Hide table of contents

Please note that the following grants are only recommendations, as all grants are still pending an internal due diligence process by CEA.

This post contains our allocation and some explanatory reasoning for our Q1 2019 grant round. We opened up an application for grant requests earlier this year which was open for about one month, after which we received an unanticipated large donation of about $715k. This caused us to reopen the application for another two weeks. We then used a mixture of independent voting and consensus discussion to arrive at our current grant allocation.

What is listed below is only a set of grant recommendations to CEA, who will run these by a set of due-diligence tests to ensure that they are compatible with their charitable objectives and that making these grants will be logistically feasible.

Grant Recipients

Each grant recipient is followed by the size of the grant and their one-sentence description of their project.

  • Anthony Aguirre ($70,000): A major expansion of the Metaculus prediction platform and its community
  • Tessa Alexanian ($26,250): A biorisk summit for the Bay Area biotech industry, DIY biologists, and biosecurity researchers
  • Shahar Avin ($40,000): Scaling up scenario role-play for AI strategy research and training; improving the pipeline for new researchers
  • Lucius Caviola ($50,000): Conducting postdoctoral research at Harvard on the psychology of EA/long-termism
  • Connor Flexman ($20,000): Performing independent research in collaboration with John Salvatier
  • Ozzie Gooen ($70,000): Building infrastructure for the future of effective forecasting efforts
  • Johannes Heidecke ($25,000): Supporting aspiring researchers of AI alignment to boost themselves into productivity
  • David Girardo ($30,000): A research agenda rigorously connecting the internal and external views of value synthesis
  • Nikhil Kunapuli ($30,000): A study of safe exploration and robustness to distributional shift in biological complex systems
  • Jacob Lagerros ($27,000): Building infrastructure to give X-risk researchers superforecasting ability with minimal overhead
  • Lauren Lee ($20,000): Working to prevent burnout and boost productivity within the EA and X-risk communities
  • Alex Lintz ($17,900): A two-day, career-focused workshop to inform and connect European EAs interested in AI governance
  • Orpheus Lummis ($10,000): Upskilling in contemporary AI techniques, deep RL, and AI safety, before pursuing a ML PhD
  • Vyacheslav Matyuhin ($50,000): An offline community hub for rationalists and EAs
  • Tegan McCaslin ($30,000): Conducting independent research into AI forecasting and strategy questions
  • Robert Miles ($39,000): Producing video content on AI alignment
  • Anand Srinivasan ($30,000): Formalizing perceptual complexity with application to safe intelligence amplification
  • Alex Turner ($30,000): Building towards a “Limited Agent Foundations” thesis on mild optimization and corrigibility
  • Eli Tyre ($30,000): Broad project support for rationality and community building interventions
  • Mikhail Yagudin ($28,000): Giving copies of Harry Potter and the Methods of Rationality to the winners of EGMO 2019 and IMO 2020
  • CFAR ($150,000): Unrestricted donation
  • MIRI ($50,000): Unrestricted donation
  • Ought ($50,000): Unrestricted donation

Total distributed: $923,150

Grant Rationale

Here we explain the purpose for each grant and summarize our reasoning behind their recommendation. Each summary is written by the fund member who was most excited about recommending the relevant grant (plus some constraints on who had time available to write up their reasoning). These differ a lot in length, based on how much available time the different fund members had to explain their reasoning.

Writeups by Helen Toner

Alex Lintz ($17,900)

A two-day, career-focused workshop to inform and connect European EAs interested in AI governance

Alex Lintz and some collaborators from EA Zürich proposed organizing a two-day workshop for EAs interested in AI governance careers, with the goals of giving participants background on the space, offering career advice, and building community. We agree with their assessment that this space is immature and hard to enter, and believe their suggested plan for the workshop looks like a promising way to help participants orient to careers in AI governance.

Writeups by Matt Wage

Tessa Alexanian ($26,250)

A biorisk summit for the Bay Area biotech industry, DIY biologists, and biosecurity researchers

We are funding Tessa Alexanian to run a one day biosecurity summit, immediately following the SynBioBeta industry conference. We have also put Tessa in touch with some experienced people in the biosecurity space who we think can help make sure the event goes well.

Shahar Avin ($40,000)

Scaling up scenario role-play for AI strategy research and training; improving the pipeline for new researchers

We are funding Shahar Avin to help him hire an academic research assistant and for other miscellaneous research expenses. We think positively of Shahar’s past work (for example this report), and multiple people we trust recommended that we fund him.

Lucius Caviola ($50,000)

Conducting postdoctoral research at Harvard on the psychology of EA/long-termism

We are funding Lucius Caviola for a 2-year postdoc at Harvard working with Professor Joshua Greene. Lucius plans to study the psychology of effective altruism and long-termism, and an EA academic we trust had a positive impression of him. We are splitting the cost of this project with the EA Meta Fund because some of Caviola’s research (on effective altruism) is a better fit for the Meta Fund while some of his research (on long-termism) is a better fit for our fund.

Ought ($50,000)

We funded Ought in our last round of grants, and our reasoning for funding them in this round is largely the same. Additionally, we wanted to help Ought diversify its funding base because it currently receives almost all its funding from only two sources and is trying to change that.

Our comments from last round:

Ought is a nonprofit aiming to implement AI alignment concepts in real-world applications. We believe that Ought’s approach is interesting and worth trying, and that they have a strong team. Our understanding is that hiring is currently more of a bottleneck for them than funding, so we are only making a small grant. Part of the aim of the grant is to show Ought as an example of the type of organization we are likely to fund in the future.

Writeups by Alex Zhu

Nikhil Kunapuli ($30,000)

A study of safe exploration and robustness to distributional shift in biological complex systems

Nikhil Kunapuli is doing independent deconfusion research for AI safety. His approach is to develop better foundational understandings of various concepts in AI safety, like safe exploration and robustness to distributional shift, by exploring these concepts in complex systems science and theoretical biology, domains outside of machine learning for which these concepts are also applicable. To quote an illustrative passage from his application:

When an organism within an ecosystem develops a unique mutation, one of several things can happen. At the level of the organism, the mutation can either be neutral in terms of fitness, maladaptive and leading to reduced reproductive success and/or death, or adaptive. For an adaptive mutation, the upgraded fitness of the organism will change the fitness landscape for all other organisms within the ecosystem, and in response, the structure of the ecosystem will either be perturbed into a new attractor state or destabilized entirely, leading to ecosystem collapse. Remarkably, most mutations do not kill their hosts, and most mutations also do not lead to ecosystem collapse. This is actually surprising when one considers the staggering complexity present within a single genome (tens of thousands of genes deeply intertwined through genomic regulatory networks) as well as an ecosystem (billions of organisms occupying unique niches and constantly co-evolving). One would naïvely think that a system so complex must be highly sensitive to change, and yet these systems are actually surprisingly robust. Nature somehow figured out a way to create robust organisms that could respond to and function in a shifting environment, as well as how to build ecosystems in which organisms could be free to safely explore their adjacent possible new forms without killing all other species.

Nikhil spent a summer doing research for the New England Complex Systems Institute. He also spent 6 months as the cofounder and COO of an AI hardware startup, which he left because he decided that direct work on AI safety is more urgent and important.

I recommended that we fund Nikhil because I think Nikhil’s research directions are promising, and because I personally learn a lot about AI safety every time I talk with him. The quality of his work will be assessed by researchers at MIRI.

Anand Srinivasan ($30,000)

Formalizing perceptual complexity with application to safe intelligence amplification

Anand Srinivasan is doing independent deconfusion research for AI safety. His angle of attack is to develop a framework that will allow researchers to make provable claims about what specific AI systems can and cannot do, based off of factors like their architectures and their training processes. For example, AlphaGo can “only have thoughts” about patterns on Go boards and lookaheads, which aren’t expressive enough to encode thoughts about malicious takeover.

AI researchers can build safe and extremely powerful AI systems by relying on intuitive judgments of their capabilities. However, these intuitions are non-rigorous and prone to error, especially since powerful optimization processes can generate solutions that are totally novel and unexpected to humans. Furthermore, competitive dynamics will incentivize rationalization about which AI systems are safe to deploy. Under fast takeoff assumptions, a single rogue AI system could lead to human extinction, making it particularly unreliable for us to rely exclusively on intuitive judgments about which AI systems are safe. Anand’s goal is to develop a framework that formalizes these intuitions well enough to permit future AI researchers to make provable claims about what future AI systems can and can’t internally represent.

Anand was the CTO of an enterprise software company that he cofounded with me, where he managed a six-person engineering team for two years. Upon leaving the company, he decided to refocus his efforts toward building safe AGI. Before dropping out of MIT, Anand worked on Ising models for fast image classification and fuzzy manifold learning (which was later independently published as a top paper at NIPS).

I recommended that we fund Anand because I think Anand’s research directions are promising, and I personally learn a lot about AI safety every time I talk with him. The quality of Anand’s work will be assessed by researchers at MIRI.

David Girardo ($30,000)

A research agenda rigorously connecting the internal and external views of value synthesis

David Girardo is doing independent deconfusion research for AI safety. His angle of attack is to elucidate the ontological primitives for representing hierarchical abstractions, drawing from his experience with type theory, category theory, differential geometry, and theoretical neuroscience.

