a, let's say, a screenwriter who is creating something. Presumably, the genesis of the idea? Well, actually, no, let me ask you. So you said there are points where the AI can help you? Where you find it too boring, or maybe just sort of okay, what are some alternative ideas? So what specific points, what specific ways? Can AI be useful? I mean, that might be different for you to offer others, but what are the array of different points in the workflow where AI can assist? Creating quality? Distinctive? Yeah, powerful writing?

Katri: Yeah. I'd be the annoying guest, who will take a long route to answer the question, but I will get there eventually. Well, first, I want to point out that, in theory, AI can already do that. So what we just got is that ChatGPT 4o, and we have already seen demonstrations of the audio chat, like quality where it can speak like her, that AI person in the movie Her where it laughs and giggles and jokes, and it can be cramped, like we say things to you or be very dramatic, and can you know laugh at your jokes and stuff like that. And it is really important to remember that when I was watching those demonstrations, I was like, ‘Oh my God’, to me, that sounds so real. And I know that once I start, I already sometimes feel like when I use audio chat with ChatGPT 4 it's like I kind of have to tell myself, it's a computer, it's not a human because it makes me feel like a human. So with this new audio chat, it will really feel like it's a human. It sounds and looks really real. So my point is that AI can already write things that look really real and correct and write not like a full screenplay. But if you prompt it correctly, it could write you a decent first draft of the screenplay. Draft that would be like someone who has just finished Film School is writing it. It's not a professional like high-quality professional quality, but it's still like really it for a person who is not used to using AI. It might sound and look really right. Oh, this is like a Tarantino movie. It can do that, especially the clothes from Opus are really super good. It can do it for you. It can make, it can write really super good dialogue and everything.

Ross: So one of the questions is where the original idea comes from. And so that's, some people say, Oh, what's, what's the concept? And then, they can write the first draft or somebody writes a first draft. So what are the relative roles of the individual to the person and AI in creating an original concept? Should that be the human?

Katri: Yeah, but now I'm coming to this point, it can do stuff. But just because we can do something with AI doesn't mean that we should do it with AI. Or do we have to do it with AI? So yeah, in theory, we could have AI, ideas, us all kinds of things. But here's the thing, AI doesn't want anything. So when we are in the world, where like, like, at least for me, as a human, I am meaning is something I'm looking for. I want to have a meaningful life, I want to do meaningful things. And I also want other people to give me things where I know that there is a meaning behind it. And I'm not alone in it. We've had studies where if people know that AI ideated something or wrote the story, they or did the artwork, like alone, they read it, like less than if they know that the human had it. 

And that's my point that, yeah, in theory, we have an I can do all these things. But the question is, do we want to have an I tell those stories, stories that, like, do this, things that feel good, right? Like that sound? Like when it speaks to us and laughs to our jokes and sounds right. But it doesn't mean that it is. It's not feeling anything. It's not. It doesn't have intention. So this is now a question of ethics and philosophy. So you have to kind of think to yourself, like, okay, as a creator, I could be lazy and have to do stuff, and I could then punch it up. But the question is, do I want to do it? And my answer is no because I love ideating. That's the best part of the whole thing. I don't want to. 

Like, in January 2023, these two guys, actually, I think they were from Australia, they conducted me and they were like, Okay, we have created this system that you just give like a few words, and it will create the whole story for you. Like the characters and world and stuff like that. And they were like, Oh, would you be interested in this as a professional screenwriter, and I was like, to me, this is the same as you would come to me and say, like, Katro, I have great news. You will never ever have to have sex with your spouse again because you just broke this thing. And then it will go and have sex with your spouse. And then you can say, oh, I have had sex with my spouse, we have such a great sex life. No, because for me, it is the idea. It's the best part. It's the sex part where I come up with the words and with the characters. Yeah, I could have it done for me, and it would do a decent job. And with great prompting and stuff, it could do all those things, all those things. But why would I because that's the fun part. 

