The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • AF - Towards deconfusing wireheading and reward maximization by leogao
    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: Towards deconfusing wireheading and reward maximization, published by leogao on September 21, 2022 on The AI Alignment Forum.
    TL;DR: A response to “Reward is not the optimization target,” and to a lesser extent some other shard theory claims. I agree with the headline claim but I disagree with some of the implications drawn in the post, especially about wireheading and the "executing behaviors downstream of past reinforcement" framing.
    My main claims:
    Not only does RL not, by default, produce policies which have reward maximization as their behavioral objective, but in fact I argue that it is not possible for RL policies to care about “reward” in an embedded setting.
    I argue that this does not imply that wireheading in RL agents is impossible, because wireheading does not mean “the policy has reward as its objective”. It is still possible for an RL agent to wirehead, and in fact, is a high probability outcome under certain circumstances.
    This bears a clean analogy to the human case, viewing humans as RL agents; there is no special mechanism going on in humans that is needed to explain why humans care about things in the world.
    A few notes on terminology, which may or may not be a bit idiosyncratic:
    I define an RL policy to refer to a function that takes states and outputs a distribution over next actions.
    I define an RL agent to be an RL policy combined with some RL algorithm (i.e PPO, Q-learning) that updates the policy on the fly as new trajectories are taken (i.e I mostly consider the online setting here).
    The objective that any given policy appears to optimize is its behavioral objective (same definition as in Risks from Learned Optimization).
    The objective that the agent optimizes is the expected value the policy achieves on the reward function (which takes observations and outputs reals).
    A utility function takes a universe description and outputs reals (i.e it cares about "real things in the outside world", as opposed to just observations/reward, to the extent that those even make sense in an embedded setting).
    I mostly elide over the mesaoptimizer/mesacontroller distinction in this post, as I believe it’s mostly orthogonal.
    Thanks to AI_WAIFU and Alex Turner for discussion.
    Outside world objectives are the policy’s optimization target
    First, let us think about the mechanics of an RL agent. Each possible policy in policy-space implements a behavioral objective, and as a result has some behavior which may or may not result in trajectories which receive high reward. The RL algorithm performs optimization over the space of possible policies, to find ones that would have received high reward (the exact mechanism depends on the details of the algorithm). This optimization may or may not be local. The resulting policies do not necessarily “care” about “reward” in any sense, but they will have been selected to achieve high reward.
    Now, consider the same RL agent in an embedded setting (i.e the agent runs on a computer that is part of the environment that the agent acts in). Then, because the sensors, reward function, RL algorithm, etc are all implemented within the world, there exist possible policies that execute the strategies that result in i.e the reward function being bypassed and the output register being set to a large number, or the sensors being tampered with so everything looks good. This is the typical example of wireheading. Whether the RL algorithm successfully finds the policies that result in this behavior, it is the case that the global optima of the reward function on the set of all policies consists of these kinds of policies. Thus, for sufficiently powerful RL algorithms (not policies!), we should expect them to tend to choose policies which implement wireheading.
    There are in fact many distinct possible policies with different behavioral objective...
    8 min
  • LW - Alignment Org Cheat Sheet by Akash
    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: Alignment Org Cheat Sheet, published by Akash on September 20, 2022 on LessWrong.
    Epistemic Status: Exploratory.
    Epistemic Effort: ~6 hours of work put into this document.
    Contributions: Akash wrote this, Thomas helped edit + discuss. Unless specified otherwise, writing in the first person is by Akash and so are the opinions. Thanks to Olivia Jimenez for comments. Thanks to many others for relevant conversations.
    There have been a few attempts to summarize the AI alignment research landscape. Many of them are long and detailed.
    As an exercise, I created a “cheat sheet” describing the work of researchers, research organizations, and training programs. The goal was to summarize key people/orgs in just one sentence.
    Note that I do not aim to cover everyone; I focus on the ones that I think are most promising or popular, with a bias towards ones that I know the most (For longer summaries/analyses, I recommend the ones by Thomas Larsen & Eli Lifland and Nate Soares).
    Here’s my alignment org cheat sheet:
    Researchers & Research Organizations
    Aligned AI: Let's understand all of the possible ways a model could generalize out-of-distribution and then figure out how to select the way(s) that are safe. (see Why I'm co-founding Aligned AI) (see also Thomas’s critique of their plan here).
    Anthropic: Let’s interpret large language models by understanding which parts of neural networks are involved with certain types of cognition, and let’s build larger language models to tackle problems that will emerge as we get closer to AGI.