I decided to fund David because I think David’s research directions are very promising, and because I personally learn a lot about AI safety every time I talk with him. Tsvi Benson-Tilsen, a MIRI researcher, has also recommended that David get funding. The quality of David’s work will be assessed by researchers at MIRI.

Writeups by Oliver Habryka

I have a broad sense that funders in EA tend to give little feedback to organizations they are funding, as well as organizations that they explicitly decided not to fund (usually due to time constraints). So in my writeups below I tried to be as transparent as possible in explaining the real reasons for what caused me to believe a grant was a good idea, what my biggest hesitations are, and took a lot of opportunities to explain background models of mine that might help others get better at understanding my future decisions in this space.

For some of the grants below, I think there exist more publicly defensible (or easier to understand) arguments for the grants that I recommended. However I tried to explain the actual models that drove my decisions for these grants, which are often hard to put into a few paragraphs of text, and so I apologize in advance for some of the explanations below almost certainly being a bit hard to understand.

Note that when I’ve written about how I hope a grant will be spent, this was in aid of clarifying my reasoning and is in no way meant as a restriction on what the grant should be spent on. The only restriction is that it should be spent on the project they applied for in some fashion, plus any further legal restrictions that CEA requires.

Mikhail Yagudin ($28,000)

Giving copies of Harry Potter and the Methods of Rationality to the winners of EGMO 2019 and IMO 2020

From the application:

EA Russia has the oral agreements with IMO [International Math Olympiad] 2020 (Saint Petersburg, Russia) & EGMO [European Girls’ Mathematical Olympiad] 2019 (Kyiv, Ukraine) organizers to give HPMORs [copies of Harry Potter and the Methods of Rationality] to the medalists of the competitions. We would also be able to add an EA / rationality leaflet made by CFAR (I contacted Timothy Telleen-Lawton on that matter).

My thoughts and reasoning

[Edit & clarification: The books will be given out by the organisers of the IMO and EGMO as prizes for the 650 people who got far enough to participate, all of which are "medalists".]

My model for the impact of this grant roughly breaks down into three questions:

  1. What effects does reading HPMOR have on people?
  2. How good of a target group are Math Olympiad winners for these effects?
  3. Is the team competent enough to execute on their plan?

What effects does reading HPMOR have on people?

My models of the effects of HPMOR stem from my empirical observations and my inside view on rationality training.

  • Empirically, a substantial number of top people in our community have (a) entered due to reading and feeling a deep connection to HPMOR and (b) attributed their approach to working on the long term future in substantial part to the insights they learned from reading HPMOR. This includes some individuals receiving grants on this list, and some individuals on the grant-making team.
  • I also weight here my inside view of the skills that HPMOR helps to teach. I’ll try to point at the things I think HPMOR does exceptionally and uniquely well at, though I find it a bit hard to make my models fully explicit here in an appropriate amount of space.
    • The most powerful tools that humanity has discovered so far are methods for thinking quantitatively and scientifically about how our universe works, and using this understanding to manipulate the universe. HPMOR attempts to teach the fundamental skills behind this thinking in three main ways:
      • The first way HPMOR teaches science is that the reader is given many examples of the inside of someone’s mind when they are thinking with the goal of actually understanding the world and are reasoning with the scientific and quantitative understanding humanity has developed. HPMOR is a fictional work, containing a highly detailed world with characters whose experience a reader empathises with and storylines that evoke responses from a reader. The characters in HPMOR demonstrate the core skills of quantitative, scientific reasoning: forming a hypothesis, making a prediction, throwing out the hypothesis when the prediction does not match reality, and otherwise updating probabilistically when they don’t yet have decisive evidence.
      • The second way HPMOR teaches science is that key scientific results and mechanisms are woven into the narrative of the book. Studies in the heuristics and biases literature, genetic selection, programming loops, Bayesian reasoning, and more are all explained in an unusually natural manner. They aren’t just added on top of the narrative in order for there to be science in the book; instead, the story’s universe is in fact constrained by these theories in such a way that they are naturally brought up by characters attempting to figure out what they should do.
      • This contributes to the third way HPMOR helps teach scientific thinking: HPMOR is specifically designed to be understandable in advance of the end of the book, and many readers have used the thinking tools taught in the book to do just that. One of the key bottlenecks in individuals’ ability to affect the long-term future is the ability to deal with the universe as though it is understandable in principle, and HPMOR creates a universe where this is so and includes characters doing their best to understand it. This sort of understanding is necessary for being able to take actions that will have large, intended effects on important and difficult problems 10^n years down the line.
    • The book also contains characters who viscerally care about humanity, other conscious beings, and our collective long-term future, and take significant actions in their own lives to ensure that this future goes well.
  • It is finally worth noting that HPMOR does all of the above things while also being a highly engaging book that has been read by hundreds of thousands of readers (if not more) primarily for pleasure. It is the most reviewed Harry Potter fan fiction on fanfiction.net, which is a remarkable state of affairs.

How good of a target group are Math Olympiad winners for these effects?

I think that Math Olympiad winners are a very promising demographic within which to find individuals who can contribute to improving the long-term future. I believe Math Olympiads select strongly on IQ as well as (weakly) on conscientiousness and creativity, which are all strong positives. Participants are young and highly flexible; they have not yet made too many major life commitments (such as which university they will attend), and are in a position to use new information to systematically change their lives’ trajectories. I view handing them copies of an engaging book that helps teach scientific, practical and quantitative thinking as a highly asymmetric tool for helping them make good decisions about their lives and the long-term future of humanity.

I’ve also visited and participated in a variety of SPARC events, and found the culture there (which is likely to be at least somewhat representative of Math Olympiad culture) very healthy in a broad sense. Participants displayed high levels of altruism, a lot of willingness to help one another, and an impressive amount of ambition to improve their own thinking and affect the world in a positive way. These observations make me optimistic about efforts that build on that culture.

I think it’s important when interacting with minors, and attempting to improve (and thus change) their life trajectories, to make sure to engage with them in a safe way that is respectful of their autonomy and does not put social pressures on them in ways they may not yet have learned to cope with. In this situation, Mikhail is working with/through the institutions that run the IMO and EGMO, and I expect those institutions to (a) have lots of experience with safeguarding minors and (b) have norms in place to make sure that interactions with the students are positive.

Is the team competent enough to execute on their plan?

I don’t have a lot of information on the team, don’t know Mikhail, and have not received any major strong endorsement for him and his team, which makes this the weakest link in the argument. However, I know that they are coordinating both with SPARC (which also works to give books like HPMOR to similar populations) and the team behind the highly successful Russian printing of HPMOR, two teams who have executed this kind of project successfully in the past. So I felt comfortable recommending this grant, especially given its relatively limited downside.

Alex Turner ($30,000)

Building towards a “Limited Agent Foundations” thesis on mild optimization and corrigibility

From the application:

I am a third-year computer science PhD student funded by a graduate teaching assistantship; to dedicate more attention to alignment research, I am applying for one or more trimesters of funding (spring term starts April 1).


Last summer, I designed an approach to the “impact measurement” subproblem of AI safety: “what equation cleanly captures what it means for an agent to change its environment, and how do we implement it so that an impact-limited paperclip maximizer would only make a few thousand paperclips?”. I believe that my approach, Attainable Utility Preservation (AUP), goes a long way towards answering both questions robustly, concluding:

> By changing our perspective from “what effects on the world are ‘impactful’?” to “how can we stop agents from overfitting their environments?”, a natural, satisfying definition of impact falls out. From this, we construct an impact measure with a host of desirable properties […] AUP agents seem to exhibit qualitatively different behavior […]

Primarily, I aim both to output publishable material for my thesis and to think deeply about the corrigibility and mild optimization portions of MIRI’s machine learning research agenda. Although I’m excited by what AUP makes possible, I want to lay the groundwork of deep understanding for multiple alignment subproblems. I believe that this kind of clear understanding will make positive AI outcomes more likely.

My thoughts and reasoning

I’m excited about this because:

  • Alex’s approach to finding personal traction in the domain of AI Alignment is one that I would want many other people to follow. On LessWrong, he read and reviewed a large number of math textbooks that are useful for thinking about the alignment problem, and sought public input and feedback on what things to study and read early on in the process.
  • He wasn’t intimidated by the complexity of the problem, but started thinking independently about potential solutions to important sub-problems long before he had “comprehensively” studied the mathematical background that is commonly cited as being the foundation of AI Alignment.
  • He wrote up his thoughts and hypotheses in a clear way, sought feedback on them early, and ended up making a set of novel contributions to an interesting sub-field of AI Alignment quite quickly (in the form of his work on impact measures, on which he recently collaborated with the DeepMind AI Safety team)

Potential concerns

These intuitions, however, are a bit in conflict with some of the concrete research that Alex has actually produced. My inside views on AI Alignment make me think that work on impact measures is very unlikely to result in much concrete progress on what I perceive to be core AI Alignment problems, and I have talked to a variety of other researchers in the field who share that assessment. I think it’s important that this grant not be viewed as an endorsement of the concrete research direction that Alex is pursuing, but only as an endorsement of the higher-level process that he has been using while doing that research.