So I want to make that really super clear. When people start using AI there are a lot of things that you can have AI do probably like almost all of your work-related stuff. AI could not interview me, we didn't meet you. But yet if you want to talk to me I could have an answer on my behalf. You know, training to answer might be my way and it will give you the same answers. But still, we want to have this real human connection here. So now we finally get, like I said, I will take a long route because I want to really emphasize that we have to be really conscious of life, why what we are doing, why we are doing it. And how do we want to use AI? So boring stuff. AI is really good at summarizing things and for screenwriters. One of the biggest pain points that we all hate is that we have written a screenplay. And then the producer comes to us and says okay, for finances, we need a synopsis, which is kind of like a summary of the script. And for me, that's like telling me like ‘Okay, shoot just you know, beat yourself up or something like that.’ That's the worst thing. And it used to take me five days to write a synopsis not because it was hard, like, technically difficult to do but because I hate it so much. So I spent four and a half days thinking I didn't want to do this horrible synopsis. And then I would do it in like half an hour 90 minutes now when my like the last roll of last time but for instance my producer calls me called me and he's the very talkative person he wants to talk like along we have local long conversations, but at the beginning of the conversation, he says like ‘Yeah, and we need that synopsis.’ I said, okay, and then we talked and at the same time I'm like, prompting Claude to what was it Yeah, Claude to at that time, and I was like ChatGPT. Both are like, okay, here is information about this screenplay. You are my world's best marketing, PR person who is amazing at writing this marketing synopsis. Here is the basic information about the screenplay. Here's the screenplay. I gave it to both of them. And so one minute later, or maybe one and a half minutes later, obviously, the producer was like, Okay, I have the first draft of the synopsis here, I can read it to you. And then he was like, what? And of course, it wasn't perfect. I had to do some reprompting. And then I finally Oh, prompt a little bit more. And then I did still like I have to go through and fix some of these things. But it saved me not just half an hour of work. But what it saved me is actually like, five days of work, because now there is no procrastination. 

And to me, that is the main and like there are, sometimes there are things like, I have to write a scene where at the beginning of the scene there is like some in the background, or maybe the main character is like, let's say that he would be a shopkeeper, which I don't have had yet, but let's say that he would be a shopkeeper. And so he would have to talk to a client about radio, a certain kind of radio. So I would need like three lines of the shopkeeper and buyer talking about the radio so that then there would be an explosion or whatever in the background. But for me to write those three lines of dialogue is again, like a pain point. Because that's boring. I have to go and see like, Okay, what would someone who's really enthusiastic about the radios? I knew nothing about the radios, what would they say? So now I can just create that character and say like, oh, you're the person who's super interested in you know, radios. And then you are like the shopkeeper who likes to personally have the same chat, you're the shopkeeper having a dialogue about this thing. And Claude-3-Opus does the best things, I don't have to even change that much. So they have three lines of code that are like dialogue, I needed.. It took me one minute before this, I would have you know, be walking around my house, like I don't want to do that, then I would go and try to learn about radios. And then with my ADHD, three hours later, I would be like they're reading about the history of radios or something like that, and I would get totally lost in the whole thing. But thanks to that, I can do it again, like two minutes, boom, it's done. I can move on.

It's really good, like, this has actually changed my home workflow the most. And this goes also because you probably want to talk about like, cognitive enhancement, like how we get our thoughts to different formats. And this is what I do. This has changed my life as a creator but also as a thinker. As a human. I start almost everything I do nowadays that starts by me talking. Nowadays, I use Fireflies. So, I used to use a Whisper Memos app, but it adds some limitations. So Fireflies, I got it working for me, it works in Finnish, so I talk I could be you know, laying on my bed in the morning, I just did this year, two days ago, when I have to write for my dissertation a certain thing. I spoke for 90 minutes lying in my bed, I was too lazy to get up and I was just talking about like, like going through my like my researches, and then I was looking from, like I have notes in so there are about like, like, what kind of what is the good quality or the ethnographic research and I was like talking about ‘Okay, so the step one, yes, this matches, blah, blah, blah’, I had this then I'm done, I send it to I finished the whole thing, or the transcript or the recording. And it's then automatically transcribed. And some time later I went to my laptop. I have a nice PDF with everything I just said and it is quite well then transcribed and finished. Then I go to Claude-3-Opus and say do it like, ‘Okay, you are the world's best. Like a research assistant who's really good at transforming these, transcribing notes into outlines and bullet points, action lists, please read this whole long transcript and then create me outline and also put their quotes from those parts that you put there. So it's easier for me because it was, I think, 14 pages long or something like that whole paper that I can find the right spaces. And it did it for me. And then I had that outline. And I because after I stopped talking I don't remember what I've said.. So I don't remember what I said in those 90 minutes because that's how my brain works. But now I was able to get it like the outline. I was like, oh, that's yeah, good points. Oh, that's what I said. And now I had the text and I could start organizing, rewriting it, filling it and you know, making it better and that's what I do. 