    ARC (Alignment Research Center): Let’s interpret an AGI by building a reporter which a) honestly answers our questions and b) is smart enough to translate the AGI’s thoughts to us in a way that we can understand (see ELK report).
    Conjecture: Let’s focus on bets that make sense in worlds with short timelines, foster an environment where people can think about new and uncorrelated ways to solve alignment, create infrastructure that allows us to use large language models to assist humans in solving alignment, develop simulacra theory to understand large language models, build and interpret large language models, and maintain a culture of information security.
    DeepMind Alignment Team (according to Rohin Shah): Let’s contribute to a variety of projects and let our individual researchers pursue their own directions rather than having a unified agenda (note that DeepMind is also controversial for advancing AI capabilities; more here).
    Eliezer & Nate: Let’s help others understand why their alignment plans are likely to fail, in the hopes that people understand the problem better, and are able to take advantage of a miracle (see List of Lethalities & Why Various Plans Miss the Hard Parts).
    Encultured: Let’s create consumer-facing products that help AI safety researchers, and let’s start with a popular video game that serves as a “testing ground” where companies can deploy AI systems to see if they cause havoc (bad) or do things that human players love (good).
    Evan Hubinger: Let’s understand how we might get deceptive models, develop interpretability tools that help us detect deception, and develop strategies to reduce the likelihood of training deceptive models (see Risks from Learned Optimization and A Transparency and Interpretability Tech Tree).
    John Wentworth: Let’s become less confused about agency/alignment by understanding if human concepts are likely to be learned “naturally” by AIs (natural abstractions), understanding how selection pressures work and why they produce certain types of agents (selection theorems), and broadly generating insights that make alignment less pre-paradigmatic (see The Plan).
    MIRI: Let’s support researchers who are working on a variety of research agendas, which makes it really hard to summarize what we’re doing in one sentence, and...
    8 min
  • EA - Optimizing seed:pollen ratio to spread ideas by Holly Elmore
    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: Optimizing seed:pollen ratio to spread ideas, published by Holly Elmore on September 20, 2022 on The Effective Altruism Forum.
    Cross-posted from my blog.
    As an EA organizer when I was a grad student at Harvard, I developed an implicit model of community organizing at a university that is complimentary to the funnel model of Center for Effective Altruism (CEA) that was super popular a few years ago. I call mine “optimizing the seed:pollen ratio”. I’m not going to justify outreach as a strategy or advocate for any specific means of outreach—the point is to share my model of outreach efficacy. It’s a fun bonus that my model is well-explained by analogy to a biology concept, which I hope to explain well as a secondary objective.
    The funnel model
    CEA says, “We are trying to build a community, and one aspect of this project is encouraging people to become more deeply engaged with the community. The funnel metaphor helps us to think about the appropriate goals and audiences for our different projects.” The funnel model is focused on the deliberate composition of the community— how many people are entering the funnel, and what share of the community is at what part of the funnel at any given time. It also suggests that movement through the funnel happens in stages, from outer to middle to core, and there’s both investment from the community at every stage and friction to progressing from one stage to another.
    When this model was getting a lot of buzz, there were many discussions about what part of the funnel to be focusing on. This was happening during CEA’s big pivot toward endorsing longtermism as a fundamental tenet of EA. With that, many were arguing that there should much more focus on core EAs and the core-EA-development-pipeline, and much less emphasis on the mid- to casual level of community involvement, because most of the value of EA in the long run will be “in the tails” of the distribution, from original work, not from safer but lower value bets like getting average individuals to donate money. This view basically won and is dominant in EA today.
    Seed:pollen ratio
    (Source)
    Now for an interlude on plant reproductive strategies.1 This^ is called a “perfect flower” because it has both male and female reproductive organs. The female reproductive organ is the carpal, and it makes seeds. The male reproductive organs are the anthers and they contain pollen. Seeds are larger and contain cytoplasm, the organelles mitochondria and chloroplasts, and a reserve of nutrients. Pollen is much smaller per grain and, like sperm, basically only contains chromosomes. Whereas seeds can be as big as a coconut (which is a seed!), the biggest pollen grains are only 2.5mm long. In terms of energy and resources, seeds are much more costly and pollen is much cheaper.