As such, I think it was a necessary component of this grant that I have talked to other people in AI Alignment whose judgment I trust, who do seem excited about Alex’s work on impact measures. I think I would not have recommended this grant, or at least this large of a grant amount, without their endorsement. I think in that case I would have been worried about a risk of diverting attention from what I think are more promising approaches to AI Alignment, and a potential dilution of the field by introducing a set of (to me) somewhat dubious philosophical assumptions.

Overall, while I try my best to form concrete and detailed models of the AI Alignment research space, I don’t currently devote enough time to it to build detailed models that I trust enough to put very large weight on my own perspective in this particular case. Instead, I am mostly deferring to other researchers in this space that I do trust, a number of whom have given positive reviews of Alex’s work.

In aggregate, I have a sense that the way Alex went about working on AI Alignment is a great example for others to follow, I’d like to see him continue, and I am excited about the LTF Fund giving out more grants to others who try to follow a similar path.

Orpheus Lummis ($10,000)

Upskilling in contemporary AI techniques, deep RL and AI safety, before pursuing a ML PhD

From the application :

Notable planned subprojects:

  • Engaging with David Krueger’s AI safety reading group at Mila
  • Starting & maintaining a public index of AI safety papers, to help future literature reviews and to complement https://vkrakovna.wordpress.com/ai-safety-resources/, as a standalone wiki-page (eg at http://aisafetyindex.net)
  • From-scratch implementation of seminal deep RL algorithms
  • Going through textbooks: Goodfellow Bengio Courville 2016, Sutton Barto 2018, …
  • Possibly doing the next AI Safety camp
  • Building a prioritization tool for English Wikipedia using NLP, building on the literature of quality assessment (https://paperpile.com/shared/BZ2jzQ)
  • Studying the AI Alignment literature

My thoughts and reasoning

We funded Orpheus in our last grant round to run an AI Safety Unconference just after NeurIPS. We’ve gotten positive testimonials from the event, and I am overall happy about that grant.

I do think that of the grants I recommended this round, this is probably the one I feel least confident about. I don’t know Orpheus very well, and while I have received generally positive reviews of their work, I haven’t yet had the time to look into any of those reviews in detail, and haven’t seen clear evidence about the quality of their judgment. However, what I have seen seems pretty good, and if I had even a tiny bit more time to spend on evaluating this round’s grants, I would probably have spent it reaching out to Orpheus and talking with them more in person.

In general, I think time for self-study and reflection can be exceptionally important for people starting to work in AI Alignment. This is particularly true if they are following a more conventional academic path which could easily cause them to try to immediately work on contemporary AI capabilities research, because I generally think this has negative value even for people concerned about safety (though I do have some uncertainty here). I think giving people working on more classical ML research the time and resources to explore the broader implications of their work on safety, if they are already interested in that, is a good use of resources.

I am also excited about building out the Montreal AI Alignment community, and having someone who both has the time and skills to organize events and can understand the technical safety work seems likely to have good effects.

This grant is also the smallest grant we are funding this round, making me more comfortable with a bit less due diligence than the other grants, especially since this grant seems unlikely to have any large negative consequences.

Tegan McCaslin ($30,000)

Conducting independent research into AI forecasting and strategy questions

From the application:

1) I’d like to independently pursue research projects relevant to AI forecasting and strategy, including (but not necessarily limited to) some of the following:

I am actively pursuing opportunities to work with or under more senior AI strategy researchers [..], so my research focus within AI strategy is likely to be influenced by who exactly I end up working with. Otherwise I expect to spend some short period of time at the start generating more research ideas and conducting pilot tests on the order of several hours into their tractability, then choosing which to pursue based on an importance/tractability/neglectedness framework.


2) There are relatively few researchers dedicated full-time to investigating AI strategy questions that are not immediately policy-relevant. However, there nonetheless exists room to contribute to the research on existential risks from AI with approaches that fit into neither technical AI safety nor AI policy/governance buckets.

My thoughts and reasoning

Tegan has been a member of the X-risk network for several years now, and recently left AI Impacts. She is now looking for work as a researcher. Two considerations made me want to recommend that the LTF Fund make a grant to her.

  1. It’s easier to relocate someone who has already demonstrated trust and skills than to find someone completely new.
    1. This is (roughly) advice given by YCombinator to startups, and I think it’s relevant to the X-risk community. It’s cheaper for Tegan to move around and find the place for her to do her best work relative to an outsider who has not already worked within the X-risk network. A similarly skilled individual who is not already part of the network will need to spend a few years understanding the community and demonstrating that they can be trusted. So I think it is a good idea to help Tegan explore other parts of the community to work in.
  2. It’s important to give good researchers runway while they find the right place.
    1. For many years, the X-risk community has been funding-bottlenecked, keeping salaries low. A lot of progress has been made on this front and I hope that we’re able to fix this. Unfortunately, the current situation means that when a hire does not work out, the individual often doesn’t have much runway while reorienting, updating on what didn’t work out, and subsequently trialing at other organizations.
    2. This moves them much more quickly into an emergency mode, where everything must be optimized for short-term income, rather than long-term updating, skill building, and research. As such, I think it is important for Tegan to have a comfortable amount of runway while doing solo research and trialling at various organizations in the community.

While I haven’t spent the time to look into Tegan’s research in any depth, the small amount I did read looked promising. The methodology of this post is quite exciting, and her work there and on other pieces seems very thorough and detailed.

That said, my brief assessment of Tegan’s work was not the reason why I recommended this grant, and if Tegan asks for a new grant in 6 months to focus on solo research, I will want to spend significantly more time reading her output and talking with her, to understand how these questions were chosen and what precise relation they have to forecasting technological progress in AI.

Overall, I think Tegan is in a good place to find a valuable role in our collective X-risk reduction project, and I’d like her to have the runway to find that role.

Anthony Aguirre ($70,000)

A major expansion of the Metaculus prediction platform and its community

From the application:

The funds would be used to expand the Metaculus prediction platform along with its community. Metaculus.com is a fully-functional prediction platform with ~10,000 registered users and >120,000 predictions made to date on more than >1000 questions. The goals of Metaculus are:

  • Short-term: Provide a resource to science, tech, and (especially) EA-related communities already interested in generating, aggregating, and employing accurate predictions, and training to be better predictors.
  • Medium-term: Improve decision-making by individuals and groups by providing well-calibrated numerical predictions.
  • Long-term: encourage and backstop a widespread culture of accountable and accurate predictions and scenario planning.

There are two major high-priority expansions possible with funding in place. The first would be an integrated set of extensions to improve user interaction and information-sharing. This would include private messaging and notifications, private groups, a prediction “following” system to create micro-teams within individual questions, and various incentives and systems for information-sharing.

The second expansion would link questions into a network. Users would express links between questions, from very simple (“notify me regarding question Y when P(X) changes substantially) to more complex (“Y happens only if X happens, but not conversely”, etc.) Information can also be gleaned from what users actually do. The strength and character of these relations can then generate different graphical models that can be explored interactively, with the ultimate goal of a crowd-sourced quantitative graphical model that could structure event relations and propagate new information through the network.

My thoughts and reasoning

For this grant, and also the grants to Ozzie Gooen and Jacob Lagerros, I did not have enough time to write up my general thoughts on forecasting platforms and communities. I hope to later write a post with my thoughts here. But for a short summary, see my thoughts on Ozzie Gooen’s grant.

I am generally excited about people building platforms for coordinating intellectual labor, particularly on topics that are highly relevant to the long-term future. I think Metaculus has been providing a valuable service for the past few years, both in improving our collective ability to forecast a large variety of important world events and in allowing people to train and demonstrate their forecasting skills, which I expect to become more relevant in the future.

I am broadly impressed with how cooperative and responsive the Metaculus team has been in helping organizations in the X-risk space get answers to important questions, or provide software services to them (e.g. I know that they are helping Jacob Lagerros and Ben Goldhaber set up a private Metaculus instance focused on AI)

I don’t know Anthony well, and overall I am quite concerned that there is no full-time person on this project. My model is that projects like this tend to go a lot better if they have one core champion who has the resources to fully dedicate themselves to the project, and it currently doesn’t seem that Anthony is able to do that.

My current model is that Metaculus will struggle as a platform without a fully dedicated team or at least individual champion, though I have not done a thorough investigation of the Metaculus team and project, so I am not very confident of this. One of the major motivations for this grant is to ensure that Metaculus has enough resources to hire a potential new champion for the project (who ideally also has programming skills or UI design skills to allow them to directly work on the platform). That said, Metaculus should use the money as best they see fit.

I am also concerned about the overlap of Metaculus with the Good Judgment Project, and currently have a sense that it suffers from being in competition with it, while also having access to substantially fewer resources and people.

The requested grant amount was for $150k, but I am currently not confident enough in this grant to recommend filling the whole amount. If Metaculus finds an individual new champion for the project, I can imagine strongly recommending that it gets fully funded, if the new champion seems competent.

