The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • EA - Holly Elmore on reducing rodent suffering and wild animal welfare by Karthik Palakodeti
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Holly Elmore on reducing rodent suffering and wild animal welfare, published by Karthik Palakodeti on June 12, 2023 on The Effective Altruism Forum.
    We discuss the issue of rodent suffering and the likelihood of rodent birth control kicking off, touch on what some of the challenges are along the way in promoting its usage by governments, and delve deep into some of the tricky questions people have about wild animal welfare such as the ethics of killing carnivorous animals and how to know if an intervention does more good than harm with the limited knowledge we have so far. We are joined by Holly elmore who currently is a Researcher at ReThink Priorities, an Effective Altruism Think Tank, where she researches wild animal welfare. She completed her PhD at Harvard in Evolutionary Biology where she was also the President of the Harvard University Effective altruism student group. She has written a lot about rodent welfare and what we should do to humanely solve the problem of rodent infestation.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org
    2 min
  • EA - Historical Global Health R&D “hits”: Development, main sources of funding, and impact by Rethink Priorities
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Historical Global Health R&D “hits”: Development, main sources of funding, and impact, published by Rethink Priorities on June 12, 2023 on The Effective Altruism Forum.
    Editorial note
    This report was produced by Rethink Priorities during December 2022 and January 2023. The project was commissioned and supported by Open Philanthropy, which does not necessarily endorse our conclusions.
    The report focuses on the development story and impact of some global health R&D ‘hits’ or ‘success stories’ of the last two decades: injectable artesunate, rectal artesunate, Coartem Dispersible, long-lasting insecticidal nets, and dolutegravir. We did not classify these developments as ‘hits’ ourselves, or compare them to any other treatments, but we were asked to look into them based on the expectation that they have saved or could save many lives, and could constitute some of the most successful treatments of the past couple of decades. We analyzed the steps that led to their development, with a particular focus on the roles different stakeholders played. Additionally, we estimated the number of deaths these products have averted to date and might avert in the future.
    We have tried to flag major sources of uncertainty in the report and are open to revising our views as more information becomes available.
    Executive summary
    In this report, we attempted to reconstruct the development story behind five global health R&D hits of the last two decades, four types of antimalarial treatments and one antiretroviral used for HIV, and estimate the number of deaths each of them has averted to date and might avert in total. The latter ranges from a few tens of thousands (for rectal artesunate) to a few million (for injectable artesunate and dolutegravir). You can find below two plots (Figures 1 and 2) summarizing our models’ estimates for the lives saved by each of these products since their introduction and until obsolescence (you can also find the plots and associated data here).
    Figure 1: Cumulative number of deaths averted by each product since its introduction
    Figure 2: Total number of deaths averted by each product until obsolescence
    In most of the examples studied, the drugs were developed through partnerships involving: (1) a pharmaceutical company that carried out the development of the product, (2) academics who ran the trials to show the product’s effects and/or compare it to the standard of care, and (3) philanthropic and public actors that provided funding for various stages. Philanthropic funds were also used in some cases to support local manufacturers and help them get WHO approval. One notable exception is dolutegravir, for which the development seems to have been entirely funded by industry.
    You can find here a table including the partners involved in the development of each of the products, and the cost and funders for different parts of the process where we were able to find the information. As noted above, philanthropic funds seem to have been pivotal to the development of all products except for dolutegravir. Even though the cost incurred by the pharmaceutical companies is unknown, and likely to be quite large, it seems as though these partnerships are quite frequent and necessary when developing products for diseases mainly affecting low and middle income countries.
    For all cases in which philanthropic funding played a significant role, the funds needed seem to have been in the range of a few million dollars to the low tens of millions. It should be noted that the total funds these organizations invested toward R&D of products related to these diseases is likely much larger, including funds devoted to these particular drugs through unsuccessful initiatives we were not able to find, and funds devoted to the development of other drugs that were unsuccessful (or just less succe...
    4 min
  • EA - What can superintelligent ANI tell us about superintelligent AGI? by Ted Sanders
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: What can superintelligent ANI tell us about superintelligent AGI?, published by Ted Sanders on June 12, 2023 on The Effective Altruism Forum.
    To what extent can humans forecast the impacts of superintelligent AGI?
    From one point of view, trying to understand superintelligence seems utterly intractable. Just as a dog or chimpanzee has little hope of comprehending the motivations and powers of humans, why should humans have any hope of comprehending the motivations and powers of superintelligence?