    You may be familiar with human evolutionary psychology ideas about male vs. female reproductive strategies. What’s interesting about hermaphroditic plant species is that individuals are not committed to one strategy or the other, but can vary the amount of investment they put into seeds vs. pollen, either over evolutionary time or as a plastic response to environmental conditions. Per expected offspring, seeds are a safer bet. The average seed is FAR more likely to grow into a plant than the average pollen grain, which makes sense because pollen outnumbers seeds by a factor of 100-1000x (depending on the species— this number is for Cannabis sativa, because for some reason it’s very easy to get estimates of these quantities online for Cannabis :P) But you can make a LOT of pollen for the cost of one seed.
    Pollen can be carried by pollinators or the wind (or a number of other clever strategies) to sometimes vast distances (a genetically modified pollen grain fertilized a grass seed 21 kilometers away, using nothing but the wind ...
    11 min
  • EA - The Mistakes of Focusing on Student Outreach in EA by DavidNash
    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: The Mistakes of Focusing on Student Outreach in EA, published by DavidNash on September 20, 2022 on The Effective Altruism Forum.
    Summary
    There has been a surge in EA outreach towards university and high school students.
    Generally a higher proportion of students show interest in EA compared to professionals, but just because it's easier to get people interested does not mean that it's the most impactful use of time.
    This is potentially a mistaken use of limited community building resources and could lead to an imbalance in the age distribution of people interested in EA and bottleneck the future growth of EA or worse, lock EA into being a permanent youth movement.
    To improve this situation, EA movement building organisations can make changes in the actions they promote for students. Students interested in community building could work in more neglected areas.
    Why has EA community building been focused on students?
    It seems that this is a historical coincidence rather than a deliberately chosen path. EA was founded in Oxford, mainly by students, and their initial community building was focused on Oxford, Cambridge and similar universities. Over the last ten years there has been multiple iterations, lessons learned and funding for university group organising. So it isn’t that surprising that there are lots of stories of people who heard about EA in university and then went on to get involved in EA.
    This is in contrast to community building efforts for professional, cause and interest groups which have started to become more organised in the last couple of years, with some of them having full time organisers for the first time this year (EA Consulting Network, UK Civil Service). There has been a lot less time to see the potential outcomes of this work in direct comparison to university group organising.
    Some people have said that they focus on students because they show more interest in EA than professionals. This is likely true true, it definitely seems that it’s easier for students to attend events and get more involved with EA sooner. I think even if students are 10 times more likely to get involved, this may be a mistaken strategy to pursue.
    It's a mistake to focus on students
    Professionals have more experience, influence and networks. As EA is growing there is more need for people who have experience managing people and supporting larger organisations. It can take a long time to wait for students to graduate and gain those skills whilst maintaining interest in EA (this may be especially relevant for people with shorter AI timelines).
    If we keep focusing on students we won't be able to fix the mentorship gap. At the moment we have lots of people looking for guidance and very few people able to provide the right level of support. Below there are two age distributions to show how EA currently seems vs what may be a more ideal distribution.
    The average age in EA used to be going up one year every one or two years. The EA survey from 2020 shows that the trend had reversed with the mean and median both going down one year. According to 80,000 Hours the age that people tend to have the most impact is usually 36+.
    EA can give off the impression of being a youth movement. This can lead to professionals bouncing when they encounter EA and assuming that it is not a place meant for them. Even professionals who have been interested for a while may find it hard working directly at EA organisations if there is no consideration for having to support a family, healthcare and uncertainty around funding.
    We shouldn't ignore students though
    I don’t think we should stop student community building, this is more about what we should do on the margin. There is a lot to be gained by having organisers at the top 20-50 universities. There are also plenty of ways that talented young people ca...
    5 min
  • EA - $13,000 of prizes for changing our minds about who to fund (Clearer Thinking Regrants Forecasting Tournament) by spencerg
    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: $13,000 of prizes for changing our minds about who to fund (Clearer Thinking Regrants Forecasting Tournament), published by spencerg on September 20, 2022 on The Effective Altruism Forum.
    We have $13,000 of prizes you can win for our Clearer Thinking Regrants Forecasting Tournament on Manifold Markets! You can win money by:
    (1) providing us with arguments or evidence that changes our minds about which of the 28 finalist projects to fund or how much to fund each project
    (2) being one of the 20 most accurate forecasters at predicting which projects we end up selecting for funding in the Clearer Thinking Regrants program
    You can learn more about the tournament (including all terms and conditions) here:
    And you can begin forecasting here:
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • LW - You Are Not Measuring What You Think You Are Measuring by johnswentworth
    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: You Are Not Measuring What You Think You Are Measuring, published by johnswentworth on September 20, 2022 on LessWrong.