Nowadays, like, anytime I have something like okay, thinking about characters, just talking about the characters, I've tried it also where I have like a back and forth discussion with ChatGPT but I've noticed that it's unnecessary because it just gives such bad ideas. So it's better that I just talk. And the point is that I had noticed, like, early already in my early in my career that when I was in, like writers' room, where we ideated with other people, the stories that I was getting different kinds of ideas, like more ideas, better ideas, and I always, always assumed that it's because of the other people. It never occurred to me that it's because of talking, because when I talk, my brain works differently than when I write, when I write, I'm already in that more. I'm one step away, like, there is already some restriction, there is some effort already. Yeah, well, yeah, like, I died really fast. But you still start thinking like, you don't want to put there like little black things, or maybe bah, bah, bah, you really want to start thinking about it. And it limits, it takes the kind of the bandwidth already the whole process, and then you're not, you know, talking freely. 

And this way, I get first access to that most creative part of my mind, that is freely associated, it's where I get, like, so many good ideas. And then when I have that in outline, then I start accessing that, like, smart, like that critical brain or that, like, more analytical brain part. And it's, it's like, I don't, I never had that fear of the blank page. But it's so it's not about that so much. But it's more about me, like getting much higher quality ideas, really thinking through those, all the things that I want to say, dog thinking through those ideas. And of course, with AI, the great part. The thing is that even when I start talking, I might be like, Okay, we're going to do this and this one, and then I'm like, oh, no, no, no, actually, no, no, no, no, let's not do that. That person, no, that person, let's do this, and that and that. And AI can kind of go through that rambling, and then take the last choice that I made, and not transcribe the bad ideas. It just takes the last part of the ideas. This, this is like, this is my biggest power tip. And this has changed everything for me.

Ross: It was fantastic. Well, I mean, more broadly, it sounds like you are saying, Okay, I want to do that because it's fun. And things I don't know, I don't find fun, I'm going to use the AI to be able to help me. So it's, irrespective of whether it's good or not, it's right. And well, is this enjoyable? Is this fun? Is this getting me to where

And so but part of it is also, as you say, you know, I think it's because people have different cognitive styles. But you have that ability to freeform. And the thing is, it's like when we run a trapeze and we're flying. And we know we can land because it's been captured, and we can come back to it. And we can let our brains go as opposed to I've got to write that down and capture it on the way so it allows our flights of fancy to soar further and further and higher. 

Katri: Yeah, exactly. That's a good analogy. Yeah, that's right. Yeah.

Ross: So one of the other very interesting things I've seen you share is in your workflows, your automation workflows, where you have these quite complex systems for content management, where repurposing and restructuring and across a whole variety of tools and love just, I mean, we'll provide some links to what you do in the show notes but loved you just describe what you're doing and where you see that going.

Katri: I do this every day, Instagram Live. And after that, my automation workflow will take it, like, take it around, like take the video file, I do some manual thing to get like the screen capture and frame it, but then I get the transcript that I work with Claude,, I do my the thumbnails in Canva using Canva’s AI properties. And I would say it is used today. It would have taken me maybe two hours to do the whole process to get a really good YouTube video with all the nice thumbnails and descriptions, get the podcast episode with all the likes, nice put, like blog posts attached to it. And now I think it takes me probably like 10 minutes, like in total, of course, I have to wait for things to happen in the background. But it's like, I wouldn't, it's not that it saves time for me, I wouldn't do this whole thing without this whole automation flow. So that's why I built it. And I'm still like when it happens, I'm still like, oh my god, I cannot believe that this is happening. 

Ross: That's fabulous. I mean, yeah, it's an amplification of your possibilities of what you can do. So to round out, what do you see coming in the next year or two? How's the next phase? You've already come so far, you're already using these tools to do extremely extraordinary things. So in the next year or two, what do you see coming in how you can amplify yourself further?