Lauren Lee ($20,000)

Working to prevent burnout and boost productivity within the EA and X-risk communities

From the application:

(1) After 2 years as a CFAR instructor/researcher, I’m currently in a 6-12 month phase of reorienting around my goals and plans. I’m requesting a grant to spend the coming year thinking about rationality and testing new projects.

(2) I want to help individuals and orgs in the x-risk community orient towards and achieve their goals.

(A) I want to train the skill of dependability, in myself and others.

This is the skill of a) following through on commitments and b) making prosocial / difficult choices in the face of fear and aversion. The skill of doing the correct thing, despite going against incentive gradients, seems to be the key to virtue.

One strategy I’ve used is to surround myself with people with shared values (CFAR, Bay Area) and trust the resulting incentive gradients. I now believe it is also critical to be the kind of person who can take correct action despite prevailing incentive structures.

Dependability is also related to thinking clearly. Your ability to make the right decision depends on your ability to hold and be with all possible realities, especially painful and aversive ones. Most people have blindspots that actively prevent this.

I have some leads on how to train this skill, and I’d like both time and money to test them.

(B) Thinking clearly about AI risk

Most people’s decisions in the Bay Area AI risk community seem model-free. They themselves don’t have models of why they’re doing what they’re doing; they’re relying on other people “with models” to tell them what to do and why. I’ve personally carried around such premises. I want to help people explore where their ‘placeholder premises’ are and create safety for looking at their true motivations, and then help them become more internally and externally aligned.

(C) Burnout

Speaking of “not getting very far.” My personal opinion is that most ex-CFAR employees left because of burnout; I’ve written what I’ve learned here, see top 2 comments: [https://forum.effectivealtruism.org/posts/NDszJWMsdLCB4MNoy/burnout-what-is-it-and-how-to-treat-it#87ue5WzwaFDbGpcA7]. I’m interested in working with orgs and individuals to prevent burnout proactively.

(3) Some possible measurable outputs / artifacts:

  • A program where I do 1-on-1 sessions with individuals or orgs; I’d create reports based on whether they self-report improvements
  • X-risk orgs (e.g. FHI, MIRI, OpenPhil, BERI, etc.) deciding to spend time/money on my services may be a positive indicator, as they tend to be thoughtful with how they spend their resources
  • Writings or talks
  • Workshops with feedback forms
  • A more effective version of myself (notable changes = gaining the ability to ride a bike / drive a car / exercise—a PTSD-related disability, ability to finish projects to completion, others noticing stark changes in me)

My thoughts and reasoning

Lauren worked as an instructor at CFAR for about 2 years, until Fall 2018. I review CFAR’s impact as an institution below; in general, I believe it has helped set a strong epistemic foundation for the community and been successful in recruitment and training. I have a great appreciation for everyone who helps them with their work.

Lauren is currently in a period of reflection and reorientation around her life and the problem of AGI, in part due to experiencing burnout in the months before she left CFAR. To my knowledge, CFAR has never been well-funded enough to offer high salaries to its employees, and I think it is valuable to ensure that people who work at EA orgs and burn out have the support to take the time for self-care after quitting due to long-term stress. Ideally, I think this should be improved by higher salaries that allow employees to build significant runway to deal with shocks like this, but I think that the current equilibrium of salary levels in EA does not make that easy. Overall, I think it’s likely that staff at highly valuable EA orgs will continue burning out, and I don’t currently see it as an achievable target to not have this happen (though I am in favor of people people working on solving the problem).

I do not know Lauren well enough to evaluate the quality of her work on the art of human rationality, but multiple people I trust have given positive reviews (e.g. see Alex Zhu above), so I am also interested to read her output on the subjects she is thinking about.

I think it’s very important that people who work on developing an understanding of human rationality take the time to add their knowledge into our collective understanding, so that others can benefit from and build on top of it. Lauren has begun to write up her thoughts on topics like burnout, intentions, dependability, circling, and curiosity, and her having the space to continue to write up her ideas seemed like a significant additional positive outcome of this grant.

I think that she should probably aim to make whatever she does valuable enough that individuals and organizations in the community wish to pay her directly for her work. It’s unlikely that I would recommend renewing this grant for another 6 month period in the absence of a relatively exciting new research project/direction, and if Lauren were to reapply, I would want to have a much stronger sense that the projects she was working on were producing lots of value before I decided to recommend funding her again.

In sum, this grant hopefully helps Lauren to recover from burning out, get the new rationality projects she is working on off the ground, potentially identify a good new niche for her to work in (alone or at an existing organization), and write up her ideas for the community.

Ozzie Gooen ($70,000)

Build infrastructure for the future of effective forecasting efforts

From the application:

What I will do

I applied a few months ago and was granted $20,000 (thanks!). My purpose for this money is similar but greater in scope to the previous round. The previous funding has given me the security to be more ambitious, but I’ve realized that additional guarantees of funding should help significantly more. In particular, engineers can be costly and it would be useful to secure additional funding in order to give possible hires security.

My main overall goal is to advance the use of predictive reasoning systems for purposes most useful for Effective Altruism. I think this is an area that could eventually make use of a good deal of talent, so I have come to see my work at this point as foundational.

This work is in a few different areas that I think could be valuable. I expect that after a while a few parts will emerge as the most important, but think it is good to experiment early when the most effective route is not yet clear.

I plan to use additional funds to scale my general research and development efforts. I expect that most of the money will be used on programming efforts.


Foretold is a forecasting application that handles full probability distributions. I have begun testing it with users and have been asked for quite a bit more functionality. I’ve also mapped out the features that I expect people will eventually desire, and think there is a significant amount of work that would be significantly useful.

One particular challenge is figuring out the best way to handle large numbers of questions (1000 active questions plus, at a time.) I believe this requires significant innovations in the user interface and backend architecture. I’ve made some wireframes and have experimented with different methods, and believe I have a pragmatic path forward, but will need to continue to iterate.

I’ve talked with members of multiple organizations at this point who would like to use Foretold once it has a specific set of features, and cannot currently use any existing system for their purposes. […]


Ken is a project to help organizations set up and work with structured data, in essence allowing them to have private versions of Wikidata. Part of the project is Ken.js, a library which I’m beginning to integrate with Foretold.

Expected Impact

The main aim of EA forecasting would be to better prioritize EA actions. I think that if we could have a powerful system set up, it could make us better at predicting the future, better at understanding what things are important and better at coming to a consensus on challenging topics.


In the short term, I’m using heuristics like metrics regarding user activity and upvotes on LessWrong. I’m also getting feedback by many people in the EA research community. In the medium to long term, I hope to set up evaluation/estimation procedures for many projects and would include this one in that process.

My thoughts and reasoning

This grant is to support Ozzie Gooen in his efforts to build infrastructure for effective forecasting. Ozzie requested $70,000 to hire a software engineer who would support him on his work on the prediction platform www.foretold.iothat he is working on.

  • When thinking about how to improve the long-term future, I think we are confused about what counts as progress and what specific problems need solving. We can already see that there are a lot of technical and conceptual problems that have to be solved to make progress on a lot of the big problems we think are important.
  • I think that in order to make effective intellectual progress, you need some way for many people to collaborate on solving problems and to document the progress they have made so far.
  • I think there is potentially a lot of low-hanging fruit in designing better online platforms for making intellectual progress (which is why I chose to work on LessWrong + AI Alignment Forum + EA Forum). Ozzie works in this space too, and previously built Guesstimate (a spreadsheet where every cell is a probability distribution), which I think displayed some real innovation in the way we can use technology to communicate and clarify ideas. It was also produced to a very high standard of quality.
  • Forecasting platforms in particular have already displayed significant promise and tractability, with recent work by Philip Tetlock showing that a simple prediction platform can outperform major governmental institutions like the CIA, and older work by Robin Hanson, showing ways that prediction markets could help us make progress on a number of interesting problems.
  • The biggest concerns I have with Ozzie’s work, as well as the work on other prediction and aggregation platforms, is that the problem of getting people to actually use the product turns out to be very hard. Matt Fallshaw’s team at Trike Apps built https://predictionbook.com/, but then found it hard to get people to actually use it. Ozzie’s last project, Guesstimate, seemed quite well-executed, but similarly faltered due to low user numbers and a lack of interest from potential customers in industry. As such, I think it’s important not to underestimate the difficulty of making the product good enough that people actually use it.
  • I do think that the road to building knowledge aggregation platforms will include many failed projects and many experiments that never get traction; as such, I do think that one should not over-update on the lack of users for some of the existing platforms. As a positive counterexample, the Good Judgment Project seems to have a consistently high number of people making predictions.
  • I’ve also frequently interacted with Ozzie in person, and generally found his reasoning and judgment in this domain to be good. I also think it is quite good that he has been writing up his thinking for the community to read and engage with, which will allow other people to build off of his thinking and efforts, even if he doesn’t find traction with this particular project.