    But from another point of view, forecasting the impacts of superintelligence may yet be possible. The laws of reality that constrain us will similarly constrain any superintelligence. Even if a superintelligence achieves a more refined understanding of physics than us humans, it very likely won’t overturn laws already known. Thus, any inventions optimized against those physical laws, even if superior to our own, may end up looking familiar rather than alien.
    No matter how intelligent an AGI is, it will still be bound by physics. No matter how smart you are, you still must obey the law of conservation of energy. Just like us, an AGI wishing to affect the world will require an energy industry full of equipment to extract energy from natural sources. Just like us, its energy will have to come from somewhere, whether it’s the sun (solar, wind, biofuels, fossil fuels, hydro), the Earth (geothermal, nuclear), or the Moon (tidal, nuclear). Just like us, any heat engines will be limited by Carnot efficiency. Just like us, energy will need to be transported from where it is collected to where it is consumed, likely by electromagnetic fields in the presence of bound electrons (e.g., chemical fuels) or unbound electrons (e.g., electricity) or neither (e.g., lasers).
    If there are economies of scale, as there likely will be, that transportation will take place across networks with fractal network topologies, similar to our electric grids, roads, and pipelines. The physics of energy production are so constrained and so well understood that no matter what a superintelligence might build (even fusion electricity, or superconducting power lines, or wireless power), I suspect it will be something that humans had at least considered, even if our attempts were not as successful.
    One way to preview superintelligent AGI is to consider the superintelligent narrow AIs humanity has attempted to develop, such as chess AI.
    Lessons from chess AI: superintelligence is not omnipotence
    In 2017, DeepMind revealed AlphaZero. In less than 24 hours of (highly parallelized) training, it was able crush Stockfish, the reigning AI world chess champion. AlphaZero was trained entirely de novo, with no learning from human games and no human tuning of chess-specific parameters.
    AlphaZero is superhuman at chess. AlphaZero is so good at chess that it could defeat all of us combined with ease. Though the experiment has never been done, were we to assemble all the world’s chess grandmasters and give them the collective task of coming up with a single move a day to play against AlphaZero, I’d bet my life savings that AlphaZero would win 100 games before the humans won 1.
    From this point of view, AlphaZero is godlike.
    Its margin of strength over us is so great that even if the entire world teamed up, it could defeat all of us combined with ease
    It plays moves so subtle and counterintuitive that they are beyond the comprehension of the world’s smartest humans (or at least beyond the tautological comprehension of ‘I guess it wins because the computer says it wins’).
    ...but on the other hand, pay attention to all the things that didn’t happen:
    AlphaZero’s play mostly aligned with human theory—it didn’t discover any secret winning shortcuts or counterintuitive openings.
    AlphaZero rediscovered openings commonly played by humans ...
    9 min
  • AF - Explicitness by Tsvi Benson-Tilsen
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Explicitness, published by Tsvi Benson-Tilsen on June 12, 2023 on The AI Alignment Forum.
    [Metadata: crossposted from. First completed March 3, 2023.]
    Explicitness is out-foldedness. An element of a mind is explicit when it is available to relate to other elements when suitable.
    Thanks to Sam Eisenstat for related conversations.
    Note: The ideas of explicitness and inexplicitness require more explication.
    Explicitness and inexplicitness
    Elements can be more or less explicit, less or more inexplicit.
    (This statement wants to be unpacked.)
    In general, inexplicitness is the lack of explicitness, and explicitness is when elements that have good reason to be related, are able to be related. That is, when structure is explicit, it can be brought into relation with other structure when suitable.
    Structure is explicit when it's out-folded: when it already makes itself available (visible, applicable, informative, copyable, tunable, directable, modifiable, predictable, combinable, interoperable), so that nothing is hidden or blocked.
    An explicit element is an element high in explicitness, i.e. it can be brought into relation with other elements when suitable.
    Explicitizing
    Elements can become more explicit.
    By default, structure is fully inexplicit for a mind. That is, it's fully ectosystemic for the mind: it's not available for elements of the mind to relate to.
    Structure can be brought into explicitness.
    For example, these processes make structure more explicit: reflection, analysis, description, expression, joint-carving, separating, factoring, refactoring, modularization, indexing, interfacing, connecting, disentangling.
    The early stages of explicitizing involve incomplete or deficient participation——like a blind man touching an elephant's tail, or entering the outer regions of a nexus of reference. E.g., the relationship that the Ancient Greek mathematicians had to Cartesian algebraic geometry.