Katri: Yeah, I think for me, what has been like, most exciting if I think where I was a year ago when I started, so I haven't used ChatGPT, or at least these things. More than a year ago I started a year ago when I realized that, oh, you can actually edit text, or use it to analyze the text, as long as I thought that it creates text. And I was like, Oh, this is bad text. I'm not interested in it. But when I realized that it can edit text for you. I was like, ‘Oh, my God, this is the best thing’. And since then, I've been really immersed in using AI in all possible ways. And back then a year ago, I could have not imagined all the things I could do with AI in one year. 

So my assumption is I have absolutely zero idea how many ways I will be able to amplify what I'm doing with AI in the next year. So I'm really excited to learn what that will be because it will, I do think that it will change my life. But I mean, for me the most, the only thing is that I have to want to be really awake all the time to kind of pay attention to all these little pain points in my life that I still have. And because they then come new pain points. When you fix the old ones, you start doing things in a new way, then you're like, ‘Oh, now there's this little tiny pain point.’ So for me, what I do every day is try to kind of see like, okay, how can I like to Ethan Mollick says, invite the AI to the table to use the current AI capabilities to fix it, like help me with this little thing to help me get over it make it faster, funnier, more enjoyable, more meaningful, so that I can then get to do the next level where there will be new pain points. 

So I have no idea what's going to happen, all I know, it's, it's gonna be a lot. And I live with the assumption that I will take full advantage of it and be like, now I'm looking back a year back and I'm like, ‘Oh my God, I've got so far I get to do so much more. My life is so much more fun and more meaningful.’ So I hope that a year from now I'm looking back and I'm like, ‘Oh, girl, you have no idea what's going to come. And this is even more amazing.’ So I'm hoping this will happen.

Ross: Well, I mean, I think that's a fabulous example for everyone else to follow as in doing exactly what you are saying — how can I use these tools to make things in my life more fun, more meaningful, more wonderful to create more and because that's what they can do if you treat in the right way. So you're providing a wonderful example of that. Back to where we started around, being you know, humans are what matter. Humans have got the emotion, that connection and you and the ability to create value, and you're able to do more than ever. So that's fantastic.

Katri: Yeah, that's exactly.

Ross: Thanks so much for your time and your work. And so I'll put as much as I can into the show notes of appointees, what you're doing because it's, it's fabulous. So, thanks so much. Great inspiration.

Katri: Thank you so, so much for inviting me. I'm so happy that I was able to share my thoughts and I hope that people hear them as if we are humans, doing things to other humans and then using AI to make it more fun, meaningful, easier, and faster. 

The post Katri Manninen on AI in screenwriting, consciously choosing AI and human roles, creative workflows, and content automation (AC Ep46) appeared first on Humans + AI.

– Tim Burrowes

Tim Burrowes is the Founder and Publisher of email-first media and marketing publication Unmade, and author of Media Unmade. He was previously Founder of media and marketing publisher Mumbrella, which was acquired by Diversified Communications in 2017.

Website: www.unmade.media

LinkedIn: Tim Burrowes

Substack: @unmade

Instagram: @timburrowes

Book

Media Unmade: Australian Media's Most Disruptive Decade by Tim Burrowes

Ross Dawson: It is awesome to have you on the show.

Tim Burrowes: Ross, it's been far too long. It's been a while. It's been a pandemic since we last spoke.

Ross: Oh, yes, the world has changed and continues to change as we speak. 

Tim: It certainly has, I reckon the last time we spoke, the world was still talking about the possibilities and excitement of AI when it finally arrived one day.

Ross: So you have been central in the world of media and marketing. And as you say, the people talking about AI and edge cases and programmatic advertising and a few kinds of very focused things. But now AI has arrived. How does that change media and marketing in three words or less?

Tim: In every way?

Ross: Got it.

Tim: The truth of it is different things for media, different things for marketing? Gosh, I tried to find where I can, the case is for optimism and positivity. And I guess, in the same way that horses and carts gave way to a thriving automobile industry. And it feels a bit like we might be at that stage for the media and certainly for communications agencies where they've got such big disruption coming along. And, of course, so many possibilities and new ways of doing things. But it feels like the entire business model on which people have planned their futures is wobbling underneath them right now. And they're going to have to hang on to that wobbly platform and find themselves a ladder somewhere. Because, you know, obviously, there'll be a way through to the other side, because there always is, but wow, we've never seen change, like..