Johannes Heidecke ($25,000)

Supporting aspiring researchers of AI alignment to boost themselves into productivity

From the application:

(1) We would like to apply for a grant to fund an upcoming camp in Madrid that we are organizing. The camp consists of several weeks of online collaboration on concrete research questions, culminating in a 9-day intensive in-person research camp. Participants will work in groups on tightly-defined research projects in strategy and technical AI safety. Expert advisors from AI Safety/Strategy organizations will help refine proposals to be tractable and relevant. This allows for time-efficient use of advisors’ knowledge and research experience, and ensures that research is well-aligned with current priorities. More information: https://aisafetycamp.com/

(2) The field of AI alignment is talent-constrained, and while there is a significant number of young aspiring researchers who consider focussing their career on research on this topic, it is often very difficult for them to take the first steps and become productive with concrete and relevant projects. This is partially due to established researchers being time-constrained and not having time to supervise a large number of students. The goals of AISC are to help a relatively large number of high-talent people to take their first concrete steps in research on AI safety, connect them to collaborate, and efficiently use the capacities of experienced researchers to guide them on their path.

(3) We send out evaluation questionnaires directly after the camp and in regular intervals after the camp has passed. We measure impact on career decisions and collaborations and keep track of concrete output produced by the teams, such as blog posts or published articles.

We have successfully organized two camps before and are in the preparation phase for the third camp taking place in April 2019 near Madrid. I was the main organizer for the second camp and am advising the core team of the current camp, as well as organizing funding.

An overview of previous research projects from the first 2 camps can be found here:



We have evaluated the feedback from participants of the first two camps in the following two documents:



My thoughts and reasoning

I’ve talked with various participants of past AI Safety camps and heard broadly good things across the board. I also generally have a positive impression of the people involved, though I don’t know any of the organizers very well.

The material and testimonials that I’ve seen so far suggest that the camp successfully points participants towards a technical approach to AI Alignment, focusing on rigorous reasoning and clear explanations, which seems good to me.

I am not really sure whether I’ve observed significant positive outcomes of camps in past years, though this might just be because I am less connected to the European community these days.

I also have a sense that there is a lack of opportunities for people in Europe to productively work on AI Alignment related problems, and so I am particularly interested in investing in infrastructure and events there. This does however make this a higher-risk grant, since I think this means this event and the people surrounding it might become the main location for AI Alignment in Europe, and if the quality of the event and the people surrounding it isn’t high enough, this might cause long-term problems for the AI Alignment community in Europe.


  • I think organizing long in-person events is hard, and conflict can easily have outsized negative effects. The reviews that I read from past years suggest that interpersonal conflict negatively affected many participants. Learning how to deal with conflict like this is difficult. The organizers seem to have considered this and thought a lot about it, but the most likely way I expect this grant to have large negative consequences is still if there is some kind of conflict at the camp that results in more serious problems.
  • I think it’s inevitable that some people won’t get along with organizers or other participants at the camp for cultural reasons. If that happens, I think it’s important for these people to have some other way of getting connected to people working on AI Alignment. I don’t know the best way to arrange this, but I would want the organizers to think about ways to achieve it.

I also coordinated with Nicole Ross from CEA’s EA Grants project, who had considered also making a grant to the camp. We decided it would be better for the LTF Fund team to make this grant, though we wanted to make sure that some of the concerns Nicole had with this grant were summarized in our announcement:

  • AISC could potentially turn away people who would be very good for AI Safety or EA, if those people have negative interactions at the camp or if they are much more talented than other participants (and therefore develop a low opinion of AI Safety and/or EA).
  • Some negative interactions with people at the camp could, as with all residential programs, lead to harm and/or PR issues, (for example, if someone at the camp were sexually harassed). Being able to handle such issues thoughtfully and carefully is a hard task, and additional support or advice may be beneficial.

This seems to roughly mirror my concerns above.

I would want to engage with the organizers a fair bit more before recommending a renewal of this grant, but I am happy about the project as a space for Europeans to get engaged with alignment ideas and work on them for a week together with other technical and engaged people.

Broadly, the effects of the camp seem very likely to be positive, while the (financial) cost of the camp seems small compared to the expected size of the impact. This makes me relatively confident that this grant is a good bet.

Vyacheslav Matyuhin ($50,000)

An offline community hub for rationalists and EAs

From the application:

Our team is working on the offline community hub for rationalists and EAs in Moscow called Kocherga (details on Kocherga are here).

We want to make sure it keeps existing and grows into the working model for building new flourishing local EA communities around the globe.

Our key assumptions are:

  1. There’s a gap between the “monthly meetup” EA communities and the larger (and significantly more productive/important) communities. That gap is hard to close for many reasons.
  2. Solving this issue systematically would add a lot of value to the global EA movement and, as a consequence, the long-term future of humanity.
  3. Closing the gap requires a lot of infrastructure, both organizational and technological.

So we work on building such an infrastructure. We also keep in mind the alignment and goodharting issues (building a big community of people who call themselves EAs but who don’t actually share EA virtues would be bad, obviously).


Concretely, we want to:

  1. Add 2 more people to our team.
  2. Implement our new community building strategy (which includes both organizational tasks such as new events and processes for seeding new working groups, and technological tasks such as implementing a website which allows people from the community to announce new private meetups or team up for coaching or mastermind groups)
  3. Improve our rationality workshops (in terms of scale and content quality). Workshops are important for attracting new community members, for keeping the high epistemic standards of the community and for making sure that community members can be as productive as possible.

To be able to do this, we need to cover our current expenses somehow until we become profitable on our own.

My thoughts and reasoning

The Russian rationality community is surprisingly big, which suggests both a certain level of competence from some of its core organizers and potential opportunities for more community building. The community has:

  • Successfully translated The Sequences and HPMOR into Russian, as can be seen at the helpful LessWrong.ru site.
  • Executed a successful kickstarter campaign to distribute physical copies of HPMOR (over 7,000 copies).
  • Built a community hub in Moscow called Kocherga, which is a financially self-sustaining anti-cafe (a cafe where you pay for time spent there rather than drinks/snacks) that hosts a variety of rationality events for roughly 100 attendees per week.

This grant is to the team that runs the Kocherga anti-cafe.

Their LessWrong write-up suggests:

  • They have good skills at building spaces, running events, and generally preserving their culture while still being financially sustainable
  • They’ve seen steady increases over time in available funding and attendees
  • They’ve succeeded at being largely self-sufficient for 4 years
  • They’ve successfully engaged with other local intellectual communities
  • Their culture seems to value careful thinking and good discourse a lot, and they seem to have put serious effort into developing the art of rationality, including caring about the technical aspects and incorporating CFAR’s work into their thinking

I find myself having slightly conflicted feelings about the Russian rationality community trying to identify and integrate more with the EA community. I think a major predictor of how excited I have historically been about community building efforts has been a group’s emphasis on improving members’ judgement and thinking skills, as well as the degree to which it emphasizes high epistemic standards and careful thinking. I am quite excited about how Kocherga seems to have focused on those issues so far, and I am worried that this integration and change of identity will reduce that focus (as I think it has for some local and student groups that made a similar transition). That said, I think the Kocherga group has shown quite good judgement on this dimension (see here), which addresses many of my concerns, though I am still interested in thinking and talking about these issues further.

I’m somewhat concerned that I’m not aware of any major insights or unusually talented people from this community, but I expect the language barrier to be a big part of what is preventing me from hearing about those things. And I am somewhat confused about how to account for interesting ideas that don’t spread to the projects I care most about.

I think there are benefits to having an active Russian community that can take opportunities that are only available for people in Russia, or at least people who speak Russian. This particularly applies to policy-oriented work on AI alignment and other global catastrophic risks, which is also a domain that I feel confused about and have a hard time evaluating.

For a lot of the work that I do feel comfortable evaluating, I expect the vast majority of intellectual progress to be made in the English-speaking world, and as such, the question of how talent can flow from Russia to the existing communities working on the long-term future seems quite important. I hope this grant can facilitate a stronger connection between the rest of the world and the Russian community, to improve that talent and idea flow.

This grant seemed like a slightly better fit for the EA Meta fund. They decided not to fund it, so we made it instead, since it still seemed like a strong proposal to us.

What I have seen so far makes me confident that this grant is a good idea. However, before we make more grants like this, I would want to talk more to the organizers involved and generally get more information on the structure and culture of the Russian EA and rationality communities.

Jacob Lagerros ($27,000)

Building infrastructure to give x-risk researchers superforecasting ability with minimal overhead

From the application:

Build a private platform where AI safety and policy researchers have direct access to a base of superforecaster-equivalents, and where aspiring EAs with smaller opportunity costs but excellent calibration perform useful work.


I previously received two grants to work on this project: a half-time salary from EA Grants, and a grant for direct project expenses from BERI. Since then, I dropped out of a Master’s programme to work full-time on this, seeing that was the only way I could really succeed at building something great. However, during that transition there were some logistical issues with other grantmakers (explained in more detail in the application), hence I applied to the LTF for funding for food, board, travel and the runway to make more risk-neutral decisions and capture unexpected opportunities in the coming ~12 months of working on this.”