    A diagram:
    Examples
    An example of explicitizing also furnishes examples of inexplicitness (before the explicitizing) and explicitness (after the explicitizing), and likewise an example of inexplicitizing also furnishes examples of explicitness and inexplicitness.
    Classes of examples of explicitizing
    Internal sharing of elements. See here for examples of inexplicitness.
    Making an analogy. By grasping a partial isomorphism, the partially isomorphic aspects of the analogands can be transferred back and forth between them.
    Putting a word to something. When the word comes up, the element is accessed, and vice versa. That helps different contexts in which the element is relevant communicate with each other through the unfolding of the element.
    Refactoring code. Separating concerns A and B renders code for dealing with just A useful for tasks that deal with A but not B, whereas the unseparated code might for example assume the existence of B (e.g. as an argument or global variable). The separation makes explicit the dependence and non-dependence of code on A or B. Or for example rewriting a function so that it accepts a standard rather than non-standard input format, so that elements expecting a function to accept standard input can easily relate to the function.
    Deduction. Drawing out the implications of an element makes the element available for comparison with other elements and makes the element available to recommend action.
    Writing things down.
    Expressing and recording something in shared language makes it available to others.
    Storing something in memory makes it available to your future self.
    Abbreviating something makes synopsis and large-scale modification more feasible. For example, mathematical notation makes big thoughts viewable all together, and makes algebra more efficient.
    Attaching abbreviations to an element makes that element easier to find.
    Drawing a picture makes somet...
    25 min
  • EA - Effective altruism organizations should avoid using “polarizing techniques” by Joey
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Effective altruism organizations should avoid using “polarizing techniques”, published by Joey on June 12, 2023 on The Effective Altruism Forum.
    TL;DR: The EA movement should not use techniques that alienate people from the EA community as a whole if they do not align with a particular subgroup within the community. These approaches not only have an immediate negative impact on the EA community, but also have long-term repercussions on the sub-community utilizing them. Right now, the EA movement uses these sorts of tactics too often.
    People connect with the EA movement through many different channels, and often encounter sub-communities before they have a full understanding of the movement and the wide variety of opinions and viewpoints within it. These sub-communities can sometimes make the mistake of using "polarizing techniques". By this, I mean strategies that alienate people or burn bridges with the broader community. This could be from pushing a sub-perspective too hard, or being aggressively dismissive of other views.
    An example of this might be if I met a talented person at a party and they said they wanted to change to an impactful career, but had never heard of EA. If I then proceeded to aggressively push founding a charity through Charity Entrepreneurship (the organization) as a career path to them, to the point where they got turned off of EA altogether if they don’t come on board with my claims, I would consider that a polarizing approach: either they choose charity entrepreneurship as a path, or they don’t engage with effective altruism at all. Note that in the short term, all Charity Entrepreneurship really measures impact-wise is how many great charities get started, so a good person going into policy due to me connecting them to Probably Good means nothing to our organizational impact. Taken to an extreme, it might be worth pushing quite hard if I think that founding nonprofits is many times more important as a career path than policy.
    However, I think this style hurts both the community and Charity Entrepreneurship long-term.
    This phenomenon occurs across a diverse range of people, both in terms of funding and career transitions. Most often, it revolves around cause prioritization. It can be disappointing when someone does not share your enthusiasm for your preferred causes, but there is still a lot of value in directing them to the most impactful path they would in fact consider pursuing.
    Why it Hurts the Community
    The clearest way this technique is damaging, is that turning off someone from one part of the community often demotivates them to engage positively in other parts of the community. It makes them more likely to become an active critic instead of a neutral or contributing member of a different sub-community, or to the philosophy of effective altruism as a whole.
    Different sub-communities look for different types of people and resources. It's difficult for one person to have a bird's eye view on all sub-communities within EA, and it’s easy to overvalue your own community's certain needs or strengths. On numerous occasions, I have witnessed instances where one sub-community dismisses individuals possessing skills that would be immensely valuable in another segment of the community. It seems worth exercising a degree of modesty in determining who exactly is a good fit for the community as a whole.
    The EA community is diverse in its viewpoints, cause areas, and approaches, and this diversity brings many benefits. The most promising career, the most neglected cause area, or what seems to be missing most from the community has changed many times over the years. Even if someone has a comprehensive understanding of all EA sub-communities, accurately predicting their long-term needs can be challenging. It is not uncommon for a community to inadvert...