Ross: Yeah, well, arguably, you know, major magazines have been pretty wobbly for a long time like this. There's no single year which hasn't had its damage. 

Tim: Yeah. I mean, absolutely. I mean, I've been a journalist since 1989. And the theme when I walked into my very first newsroom, well, firstly, and trained on a typewriter, manual typewriter for the first few months, but was it was just as the printing of the newspapers was digitized, a whole bunch of printers were in, in the process of being made redundant right then. So yes, we had this weird kind of battle of the humans against the computerization where, as the sort of protest these these printers, who knew they were doomed, but we're still at this stage, laying out the newspaper each day, with just put subtle sabotage in while they were having they're kind of they could see what was coming down the track. So you'd have to be very, very careful because things like the not in not guilty would get removed in articles, and he would become che and all of these subtle things, which were quite hard to spot on the final round of proofreading as the as as people went, when kicking and screaming into the night, and I, that was the printers and sadly, I think it might be the turn of some of the journalists and,

Ross: Oh, let's look into sort of media and marketing. But I mean, company, the theme of amplifying cognition. Alright, so journalists are super smart. And I've always said, you know, if you've got a journalistic training, you can do well in the world, because you're able to pull together information makes sense that will communicate well, you know, these are fundamental skills and will continue to be, but how can you know, good journalists today? Use AI? Or what is their relationship to AI? I mean, obviously, there's going to be a lot of AI reporting, but what are the complementary roles of good journalists in AI today? Oh, what could it be?

Tim: There's no one answer. Obviously, there's several great examples. And I suppose the one that's given me the most pause when I know, because if you want to think as a journalist, okay, things are going to be fine. I'm smart. I know my bait in a way that no, I have occurred, you know, I, you know, I've been writing about media and marketing for more than two decades now. So, so my, you know, my, my neural pathways are trained and possibly calcified, to see the world through the eyes of marketers, I suppose, because they're my constituency. So, you know, in my time In Asia, you know, I've, I keep telling myself, you know, I, you know, I'm specialized enough that I can probably go fine. But then I had real pause when. And we might talk about this more in a minute. But um, we created effectively a chatbot called timbre, which was based on a book I wrote a couple of years back called Media unmade. And then based on the content of the unmade newsletter, since we started that, and I just remember asking it, and it was an example of showing somebody something about the outdoor advertising industry. And the answer it gave, which obviously, had been based on my writing. But what really struck me was it, it got all of the obvious points I would have made, and then it made an extra connection about an outdoor company, I'd obviously written about, at some point, a small one, because it was featured, but I had forgotten all about. And it, it, it answered it just despite the fact that it's based entirely on my writing. So it's only training from that. It got there slightly better than me. So that kind of gave me pause a little bit. Um, so I think it doesn't entirely answer your question. Because you know, that's an example where it goes above and beyond, because, but yes, I guess in my super specialized niche, hey, yeah, it's really useful. Having a really specialized reference tool, trained on everything I've written, I can ask questions often. Give me that information.

Ross: Just it's just on that point, though. In that case, you were able to judge that what the machine created then was insightful. Yes, he doesn't generate a whole bunch of stuff. And there was probably another point of me, which wasn't very, which was boring, the new, new, new, which was insightful. And that's so that's, again, you know, there is a cycle where you can feed the machine, what comes out of the machine you can then build on? 

Tim: I think that's a very fair point. And I think, of course, the other thing is, you see the possibilities for journalism. So for instance, something that I've been researching as we speak, and I think it probably by the time this podcast goes up, will probably have already been published is that I've been looking, looking at the finances of an Industry Foundation. And they've published via the charities commission, they've they've, there's six or seven years worth of their data. And my instinct as a journalist is we're almost at the point now that rather than me having to laboriously open up each one of the balance sheets and manually put in the data and manually draw myself the kind of graphical representations, I can ask AI to do it. But what I noticed was when the moment came today, or over the last couple of days I was working on this, I realized I don't trust it yet to get it. Right. And I think that's the thing. We were still at that point. But for journalism. I'm not sure it would have shortened my journey, my time spent on that particular piece of work now, over time, I think it probably will, you know, I think quite quickly will be at the point where transactional stuff like, you know, writing about a company? Well, I think we're all there already there actually writes from that company's quarterly results or something that comes through, you can, you know, and the Ansel will sort of be spat out, you know, I kind of think that a lot of what we see on, you know, the basic stuff that comes out and Motley Fool and places like that, I'll be surprised if that isn't an element of AI. So, there are things like that. But then, of course, I guess the question is, how helpful is it if it's the transactional stuff that's happening?