My thoughts and reasoning

There were three main factors behind my recommending this grant:

  1. My object-level reasons for recommending this grant are quite similar to my reasons for recommending Ozzie Gooen’s and Anthony Aguirre’s.
  2. Jacob has been around the community for about 3 years. The output of his that I’ve seen has included (amongst other things) competently co-directing EAGxOxford 2016, and some thoughtful essays on LessWrong (e.g. 1, 2, 3, 4).
  3. Jacob’s work seems useful to me, and is being funded on the recommendation of the FHI Research Scholars Programme and the Berkeley Existential Risk Initiative. He is also collaborating with others I’m excited about (Metaculus and Ozzie Gooen).

However, I did not assess the grant in detail, as the only reason Jacob asked for a grant was due to logistical complications with other grantmakers. Since FHI and BERI have already investigated the project in more detail, I was happy to suggest we pick up the slack to ensure Jacob has the runway to pursue his work.

Connor Flexman ($20,000)

Perform independent research in collaboration with John Salvatier

I am recommending this grant with more hesitation than most of the other grants in this round. The reasons for hesitation are as follows:

  • I was the primary person on the grant committee on whose recommendation this grant was made.
  • Connor lives in the same group house that I live in, which I think adds a complicating conflict of interest to my recommendation.
  • I have generally positive impressions of Connor, but I have not personally seen concrete, externally verifiable evidence that clearly demonstrates his good judgment and competence, which in combination with the other two factors makes me more hesitant than usual.

However, despite these reservations, I think this grant is a good choice. The two primary reasons are:

  1. Connor himself has worked on a variety of research and community building projects, and both by my own assessment and other people I talked to, has significant potential in becoming a strong generalist researcher, which I think is an axis on which a lot of important projects are bottlenecked.
  2. This grant was strongly recommended to me by John Salvatier, who is funded by an EA Grant and whose work I am generally excited about.

John did some very valuable community organizing while he lived in Seattle and is now working on developing techniques to facilitate skill transfer between experts in different domains. I think it is exceptionally hard to develop effective techniques for skill transfer, and more broadly techniques to improve people’s rationality and reasoning skills, but am sufficiently impressed with John’s thinking that I think he might be able to do it anyway (though I still have some reservations).

John is currently collaborating with Connor and requested funding to hire him to collaborate on his projects. After talking to Connor I decided it would be better to recommend a grant to Connor directly, encouraging him to continue working with John but also allowing him to switch towards other research projects if he finds he can’t contribute as productively to John’s research as he expects.

Overall, while I feel some hesitation about this grant, I think it’s very unlikely to have any significant negative consequences, and I assign some significant probability that this grant can help Connor develop into an excellent generalist researcher of a type that I feel like EA is currently quite bottlenecked on.

Eli Tyre ($30,000)

Broad project support for rationality and community building interventions

Eli has worked on a large variety of interesting and valuable projects over the last few years, many of them too small to have much payment infrastructure, resulting in him doing a lot of work without appropriate compensation. I think his work has been a prime example of picking low-hanging fruit by using local information and solving problems that aren’t worth solving at scale, and I want him to have resources to continue working in this space.

Concrete examples of projects he has worked on that I am excited about:

  • Facilitating conversations between top people in AI alignment (I’ve in particular heard very good things about the 3-day conversation between Eric Drexler and Scott Garrabrant that Eli helped facilitate)
  • Organizing advanced workshops on Double Crux and other key rationality techniques
  • Doing a variety of small independent research projects, like this evaluation of birth order effects in mathematicians
  • Providing many new EAs and rationalists with advice and guidance on how to get traction on working on important problems
  • Helping John Salvatier develop techniques around skill transfer

I think Eli has exceptional judgment, and the goal of this grant is to allow him to take actions with greater leverage by hiring contractors, paying other community members for services, and paying for other varied expenses associated with his projects.

Robert Miles ($39,000)

Producing video content on AI alignment

From the application:

My goals are:

  1. To communicate to intelligent and technically-minded young people that AI Safety:
    1. is full of hard, open, technical problems which are fascinating to think about
    2. is a real existing field of research, not scifi speculation
    3. is a growing field, which is hiring
  2. To help others in the field communicate and advocate better, by providing high quality, approachable explanations of AIS concepts that people can share, instead of explaining the ideas themselves, or sharing technical documents that people won’t read
  3. To motivate myself to read and internalise the papers and textbooks, and become a technical AIS researcher in future

My thoughts and reasoning

I think video is a valuable medium for explaining a variety of different concepts (for the best examples of this, see 3Blue1Brown, CGP Grey, and Khan Academy). While there are a lot of people working directly on improving the long term future by writing explanatory content, Rob is the only person I know who has invested significantly in getting better at producing video content. I think this opens a unique set of opportunities for him.

The videos on his Youtube channel pick up an average of ~20k views. His videos on the official Computerphile channel often pick up more than 100k views, including for topics like logical uncertainty and corrigibility (incidentally, a term Rob came up with).

More things that make me optimistic about Rob’s broad approach:

  • He explains that AI alignment is a technical problem. AI safety is not primarily a moral or political position; the biggest chunk of the problem is a matter of computer science. Reaching out to a technical audience to explain that AI safety is a technical problem, and thus directly related to their profession, is a type of ‘outreach’ that I’m very happy to endorse.
  • He does not make AI safety a politicized matter. I am very happy that Rob is not needlessly tribalising his content, e.g. by talking about something like “good vs bad ML researchers”. He seems to simply portray it as a set of interesting and important technical problems in the development of AGI.
  • His goal is to create interest in these problems from future researchers, and not to simply get as large of an audience as possible. As such, Rob’s explanations don’t optimize for views at the expense of quality explanation. His videos are clearly designed to be engaging, but his explanations are simple and accurate. Rob often interacts with researchers in the community (at places like DeepMind and MIRI) to discuss which concepts are in need of better explanations. I don’t expect Rob to take unilateral action in this domain.

Rob is the first skilled person in the X-risk community working full-time on producing video content. Being the very best we have in this skill area, he is able to help the community in a number of novel ways (for example, he’s already helping existing organizations produce videos about their ideas).

Rob made a grant request during the last round, in which he explicitly requested funding for a collaboration with RAISE to produce videos for them. I currently don’t think that working with RAISE is the best use of Rob’s talent, and I’m skeptical of the product RAISE is currently trying to develop. I think it’s a better idea for Rob to focus his efforts on producing his own videos and supporting other organizations with his skills, though this grant doesn’t restrict him to working with any particular organization and I want him to feel free to continue working on RAISE if that is the project he thinks is currently most valuable.

Overall, Rob is developing a new and valuable skill within the X-risk community, and executing on it in a very competent and thoughtful way, making me pretty confident that this grant is a good idea.

MIRI ($50,000)

My thoughts and reasoning

  • MIRI is a 20-year-old research organization that seeks to resolve the core difficulties in the way of AGI having a positive impact.
    • My model of MIRI’s approach looks something like an attempt to join the ranks of Turing, Shannon, von Neumann and others, in creating a fundamental piece of theory that helps humanity to understand a wide range of powerful phenomena. Gaining an understanding of the basic theory of intelligent agents well enough to think clearly about them is plausibly necessary for building an AGI that ensures the long term future goes well.
    • It seems to me that they are making real progress (although I’m not confident of the rate of that progress) - for example, MIRI has discovered a Solomonoff-induction-style algorithm that can reason well under logical uncertainty, learning reasonable probabilities for mathematical propositions before they can be proved, which I found surprising. While I am uncertain about the usefulness of this particular insight on the path to further basic theory, I take it as some evidence that they’re using methods that can in principle make progress, which is something that I have historically been pessimistic about.
  • Only in recent years have there been routes to working on alignment that have also given you funding, status, and a stable social life. Nowadays many others are helping out the work of solving alignment, but MIRI core staff worked on the problem while all the incentives pulled in other directions. For me this is a strong sign of their integrity, and makes me expect they will make good decisions in many contexts where the best action isn't the locally incentivized action. It is also evidence that if I can’t understand why their weird action is good, that they will often still be correct to do it, and this is an outside view in favor of funding them in cases where I don't have my own inside-view model of why the project they're working on is good.
  • On that note, MIRI has also worked on a number of other projects that have attempted to teach the skills behind their general methodology for reasoning quantitatively and scientifically about the world and taking right action. I regret not having the time to detail all the impacts of these projects, but they include (and are not limited to): LessWrong, The Sequences, HPMOR, Inadequate Equilibria, Embedded Agency, and CFAR (an organization I discuss below). I view these as some of the main reasons the x-risk community exists.
  • Another outside view to consider is the support of MIRI by so many others whom I trust. Their funders have included Open Phil, BERI, FLI, and Jaan Tallinn, plus a variety of smaller donors I trust, and they are advised by Stuart Russell and Nick Bostrom. They’ve also been supported by other people who I don’t necessarily trust directly, but who I do think have interesting and valuable perspectives on the world, like Peter Thiel and Vitalik Buterin .
  • I also judge the staff to be exceptionally competent. Some examples:
    • The programming team has taken very early hires from multiple good startups such as Triplebyte, Recursion Pharmaceuticals, and Quixey, and also includes the Haskell core-developer Edward Kmett.
    • The ops staff are currently, in my evaluation, the most competent operations team of any of the organizations that I have personally interacted with.