    6 min
  • EA - Family Empowerment Media: track record, cost-effectiveness, and main uncertainties by Rethink Priorities
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Family Empowerment Media: track record, cost-effectiveness, and main uncertainties, published by Rethink Priorities on June 12, 2023 on The Effective Altruism Forum.
    Editorial note and disclaimer
    This report assesses the nonprofit Family Empowerment Media (FEM). The project was commissioned and supported by a donor who wishes to remain anonymous. It was conducted in February 2023 over two weeks.
    This report attempts to estimate the cost-effectiveness of FEM, mainly relying on two existing cost-effectiveness analyses (from GiveWell and Founders Pledge). We were asked to look into the cost-effectiveness of the organization in an unbiased way, and we started the research without a prior for whether this organization would be cost-effective or not. In this report, we express cost-effectiveness as a function of the cost-effectiveness of cash transfers (like the original models mentioned above). We did not analyze the cost-effectiveness of any other family planning interventions, so we cannot and do not make any claims about how FEM compares to other family planning interventions.
    In addition to looking at the cost-effectiveness of FEM’s program, we describe their track record to date, and try to determine the main uncertainties about the program and its (cost-)effectiveness.
    We relied on four expert interviews to attempt to answer some of the questions: We interviewed Anna Christina Thorsheim (Executive Director and cofounder of FEM), Andrew Martin (Senior Research Analyst at GiveWell), Rosie Bettle (Applied Researcher at Founders Pledge), and Dr. Mukhtar Muhammad (medically trained expert in social and behavior change communication with a historical focus on family planning programming).
    We have tried to flag major sources of uncertainty in the report and are open to revising our views as more information becomes available.
    We would like to mention that Melanie Basnak previously worked at Charity Entrepreneurship (CE), the nonprofit organization that incubated FEM. Melanie did not overlap with Anna Christina Thorsheim during her time at CE and believes she oversaw this research in an unbiased manner.
    Executive summary
    In a randomized controlled trial in Burkina Faso from 2016-2018, a family planning mass media campaign led by the nonprofit Development Media International (DMI) led to a 5.9 percentage point (20%) increase in the modern contraceptive prevalence rate (mCPR), leading the study authors to estimate a cost effectiveness of $7.70 per couple-year protection, albeit with uncertainty. The nonprofit Family Empowerment Media (FEM) ran a pilot family planning mass media campaign in Kano, Nigeria, and they observed a 6 percentage point (75%) increase in mCPR using less rigorous pre- and post-pilot data comparisons. GiveWell conducted a cost-effectiveness analysis (CEA) of DMI’s intervention, finding it to be 1.1-5.4x the cost-effectiveness of cash transfers, depending on the context. FEM altered the model to pertain to its pilot intervention in Kano, finding that their intervention was ~27x the cost-effectiveness of cash transfers.
    Founders Pledge conducted a separate CEA, finding a cost-effectiveness of ~22x that of cash transfers, complete with a full (unpublished) report.
    We used the original models as a starting point to conduct our own cost-effectiveness estimation, tweaking certain values. In most cases, our adjustments would tend to increase these existing cost-effectiveness estimates, so we are confident that the cost-effectiveness of FEM’s intervention beats that of cash transfers, and we believe that their cost-effectiveness might be higher than GiveWell’s and Founders Pledge’s estimates. However, we would like to see several remaining uncertainties resolved before we would feel confident putting a bound on cost-effectiveness, and we discuss these uncertainti...
    5 min
  • LW - Critiques of prominent AI safety labs: Conjecture by Omega.
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Critiques of prominent AI safety labs: Conjecture, published by Omega. on June 12, 2023 on LessWrong.
    Cross-posted from the EA Forum. See the original here. Internal linking has not been updated for LW due to time constraints and will take you back to the original post.
    In this series, we consider AI safety organizations that have received more than $10 million per year in funding. There have already been several conversations and critiques around MIRI (1) and OpenAI (1,2,3), so we will not be covering them. The authors include one technical AI safety researcher (>4 years experience), and one non-technical community member with experience in the EA community. We’d like to make our critiques non-anonymously but believe this will not be a wise move professionally speaking. We believe our criticisms stand on their own without appeal to our positions. Readers should not assume that we are completely unbiased or don’t have anything to personally or professionally gain from publishing these critiques. We’ve tried to take the benefits and drawbacks of the anonymous nature of our post seriously and carefully, and are open to feedback on anything we might have done better.