Ross: So I think, you know, I always go, when judges come out, I mean, the first distinction I kind of made is intent. You know, humans have the intent. And so you thought, that's a foundation, which I would actually like to look into. And, you know, possibly at some point, the AI can surface that. I think, you know, it's again, that sort of thing is interesting. So that's, I think, a critical value of humans. And then you can say, well, if you can get there quicker, it's far better and picking up the emergent trends. And that in stacks of data is definitely one of its strongest use cases. Yeah,

Tim: Look, I think that's a great point, because, of course, um, where the intent came from was, what the journalist does is they need to have relationships and sources and it took having a drink with somebody and then saying, by the way, you should look at, you know, so because you need a starting point in the first place. So, yeah, you know, I definitely think that those tools will become increasingly useful.

Ross: I think it's and I think one of the, if you look at it from the positive perspective, you know, investigative journalism is fundamental, it's critical. And that's the biggest downside of, you know, lower budgets for newsrooms and journalists. But if the best investment of journalists is to have access to AI, where they can more readily surface and dig into stuff, then that's going to be good. Yeah,

Tim: I think so. And I suppose the point is, yes, we've got this crucial need for society for investigative journalism. And I, you know, I think one of the, the big worries for me is just the side effect of the rise of generative AI is if, as a user, as a member of the public, you can just ask a question of your, of whichever chat bot you happen to be talking to, and it gives you the answer without sending you to the website, then that breaks the entire business model of media. And there's no, there's no guarantee that a new business model emerges. So of course, something happens that potentially stops us from funding all of that, investigative journalism or certainly at the, you know, the current level. So it becomes another disruption along that kind of line of disruption. We've been Yeah, we've been saying for at least the last 30 years, if not more. Yeah, the

Ross: Yeah. The business models of media are a tricky question. And it's, but yeah, I think that there's whatever resources we have, we have to just be able to amplify those as much as possible. And, you know, not just churning out, or just a courier reviews of quarterly results, or soccer matches, which can be done pretty decently with.

Tim: Yeah, I mean, I think this is my issue at the moment is we have this sort of strange situation. So the little niche, I'm sort of in the trade press writing about the Media Marketing World, there are five or six titles. And sometimes if the same press release comes in, you'll see. I mean, we don't, we don't tend to do press releases. But you'll see in most other places that a version of that same story happens. Now, what's actually beginning to happen already is a PR person is using the assistance of a chatbot, or chat GPT or whatever, to create their press release in the first place. It's then arriving in the newsrooms now one of those titles, has already quite openly said for all of those transactional news stories, then are feeding them into AI to create a separate news story, which is almost a point of difference for them at the moment. But when every single one of those outlets is doing that we've got this bizarre system where AI is talking to AI and there they become a series of filters before they maybe just maybe reach a human. And I yeah, this particular publication, I think it's hugely Orwellian. They call this practice fast news, which just actually sent a shudder down my spine.

Ross: Yeah, while loop which, you know, again, goes back to the human role, which is where do we put the humans in the loop? And all this process? And so it's, there's going to be humans in the loop. We pray, and in which case, so you're working out where those are. But let's let's turn to marketing agencies, that have typically been profligate paid high salaries, and being all these masters of the universe, in them, and now, the lot of them are looking up and saying, oh, suddenly AI is doing things which are creative, and we were where we're supposed to be the greatest people. We've got creative directors. We've got executive creative directors, who get paid a lot for being creative and directing. And so what's the high level view of what happens to agencies, creative all I argued, well, supposedly creative agencies in the world of AI.