In sum, I think MIRI is one of the most competent and skilled teams attempting to improve the long-term future, I have a lot of trust in their decision-making, and I’m strongly in favor of ensuring that they’re able to continue their work.

Thoughts on funding gaps

Despite all of this, I have not actually recommended a large grant to MIRI.

  • This is due to MIRI’s funding situation being solid at its current level (I would be thinking very differently if I annually had tens of millions of dollars to give away). But MIRI’s marginal use of dollars at this point of funding seems lower-impact, so I only recommended $50k.
  • I feel conflicted about whether it might be better to give MIRI more money. Historically, it has been common in the EA funding landscape to only give funding to organizations when they have demonstrated concrete room for more funding, or when funding is the main bottleneck for the organization. I think this has allowed us to start many small organizations that are working on a variety of different problems.
    • A common way in which at least some funding decisions are made is to compare the effect of a marginal donation now with the effect of a marginal donation at an earlier point in the project’s lifecycle (i.e. not wanting to invest in a project after it has hit strongly diminishing marginal returns, aka “maxed out its room for more funding” or “filled the funding gap”).
    • However, when I think about this from first principles, I think we should expect a heavy-tailed (probably log-normal) distribution in the impact of different cause areas, individuals, and projects. And while I can imagine that many good opportunities might hit strong diminishing marginal returns early on, it doesn’t seem likely for most projects. Instead, I expect factors that stay constant over the life of a project, like its broader organizational philosophy, core staff, and choice of problem to solve, to determine a large part its marginal value. Thus, we should expect our best guesses to be worth investing significant further resources into.

However, this is all complicated by a variety of countervailing considerations, such as the following three:

  1. Power law distributions of impact only really matter in this way if we can identify which interventions we expect to be in the right tail of impact, and I have a lot of trouble properly bounding my uncertainty here.
  2. If we are faced with significant uncertainty about cause areas, and we need organizations to have worked in an area for a long time before we can come to accurate estimates about its impact, then it’s a good idea to invest in a broad range of organizations in an attempt to get more information. This is related to common arguments around “explore/exploit tradeoffs”.
  3. Sometimes, making large amounts of funding available to one organization can have negative consequences for the broader ecosystem of a cause area. Also, giving an organization access to more funding than it can use productively may cause it to make too many hires or lose focus by trying to scale too quickly. Having more funding often also attracts adversarial actors and increases competitive stakes within an organization, making it a more likely target for attackers.

I can see arguments that we should expect additional funding for the best teams to be spent well, even accounting for diminishing margins, but on the other hand I can see many meta-level concerns that weigh against extra funding in such cases. Overall, I find myself confused about the marginal value of giving MIRI more money, and will think more about that between now and the next grant round.

CFAR ($150,000)

[Edit: It seems relevant to mention that LessWrong is currently receiving operational support from CFAR, in a way that makes me technically an employee of CFAR (similar to how ACE and 80K were/are part of CEA for a long time). However, LessWrong operates as a completely separate entity with its own fundraising and hiring procedures, and I don't feel any hesitation or pressure to critique CFAR openly because of that relation. Though I find myself a tiny bit more hesitant to speak harshly of specific individuals, simply because I am only working a floor away from the CFAR offices and that does have some psychological effect on me. Though the same was true for CEA while LessWrong was located in the CEA office for a few months, and was true for residents of my group house while LessWrong was located in the living room of my group house for most of the past two years, so I don't think this effect is particularly large.]

I think that CFAR’s intro workshops have historically had a lot of positive impact. I think they have done so via three pathways.

  1. Establishing epistemic norms: I think CFAR workshops are quite good at helping the EA and rationality community establish norms about what good discourse and good reasoning look like. As a concrete example of this, the concept of Double Crux has gotten traction in the EA and rationality communities, which has improved the way ideas and information spread throughout the community, how ideas get evaluated, and what kinds of projects get resources. More broadly, I think CFAR workshops have helped in establishing a set of common norms about what good reasoning and understanding look like, similar to the effect of the sequences on LessWrong.
    1. I think that it’s possible that the majority of the value of the EA and rationality communities comes from having that set of shared epistemic norms that allows them to reason collaboratively in a way that most other communities cannot (in the same way that what makes science work is a set of shared norms around what constitutes valid evidence and how new knowledge gets created).
    2. As an example of the importance of this: I think a lot of the initial arguments for why AI risk is a real concern were “weird” in a way that was not easily compatible with a naive empiricist worldview that I think is pretty common in the broader intellectual world.
      1. In particular, the arguments for AI risk are hard to test with experiments or empirical studies, but hold up from the perspective of logical and philosophical reasoning and are generated by a variety of good models of broader technological progress, game theory, and related areas of study. But for those arguments to find traction, they required a group of people with the relevant skills and habits of thought for generating, evaluating, and having extended intellectual discourse about these kinds of arguments.
  2. Training: A percentage of intro workshop participants (many of whom were already working on important problems within X-risk) have seen significant improvements in competence; as a result, they became substantially more effective in their work.
  3. Recruitment: CFAR has helped many people move from passive membership in the EA and rationality community to having strong social bonds in the X-risk network.

While I do think that CFAR has historically caused a significant amount of impact, I feel hesitant about this grant because I am unsure whether CFAR can continue to create the same amount of impact in the future. I have a few reasons for this:

  • First, all of its founding staff and many other early staff have left. I broadly expect organizations to get a lot worse once their early staff leaves.
    • Some examples of people who left after working there:
      • Julia Galef (left a few years ago to start the Update Project)
      • Andrew Critch (left to join first Jane Street, then MIRI, then founded CHAI and BERI)
      • Kenzi Askhie
      • Duncan Sabien
    • Anna Salamon has reduced her involvement in the last few years and seems significantly less involved with the broader strategic direction of CFAR (though she is still involved in some of the day-to-day operations, curriculum development, and more recent CFAR programmer workshops). [Note: After talking to Anna about this, I am now less certain of whether this actually applies and am currently confused on this point]
    • Duncan Sabien is no longer involved in day-to-day work, but still does some amount of teaching at intro workshops and programmer workshops (though I think he is planning to phase that out) and will help with the upcoming instructor training.
    • I think that Julia, Anna and Critch have all worked on projects of enormous importance, and their work over the last few years has clearly demonstrated a level of competence that makes me expect that CFAR will struggle to maintain its level of quality with their involvement significantly reduced.
  • From recent conversations with CFAR, I’ve gotten a sense that the staff isn’t interested in increasing the number of intro workshops, that the intro workshops don’t feel particularly exciting for the staff, and that most staff are less interested in improving the intro workshops than other parts of CFAR. This makes it less likely that those workshops will maintain their quality and impact, and I currently think that those workshops are likely one of the best ways for CFAR to have a large impact.
  • I have a general sense that CFAR is struggling to attract top talent, partially because some of the best staff left, and partially due to a general sense of a lack of forward momentum for the organization. This is a bad sign, because I think CFAR in particular benefits from having highly talented individuals teach at their workshops and serve as a concrete example of the skills they’re trying to teach.
  • My impression is that while the intro workshops were historically focused on instrumental rationality and personal productivity, the original CFAR staff was oriented quite strongly around truth-seeking. Core rationality concepts were conveyed indirectly by the staff in smaller conversations and in the broader culture of the organization. The current staff seems less oriented around that kind of epistemic rationality, and so I expect that if they continue their current focus on personal productivity and instrumental rationality, the epistemic benefits of CFAR workshops will be reduced significantly, and those are the benefits I care about most.

However, there are some additional considerations that led me to recommending this grant.

  • First, CFAR and MIRI are collaborating on a set of programmer-focused workshops that I am also quite positive on. I think those workshops are less directly influenced by counterfactual donations than the mainline workshops, since I expect MIRI to fund them in any case, but they do still rely on CFAR existing as an institution that can provide instructors. I am excited about the opportunities the workshops will enable in terms of curriculum development, since they can focus almost solely on epistemic rationality
  • Second, I think that if CFAR does not receive a grant now, there’s a good chance they’d be forced to let significant portions of their staff go, or take some other irreversible action. CFAR decided not to run a fundraiser last fall because they felt like they’d made significant mistakes surrounding a decision made by a community dispute panel that they set up and were responsible for, and they felt like it would be in poor taste to ask the community for money before they thoroughly investigated what went wrong and released a public statement.
    • I think this was the correct course of action, and I think overall CFAR’s response to the mistakes they made last year has been quite good.
    • The lack of a fundraiser led CFAR to have a much greater need for funding than usual, and a grant this round will likely make a significant difference in CFAR’s future.

In the last year, I had some concerns about the way CFAR communicated a lot of its insights, and I sensed an insufficient emphasis on a kind of robust and transparent reasoning that I don’t have a great name for. I don’t think the communication style I was advocating for is always the best way to make new discoveries, but is very important for establishing broader community-wide epistemic norms and enables a kind of long-term intellectual progress that I think is necessary for solving the intellectual challenges we’ll need to overcome to avoid global catastrophic risks. I think CFAR is likely to respond to last year’s events by improving their communication and reasoning style in this respect (from my perspective).