    This is the second post in this series and it covers Conjecture. Conjecture is a for-profit alignment startup founded in late 2021 by Connor Leahy, Sid Black and Gabriel Alfour, which aims to scale applied alignment research. Based in London, Conjecture has received $10 million in funding from venture capitalists (VCs), and recruits heavily from the EA movement. We shared a draft of this document with Conjecture for feedback prior to publication, and include their response below. We also requested feedback on a draft from a small group of experienced alignment researchers from various organizations, and have invited them to share their views in the comments of this post.
    We would like to invite others to share their thoughts in the comments openly if you feel comfortable, or contribute anonymously via this form. We will add inputs from there to the comments section of this post, but will likely not be updating the main body of the post as a result (unless comments catch errors in our writing).
    Key Takeaways
    For those with limited knowledge and context on Conjecture, we recommend first reading or skimming the About Conjecture section.
    Time to read the core sections (Criticisms & Suggestions and Our views on Conjecture) is 22 minutes.
    Criticisms and Suggestions
    We think Conjecture’s research is low quality (read more).
    Their posts don’t always make assumptions clear, don’t make it clear what evidence base they have for a given hypothesis, and evidence is frequently cherry-picked. We also think their bar for publishing is too low, which increases the signal to noise ratio. Conjecture has acknowledged some of these criticisms, but not all (read more).
    We make specific critiques of examples of their research from their initial research agenda (read more).
    There is limited information available on their new research direction (cognitive emulation), but from the publicly available information it appears extremely challenging and so we are skeptical as to its tractability (read more).
    We have some concerns with the CEO’s character and trustworthiness because, in order of importance (read more):
    The CEO and Conjecture have misrepresented themselves to external parties multiple times (read more);
    The CEO’s involvement in EleutherAI and Stability AI has contributed to race dynamics (read more);
    The CEO previously overstated his accomplishments in 2019 (when an undergrad) (read more);
    The CEO has been inconsistent over time regarding his position on releasing LLMs (read more).
    We believe Conjecture has scaled too quickly before demonstrating they have promising research results, and believe this will make it harder for...
    55 min
  • EA - Critiques of prominent AI safety labs: Conjecture by Omega
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Critiques of prominent AI safety labs: Conjecture, published by Omega on June 12, 2023 on The Effective Altruism Forum.
    Crossposted to LessWrong.
    In this series, we consider AI safety organizations that have received more than $10 million per year in funding. There have already been several conversations and critiques around MIRI (1) and OpenAI (1,2,3), so we will not be covering them. The authors include one technical AI safety researcher (>4 years experience), and one non-technical community member with experience in the EA community. We’d like to make our critiques non-anonymously but believe this will not be a wise move professionally speaking. We believe our criticisms stand on their own without appeal to our positions. Readers should not assume that we are completely unbiased or don’t have anything to personally or professionally gain from publishing these critiques. We’ve tried to take the benefits and drawbacks of the anonymous nature of our post seriously and carefully, and are open to feedback on anything we might have done better.
    This is the second post in this series and it covers Conjecture. Conjecture is a for-profit alignment startup founded in late 2021 by Connor Leahy, Sid Black and Gabriel Alfour, which aims to scale applied alignment research. Based in London, Conjecture has received $10 million in funding from venture capitalists (VCs), and recruits heavily from the EA movement. We shared a draft of this document with Conjecture for feedback prior to publication, and include their response below. We also requested feedback on a draft from a small group of experienced alignment researchers from various organizations, and have invited them to share their views in the comments of this post.
    We would like to invite others to share their thoughts in the comments openly if you feel comfortable, or contribute anonymously via this form. We will add inputs from there to the comments section of this post, but will likely not be updating the main body of the post as a result (unless comments catch errors in our writing).
    Key Takeaways
    For those with limited knowledge and context on Conjecture, we recommend first reading or skimming the About Conjecture section.
    Time to read the core sections (Criticisms & Suggestions and Our views on Conjecture) is 22 minutes.
    Criticisms and Suggestions
    We think Conjecture’s research is low quality (read more).
    Their posts don’t always make assumptions clear, don’t make it clear what evidence base they have for a given hypothesis, and evidence is frequently cherry-picked. We also think their bar for publishing is too low, which decreases the signal to noise ratio. Conjecture has acknowledged some of these criticisms, but not all (read more).
    We make specific critiques of examples of their research from their initial research agenda (read more).
    There is limited information available on their new research direction (cognitive emulation), but from the publicly available information it appears extremely challenging and so we are skeptical as to its tractability (read more).