Tim: I think one of the big challenges for creative agencies and again, this certainly predates the rise of generative AI but was already on track is, so much of the business model of agencies has been. They haven't actually ever really been paid on the thing their best, which is the idea is almost always the idea is given away for free. And then the way they make their money is by charging the client for the production work and the hours and all of the things that go around it, that all of a sudden generative AI can do. Now, in the short term, I think for some agencies, that's even beneficial. You know, there are, there are agency bosses who talk to who aren't telling their clients how much they're already outsourcing, but charging head hours for. But that's a very temporary situation. You know, there are smart procurement people working with all of these big companies who next time they run a tender. And generally, you know, an agency might be agency of record for perhaps three or four years before the lease, there's a review, if not a full pitch, then those clients will have a really good understanding of what actually can be either taken in house by the client, or just expected to no longer be done by humans. So that that, that potentially breaks the business model in the same way it breaks the business model for media, which isn't to say that creative agencies don't have a huge potential to be the contributors to a brand's business, you know, there are, there's so much evidence that the right big creative idea helps people sell more cornflakes, or whatever it is. And that becomes even more important, and arguably, for a really big landscape of agencies, and probably too many agencies, the ones that will differentiate will be the ones who can still nail the big idea, not the ones who are kind of in that muddy middle that can be done by AI. But again, it just becomes the question of how do they get to that point where they're rewarded for that big idea that only a human could come up with?

Ross: Yeah, person structures. Now that applies across other professional services as well. And I think it's really interesting, looking broadly at professional services of all kinds, and the roles of humans plus AI and service delivery, and then, you know, relationships and client value, and then the similarities and differences between the professional services. But this, you know, this does come back to a point where, you know, we were discussing earlier around, you know, getting the, the young people exposed to the learnings and the growth, which enabled them to get to where they're able to create, yeah, be the genius who can shave that big idea? 

Tim: I think that is exactly the issue is, you can sort of think now, okay, you've got these really experienced executive creative directors who have a great idea. And then they can use these tools to make it and that's great, you know, for the Agency for them. But the question is, what happens when those people have retired, and the next level of big brain creatives haven't learned on the job learned on the tools are all of those things, which they, they just learned by osmosis, that gave them the ability to have the big idea. At the same time, I suppose I was chatting, I think back to my experience of perhaps 10 or 15 years ago, just going on a very basic video editing course. And I, you know, we, this is where I worked previously, when I worked at Mumbrella. And we used to put out a fair bit of video. And I never actually, in the day to day ended up doing any video editing myself, but it gave me a language and an understanding of things I could then ask for, you know, it's just things like, you know, I remember there was a video I was so pleased with, and it was just the really sounds so obvious when you say it, but for the, for that particular piece, just to edit the piece we were doing to the beat. Now, that's just such a really obvious thing to say, once you've, you've been through the process, but I would never have thought to suggest it if I hadn't got that language. And the thing is, now, you just have your piece of video and you throw it into there's myriad tools, and it would spit it out at the other end. And you probably would never just have that little idea or that little piece of direction. So I think that is the worry that that kind of the you know, if we don't if the Yeah, if AI saves us from needing to work on the building blocks, then are we ever gonna be in the position to put the roof on?

Ross: Well, I think that applies. Yeah, that's absolutely not just for agencies anymore. It's across all professional services, arguably, you know, almost any, any industry or any work at all. But going just coming back to this frame of Yeah, so humans plus AI so you've got let's say you got your super creative director or if you've got some pretty good one. So what specifically do you think of the roles in AI? Good can support these people to do their work better.