    Eight years ago, I worked as a data scientist at a startup, and we wanted to optimize our sign-up flow. We A/B tested lots of different changes, and occasionally found something which would boost (or reduce) click-through rates by 10% or so.
    Then one week I was puzzling over a discrepancy in the variance of our daily signups. Eventually I scraped some data from the log files, and found that during traffic spikes, our server latency shot up to multiple seconds. The effect on signups during these spikes was massive: even just 300 ms was enough that click-through dropped by 30%, and when latency went up to seconds the click-through rates dropped by over 80%. And this happened multiple times per day. Latency was far and away the most important factor which determined our click-through rates.
    Going back through some of our earlier experiments, it was clear in hindsight that some of our biggest effect-sizes actually came from changing latency - for instance, if we changed the order of two screens, then there’d be an extra screen before the user hit the one with high latency, so the latency would be better hidden. Our original interpretations of those experiments - e.g. that the user cared more about the content of one screen than another - were totally wrong. It was also clear in hindsight that our statistics on all the earlier experiments were bunk - we’d assumed that every user’s click-through was statistically independent, when in fact they were highly correlated, so many of the results which we thought were significant were in fact basically noise.
    Main point of this example: we were not measuring what we thought we were measuring. We thought we were testing hypotheses about what information the user cared about, or what order things needed to be presented in, or whether users would be more likely to click on a bigger and shinier button. But in fact, we were mostly measuring latency.
    When I look back on experiments I’ve run over the years, in hindsight the very large majority of cases are like the server latency example. The large majority of the time, experiments did not measure what I thought they were measuring. I’ll call this the First Law of Experiment Design: you are not measuring what you think you are measuring.
    Against One-Bit Experiments
    A one-bit experiment is an experiment designed to answer a yes/no question. It’s the prototypical case from high school statistics: which of two mouse diets results in lower bodyweight? Which of two button designs on a website results in higher click-through rates? Does a new vaccine design protect against COVID better than an old design (or better than no vaccine at all)? Can Muriel Bristol tell whether milk or tea was added first to her teacup? Will a neural net trained to navigate to a coin at the end of a level still go to the coin if it’s no longer at the end of a level? Can a rat navigate a maze just by smell?
    There’s an obvious criticism of such experiments: at best, they yield one bit of information. (Of course the experimenter probably observes a lot more than one bit of information over the course of the experiment, but usually people are trained to ignore most of that useful information and just report a p-value on the original yes/no question.) The First Law of Experiment Design implies that the situation is much worse: in the large majority of cases, a one-bit experiment yields approximately zero information about the thing the experimenter intended to measure. It inevitably turns out that mouse bodyweight, or Muriel Bristol’s tea-tasting, or a neural net’s coinrun performance, in fact routes through something entirely different from what we expected.
    Corollary To The First Law: If You Are Defin...
    14 min
  • AF - Doing oversight from the very start of training seems hard by Peter Barnett
    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: Doing oversight from the very start of training seems hard, published by Peter Barnett on September 20, 2022 on The AI Alignment Forum.
    TLDR: We might want to use some sort of oversight techniques to avoid inner misalignment failures. Models will be too large and complicated to be understandable by a human, so we will use models to oversee models (or help humans oversee models). In many proposals this overseer model is an ‘amplified’ version of the overseen model. Ideally you do this oversight throughout all of training so that the model never becomes even slightly misaligned without you catching it. You can’t oversee on a close to initialized model because it’s just a random soup of tensors. You also can’t use this close to initialized model to help you do oversight because it’s too dumb. We will probably need to do some amount of pretraining to make our models good enough to be interpreted and also good enough to help with this interpreting. We need to ensure that this pretraining doesn’t make the model capably misaligned.
    When we train powerful AI models, we want them to be both outer aligned and inner aligned; that is trained on the correct objective and for them to also properly learn that objective. Many proposals for achieving both outer and inner alignment look like an outer alignment proposal with some kind of oversight strapped on to deal with the inner alignment. Here ‘oversight’ means there is something with access to the internals of the model which checks that the model isn’t misaligned even if the behavior on the training distribution looks fine. In An overview of 11 proposals for building safe advanced AI, all but two of the proposals basically look like this, as does AI safety via market making.