    We have some concerns with the CEO’s character and trustworthiness because, in order of importance (read more):
    The CEO and Conjecture have misrepresented themselves to external parties multiple times (read more);
    The CEO’s involvement in EleutherAI and Stability AI has contributed to race dynamics (read more);
    The CEO previously overstated his accomplishments in 2019 (when an undergrad) (read more);
    The CEO has been inconsistent over time regarding his position on releasing LLMs (read more).
    We believe Conjecture has scaled too quickly before demonstrating they have promising research results, and believe this will make it harder for them to pivot in the future (read more).
    We are concerned that Conjecture does not have a clear plan for balancing profit an...
    55 min
  • LW - [Linkpost] Large Language Models Converge on Brain-Like Word Representations by Bogdan Ionut Cirstea
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: [Linkpost] Large Language Models Converge on Brain-Like Word Representations, published by Bogdan Ionut Cirstea on June 11, 2023 on LessWrong.
    This is a linkpost for
    One of the greatest puzzles of all time is how understanding arises from neural mechanics. Our brains are networks of billions of biological neurons transmitting chemical and electrical signals along their connections. Large language models are networks of millions or billions of digital neurons, implementing functions that read the output of other functions in complex networks. The failure to see how meaning would arise from such mechanics has led many cognitive scientists and philosophers to various forms of dualism -- and many artificial intelligence researchers to dismiss large language models as stochastic parrots or jpeg-like compressions of text corpora. We show that human-like representations arise in large language models. Specifically, the larger neural language models get, the more their representations are structurally similar to neural response measurements from brain imaging.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • LW - I can see how I am Dumb by Johannes C. Mayer
    Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: I can see how I am Dumb, published by Johannes C. Mayer on June 10, 2023 on LessWrong.
    When I'm talking to somebody, sometimes I lose the conversational thread. Or sometimes I feel like there is this thing I want to say that seems relevant now, but I just can't remember it.
    Or maybe I'm trying to solve a particular problem. I throw myself again and again at the problem, but it just won't budge. And then after some long amount of time, possibly hours, I realize that the solution was extremely simple. And I just failed to execute the right kind of solution-finding algorithm that would have found this very simple solution quickly.
    I would expect that people with more intelligence, perform better in these domains. They have probably an easier time remembering and retaining the right things. Well, that alone might be sufficient to explain a large chunk of what makes a more intelligent person able to perform better.
    If you remember the right things quickly that are relevant in the moment, and if you can keep track of more things in your head at the same time without losing track of what these things were, then that might account for a large chunk of how an intelligent person is better at performing any particular thing.
    The core point here is that I think everybody, even somebody who would be much smarter than me, can see various failure modes in their own cognition and realize that they might be just so fundamental that there is no direct way of changing them.
    I'm pretty sure that at some level what sorts of things your brain spits out into your consciousness and how useful that information is in the given situation, is something that you can't fundamentally change. I expect this to be a hard-coded algorithm, and I expect there to be many such hard-coded cognitive processes that can't be changed (at least not in major ways).
    The cognitive improvements that you can apply will be at a higher level. To me, it seems that is what much of the Sequences are about. You can understand that there is something like the sunk cost fallacy, and understanding what it is, allows you to train yourself to recognize when you fall pray to it (Though that is a separate step from understanding what it is that you actually need to do to get most of the benefit).
    And the way you would do this is by for example using TAPs. In a sense, it seems that tabs are a way to install a very small hook into your brain in the programming sense. My current model of it is that you install a little watcher program that watches your sensory input streams and your internal model of the world. And then, when it detects a specific pattern, it triggers the execution of another algorithm. The interesting thing is that if you do this well, then all of this will become subconscious. So it's not that you can't change your subconscious algorithms. TABs are a way to install tiny new subconsciously executed algorithms into your brain.
    So let me make give an example of an algorithm that I think is largely unchangeable. Let's do an experiment. Please follow the bolded instructions:
    Imagine a car in your mind's eye.
    Now, your brain will have brought probably some specific car to mind. Maybe it's a 911 Porsche, the Cybertruck, or another Tesla. The point is that based on you reading the word car your brain pulled a lot more information out of the depths of your brain, such as an image or a "feeling of carness". How is your mind doing this? You read the word car and your mind produces some qualia that is associated with the concept of car.
    Now think of a car part.
    Now what did you imagine? A steering wheel, a door, an engine, a cylinder, a wheel, a windshield, an antenna? Notice that there was one thing that came to mind first and then maybe another thing. But how did your brain generate that specific thing? Why did you for exa...
    8 min