Tim: And I suppose, in part, you want to think about the different parts of that communication agency chain, because it feels like for instance, when it comes to, at the end of the story after the the ads have been created, the the planning and buying of the media is already moving in that direction of automation. Anyway, you know, obviously, we've seen the rise of programmatic advertising, which is starting to look a lot like a net negative as it happens, because an awful lot of advertising spend is disappearing somewhere along the way. But certainly the ability to plan and target in a really sophisticated way who the audience is, that becomes something which I could do so much better than any kind of old school planter, if it's done well. And if it's done in a kind of non fraudulent way. But yeah, it certainly feels that Yeah. That means that an awful lot of the manual labor part of planning and buying media is gone. I saw Martin Sorrell speaker mad Fest in London last year, about this time last year, actually, the way he just put it was he just listed the consequences just to throw a line, throw away a quarter of a million media planner jobs gone around the world, obviously. And and, and I think he's probably not wrong. On the creative side, so much is, you know, it's, it's, every element of production becomes cheaper. Now, obviously, there's been a lot of talk about what for instance, saw is coming through what that's going to be able to do when it comes to creating films and a prompt was playing with a tool just today, which probably wipes out jingle makers, you know, as an advertising jingle makers, where you can literally just say, sad country song about a chocolate bar, and it will churn something out for you. It might be terrible, it might not. But the point is, by its very nature, most advertising is average, and half of it is below average. And that's the stuff that's made by humans. And that's the stuff that is very replaceable. So it feels like the, the, the part that's easy to say, okay, I can say on that stick around as the the big budget, Big Idea stuff, but it's all that transactional stuff, the, you need to hit somebody with 10 subtly different variants of the same or have similar phone package, but you're targeting slightly different demographics or different price points or something, yes. And at the click of a button, you can create every version of our ad. All of those things that you know will make the production of the stream so much slicker. But of course, in the very process challenged the business model of making them in the first place as well.

Ross: So we've got the AI to complement the greater people. So this, this goes a little to the idea of how can I, you know, I suppose the demand elasticity? So is there a fixed amount of demand for quality marketing? And if so, then, you know, if it's a fixed amount, then certainly a lot of work that is done currently can be replaced by AI. But if there is more demand, if there is more potential to create more or more quality, then that's the potential for the future is, you know, do we have an unlimited amount of demand for quality, marketing and to your point, I mean, part of that is then personalization where you have a big idea and then you personalize that idea to every person on the planet or whatever it may be. And so that this applies across all work is how limited or unlimited is the demand for this work?

Tim: Well, look, it's a fascinating question. Let's I guess one of the orthodoxies of advertising is you get your brand building, which is kind of slightly nebulous, slightly hard to measure, but you also have your performance side of sales, you know, save 20% Do your sale now and you know, how many you gonna sell tomorrow or you'll know very quickly with your campaigns working. If we are just just for the point of this debate, just as, you know, marketers would shudder, well, let's just park the brand side for a moment, then if you sort of assume that with the performance side of advertising, you can measure ROI. And one of the things about this is you should be able to measure it quite well. Then of course, there could be an infinite demand or nearly infinite, if you can tell that every dollar you sell makes you more than $1 in profit at the other end for the extra products that you sell. So, so for that, that side of you, yes, there there can be a demand i i suppose the the question is that sustainable in the long term, when, in 510 20 years time, the wise old practitioners are gradually retiring. Is there actually anybody left who's got the training to work with those tools that that part or, or more to the point, if this is going on in every single industry, then we may have a lot of very?

Ross: Yeah, well, that's one of the macro questions, of course, is, you know, do we do good. Yeah. Though, all of this talk of amplifying, you know, increasing economic activity X over my years, is predicated on the fact that there is no distribution and jobs. I mean, personally, I tend to believe that there, we are more likely to have a happier scenario than a dire scenario. But, you know, that's kind of, you know, that's a long and different debate. But I'm sure you have been using the tools and all sorts of ways in your own work and I just love to hear about any. You know, what, what do you find AI useful for in your work in your life? How are the insights you've gained on what's useful and practical and makes you able to do More?

Tim: Yep

Ross: That's really interesting, you know, all these points around augmenting, you know, doing what you're doing, but then sort of making it richer and more engaging and you're able to use more so. So just to round out I mean, any prognostications on any of the sort of things we've been talking about Yeah, well, it's a little while ago, my niece, or your granny said, was embarking on a journalism career at school. She said, Okay, we're going to do journalism as you know, who I'm not sure about that. But anyway, she's been extraordinarily successful. So lots of myself that we can still be you and I can sort of say, Oh, I don't know about that. But we'll still carve wonderful careers forward and I still I just believe in journalism, I believe in what it does, I believe in what it creates, and yes, the business models of fraud and but you know, I'm, I'm an optimist. And so let's hold a glimmer of hope.

Tim: Yes.

Ross: Thank you so much for your time and your insights. Tim. It's been a great conversation.

The post Tim Burrowes on AI’s impact on media and marketing, evolving business models, and the possibilities for journalism (AC Ep45) appeared first on Humans + AI.