    Examples of oversight techniques include:
    Transparency tools (either used by a human, an AI, or a human assisted by an AI)
    Adversarial inputs (giving inputs which could trick a misaligned AI into revealing itself)
    Relaxed adversarial training (which could be seen as an extension of adversarial inputs)
    Oversight loss
    We can use these oversight techniques to add an ‘oversight loss’ term to the loss function which will hopefully steer the training towards aligned models and away from misaligned ones. My model here is that we want to be doing oversight very close to the start of training to prevent it from ever becoming misaligned. If we start doing oversight too late, then the model may already be deceptively misaligned and then our oversight tools are much less likely to work (either because the model can obscure its cognition or the oversight tools are themselves misaligned). I think of this as steering the training process away from the ‘deceptive regions’ of parameter space (parameters which correspond to deceptive models), without ever having to enter the dangerous regions.
    Alternatively, rather than deception, we can think about regions where the model ‘behaves unacceptably’ on a some reasonable inputs. If taking a gradient descent step in a direction would increase the size of the set of reasonable inputs for which the model behaves unacceptably, then hopefully the oversight loss would provide some pressure away from stepping in that direction.
    Actually doing oversight
    I expect powerful AI models will be much too large and complicated for a human to understand/oversee them alone. So we will require help from other models to help with this task; this could be via models assisting the humans, or we could entirely hand the oversight process off to an oversight model.
    Here I’m taking very prosaic view: the model being overseen is a large language model, and the overseer model also as an LLM but trained to output some ‘acceptability score’ rather than generate text.
    In many proposals this oversight model is some amplified version of the model being overseen. T...
    6 min
  • EA - Announcing EA Pulse, large monthly US surveys on EA by David Moss
    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: Announcing EA Pulse, large monthly US surveys on EA, published by David Moss on September 20, 2022 on The Effective Altruism Forum.
    Rethink Priorities is excited to announce EA Pulse - a large, monthly survey of the US population aimed at measuring and understanding public perceptions of Effective Altruism and EA-aligned cause areas! This project has been made possible by a grant from the FTX Future Fund.
    What is EA Pulse?
    EA Pulse aims to serve two primary purposes:
    Tracking changes in responses to key questions relevant to EA and longtermism over time (e.g. awareness of and attitudes towards EA and longtermism, and support for different cause areas).
    Running ad hoc questions requested by EA orgs (e.g. support for particular policies, responses to different messages EAs are considering).
    We welcome requests for questions to include in the survey of either of these types. Please comment below or e-mail [email protected], ideally by October 20th.
    By tracking beliefs and attitudes towards issues related to effective altruism and longtermism, we can better get our finger on the pulse of movement building efforts over time, and potentially identify unforeseen risks to the movement. We will also be able to determine whether particular subgroups of the population appear to be missed or turned off by our outreach efforts.
    We also believe that surveying the broader public can provide a new window for looking at how the ideas generated by the EA community are being taken up by the wider population. In turn, it can help us communicate more effectively and efficiently about what matters most.
    Due to space constraints this survey is best suited to asking about relatively short, straightforward questions. If you are interested in surveys with more complex designs, a larger number of questions or experimental manipulations, complex instructions, or which involve asking respondents to read lengthy text or view videos, we are potentially able to accommodate these in separate surveys (funding permitting). Please feel free to reach out to discuss possibilities.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - EA for people with non-technical skillsets by Ronja
    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: EA for people with non-technical skillsets, published by Ronja on September 20, 2022 on The Effective Altruism Forum.
    Summary: I have noticed some people, especially when they are new to the community, see prominent messaging around AI or biorisk and conclude that they, if they don't have a technical background, don't fit in the community or cannot have an impactful career.
    To make such careers more mentally available, I have compiled a list of examples.
    Mostly aimed at people who themselves feel discouraged because they feel like they "don't fit", and who haven't been in the community long enough to see the diversity of roles themselves.
    Thoughts on going forward with increasing visibility of such roles are at the bottom.
    Some thoughts to begin with
    Your University major is not the biggest determinator of whether you will have an impactful career. Also see:
    skills that make you most employable
    And "what drives success on average" in the Five philosophies of career success post:
    General intelligence and different aspects of intelligence, such as verbal or mathematical or perceptual, judgement and rationality, and creativity
    Practice and credentials
    Work ethic: conscientiousness, emotional stability, and grit
    Integrity
    Social skills e.g. social perceptiveness, extraversion, and agreeableness
    Interest match
    Context match e.g. mentorship and cultural fit
    And remember that disadvantages in one area can be offset through other areas, and that for such huge endeavours that EA tackles, we need a community with a portfolio of very different skills and mindsets. And that, while some people may be on paths that are less clearly lined out for them in advance, that does not mean that there is not a space where they slot in and can thrive while doing good.
    Direct work
    Example 1: Operations management
    Operations management is one of the highest-impact careers overall
    Profile:/
    Podcast w Tanya Singh:/
    In general, there are very many roles in organisations that play to different strengths; you really wouldn't want to work somewhere that was solely run by researchers.
    Example 2: Art, design and illustration
    Just like other companies, EA(-aligned) organisations need people who are skilled in making things pretty and appealing to audiences.
    An example is the new/ who produce longtermist video content (they are currently looking for animators!)
    Example 3: Communications
    Job profile:/
    Journalism:/
    Of course, incl Kelsey Piper:
    Example 4: Policy and politics
    I think this is fairly well-known, so here just a few links
    Research paths
    Check out/
    Includes research ideas in media & communications, sociology, political science, law, business, and history.
    This is just a glimpse of what is possible, though! I bet you could, if you so desired, find exciting and impactful topics in any field.
    Example 1: History research
    Rose Hadshar's talk:
    A glimpse of her work:
    Research ideas in the history of social movements:
    Example 2: Behavioural science
    Daniel has a PhD in education and, when we last spoke, researched "security mindset" by interviewing people.
    Examples of people on their own paths
    Kikiope Oluwarore, who used her expertise as a veterinarian to co-found Healthier Hens.
    Liv Boeree, who used to be a famous poker player, then convinced other poker players to donate money, and who also used her influence in EA-adjacent Youtube videos. (Also note that there are people like Suzy Shepherd who shoot and cut and design those videos. Again - you don't have to be the face in the limelight)
    Thomas Moynihan, who (I think) studied history and then used that expertise to research and write a book about the history of the ideas of existential risks.
    Lizka Vaintrob, who manages the EA Forum and writes content for CEA.
    Varsha Venugopal, who studied urban and regional planning and international development, and...
    6 min
  • EA - Author Rutger Bregman about effective altruism and philanthropy by Sebastian Schwiecker
    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: Author Rutger Bregman about effective altruism and philanthropy, published by Sebastian Schwiecker on September 20, 2022 on The Effective Altruism Forum.
    Rutger Bregman, historian, and author (including Utopia for Realists and Humankind: A Hopeful History) describes his personal view on philanthropy in a conversation with Effektiv Spenden here (German here).
    In the Effektiv Spenden post donation survey he was mentioned more than any other person (e.g. more than Will MacAskill and more than Peter Singer). Our explanation is that he particularly good at reaching people from outside the existing EA community.
    Some quotes:
    On effective altruism:
    "Sometimes people can get the impression that “Oh, so you know what all the effective causes out there are and you are very dogmatic about that?” That’s not the case at all. Effective altruism is a question. It’s not an answer. It’s all about continuously asking yourself the question, is this the best use of my time, resources, and money? That’s what it’s really about. And I think intellectual humility is a really important value, and I think that’s also quite present in the movement."
    On systemic change vs. individual change:
    "There’s now this discussion going on amongst progressives and people on the left like: “Oh, we shouldn’t talk about individual change because that’s neo-liberal. We should all talk about system change”, but obviously we need to do both. If you look at the most impressive reformers and prophets and campaigners and activists throughout history, they all did it both. I’m now reading a book about Anthony Benezet who was one of the most important abolitionists, he’s called the father of abolitionism. He led the fight against the slave trade and slavery in the 18th century. If you would have said to him: “Oh, it’s all about the system. It’s not about the individual”. He would have said: “You’re a hypocrite.” Of course, it’s also about the individual, because he knew that he would be much more convincing if he actually did what he preached."
    On why he signed the Giving What We Can Pledge:
    "Because human behavior is contagious. We’re not individuals, we’re not lone atoms, but we influence each other all the time by our behavior. It’s just contagious. Giving can be like that as well. That’s why I think it’s important to be public about your giving, not to show off, you need to be a little bit careful there, but that’s also why I signed the Giving What We Can pledge to say. Look, people, if you like my work, this is what I find really important and it has made a big difference in my life to donate at least 10% of my income to highly effective causes. I think that actually, as a best seller author, you can go a little bit higher than 10%, but 10% is a good place to start."
    Since the interview is quite long feel free to share the video below with everyone who might be interested but can't be bothered for more than one minute.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min