My overall read is that CFAR is performing a variety of valuable community functions and has a strong enough track record that I want to make sure that it can continue existing as an institution. I didn’t have enough time this grant round to understand how the future of CFAR will play out; the current grant amount seems sufficient to ensure that CFAR does not have to take any drastic action until our next grant round. By the next grant round, I plan to have spent more time learning and thinking about CFAR’s trajectory and future, and to have a more confident opinion about what the correct funding level for CFAR is.

Sorted by Click to highlight new comments since:
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

I think we should think carefully about the norm being set by the comments here.

This is an exceptionally transparent and useful grant report (especially Oliver Habryka's). It's helped me learn a lot about how the fund thinks about things, what kind of donation opportunities are available, and what kind of things I could (hypothetically if I were interested) pitch the LTF fund on in the future. To compare it to a common benchmark, I found it more transparent and informative than a typical GiveWell report.

But the fact that Habryka now must defend all 14 of his detailed write-ups against bikeshedding, uncharitable, and sometimes downright rude commenters seems like a strong disincentive against producing such reports in the future, especially given that the LTF fund is so time constrained.

If you value transparency in EA and want to see more of it (and you're not a donor to the LTF fund), it seems to me like you should chill out here. That doesn't mean don't question the grants, but it does mean you should:

  • Apply even more principle of charity than usual
  • Take time to phrase your question in the way that's easiest to answer
  • Apply some filter and don't ask unimportant questions
  • Use a tone that minimizes stress for the person you're questioning

I strongly agree with this. EA funds seemed to have a tough time finding grant makers who were both qualified and had sufficient time, and I would expect that to be partly because of the harsh online environment previous grant makers faced. The current team seems to have impressively addressed the worries people had in terms of donating to smaller and more speculative projects, and providing detailed write-ups on them. I imagine that in depth, harsh attacks on each grant decision will make it still harder to recruit great people for these committees, and mean those serving on them are likely to step down sooner. That's not to say we shouldn't be discussing the grants - presumably it's useful for the committee to hear other people's views on the grants to get more information about them. But following Ben's suggestions seems crucial to EA funds continuing to be a useful way of donating into the future. In addition, to try to engage more in collaborative truthseeking rather than adversarial debate, we might try to:

  • Focus on constructive information / suggestions for future grants rather than going into depth on what's wrong with grants already given.
  • Spend at least as much time describing which grants you think are good and how, so that they can be built on, as on things you disagree with.


I think it's great that the Fund is trending towards more transparency & a broader set of grantees (cf. November 2018 grant report, cf. July 2018 concerns about the Fund).

And I really appreciate the level of care & attention that Oli is putting towards this thread. I've found the discussion really helpful.

Relatedly, is Oli getting compensated for the work he's putting in to the Longterm Future Fund?

Seems good to move towards a regime wherein:

  • The norm is to write up detailed, public grant reports
  • Community members ask a bunch of questions about the grant decisions
  • The norm is that a representative of the grant-making staff fields all of these questions, and is compensated for doing so

I don't get compensated, though I also don't think compensation would make much of a difference for me or anyone else on the fund (except maybe Alex).

Everyone on the fund is basically dedicating all of their resources towards EA stuff, and is generally giving up most of their salary potential for working in EA. I don't think it would make super much sense for us to get more money, given that we are already de-facto donating everything above a certain threshold (either literally in the case of the two Matts, or indirectly by taking a paycut and working in EA).

I think if people give more money to the fund because they come to trust the decisions of the fund more, then that seems like it would incentivize more things like this. Also if people bring up strong arguments against any of the reasoning I explained above, then that is a great win, since I care a lot about our fund distributions getting better.

Got it.

The reason compensation seems good is that it formalizes the duty of engaging with the community's discourse, which probably pushes us further towards the above regime.

Right now, the community is basically banking on you & other fund managers caring a lot about engaging with the community. This is great, and it's great that you do.

Layering on compensation seems like a way of bolstering this engagement. If someone is compensated to do this engagement, then there's increased incentive for them to do it. (Though there's probably some weirdness around Goodhart-ing here.)

cf. Role of ombudsperson in public governance

Compensation is also good in case you ever retire and someone else with different financial needs takes over (but it doesn't seem super important - there are other things you could solve first).
I think that makes sense but in practice is something that makes more sense to handle through their day jobs. (If they went the route of hiring someone for whom managing the fund was their actual day job I'd agree that generally higher salaries would be good, for mostly the same reason they'd be good across the board in EA)

Now that the dust has settled a bit, I'm curious what Habryka & the other fund managers think of the level of community engagement that occurred on this report...

  • What kinds of engagement seemed helpful?
  • What kinds of engagement seemed unnecessary?
  • What kinds of engagement were emotionally expensive to address?
  • Does it seem sustainable to write up grantmaker reasoning at this level of detail, for each grantmaking round going forward?
  • Does it seem sustainable to engage with questions & comments from the community at this level of detail, for each grantmaking round going forward?
I have a bunch of complicated thoughts here. Overall I have been quite happy with the reception to this, and think the outcomes of the conversations on the post have been quite good. I am a bit more time-strapped than usual, so I will probably wait on writing a longer retrospective until I set aside a bunch of time to answer questions on the next set of writeups.

Agree with this, especially the comments about rudeness. This also means that I disagree with Oli's comment elsewhere in this thread:

that people should feel free to express any system-1 level reactions they have to these grants.

In line with what Ben says, I think people should apply a filter to their system-1 level reactions, and not express them whatever they are.

I think that people should feel comfortable sharing their system-1 expressions, in a way that does not immediately imply judgement.

I am thinking of stuff like the non-violent communication patterns, where you structure your observation in the following steps:

1. List a set of objective observations

2. Report your experience upon making those observations

3. Then your personal interpretations of those experiences and what they imply about your model of the world

4. Your requests that follow from those models

I think it's fine to stop part-way through this process, but that it's generally a good idea to not skip any steps. So I think it's fine to just list observations, and it's fine to just list observations and then report how you feel about those things, as long as you clearly indicate that this is your experience and doesn't necessarily involve judgement. But it's a bad idea to immediately skip to the request/judgement step.

OK, that is clarifying. Maybe your original comment could have been clearer, since this framing is quite different. The issue that you raise in this comment is a big debate, and this is maybe not the place to discuss it in detail. In any case, as stated my view is that people should think carefully before they comment, and not run with their immediate feelings on sensitive topics.

I don't agree with all of the decisions being made here, but I really admire the level of detail and transparency going into these descriptions, especially those written by Oliver Habryka. Seeing this type of documentation has caused me to think significantly more favorably of the fund as a whole.

Will there be an update to this post with respect to what projects actually fund following these recommendations? One aspect that I'm not clear on is to what extent CEA will "automatically" follow these recommendations and to what extent there will be significant further review.

I will make sure to update this post with any new information about whether CEA can actually make these grants. My current guess is that maybe 1-2 grants will not be logistically feasible, but the vast majority should have no problem.

I really admire the level of detail and transparency going into these descriptions, especially those written by Oliver Habryka

Hear, hear.

I feel proud of the commitment to epistemic integrity that I see here.

Thanks Habryka for raising the bar on the amount of detail given in grant explanations.

This is the feedback that I sent to Greg about his EA-Hotel application, published with his permission. (He also provided some good responses that I hope he can reply with)

Thoughts on the EA Hotel:

The EA Hotel seems broadly pretty promising, though I do also have a good amount of concerns. First, the reasons why I am excited about the EA Hotel:

Providing a safety net: I think psychological safety matters a lot for people being able to take risks and have creative thoughts. Given that I think most of the value of the EA community comes from potentially pursuing high-risk projects and proposing new unconventional ideas, improving things on this dimension strikes me as pretty key for the success of the overall community.

I expect the EA Hotel has a chance to serve as a cheap distributed safety net for a lot of people who are worried that if they start working on EA stuff, they will run out of money soon and will then potentially end up having to take drastic actions as they run out of money. The EA Hotel can both significantly extend those people's runway, but also soften the costs of running out of money significantly for anyone who is working on EA-related ideas.

Acting on his... (read more)

My response (edited from my email to Habryka)

I think I would be in favor of giving a grant that covers the runway of the hotel for the next year.

Wow this is awesome, thanks!

# Thoughts on the EA Hotel:

Thanks for your detailed response.

First, the reasons why I am excited about the EA Hotel:
Providing a safety net ...
Acting on historical interest ...
Building high-dedication cultures ...

All good reasons, eloquently put!

1. I think the people behind the EA Hotel where initially overeager to publicize the EA Hotel via broad media outreach in things like newspapers and others media outlets with broad reach.

I think this is based on an unfortunate misconception. The whole thing with the media interest has been quite surprising to us. We have never courted the media - quite the opposite in fact. It started with The Economist approaching us. This was whilst I was on holiday and out of communication. The first I heard about it was 3 days before they went to press (the piece appeared in print whilst I was still away). The journalist was told not to come to Blackpool. I spoke to them on the phone and said I wanted more time to think about it and discuss it with people. They went ahead anyway a... (read more)

I thought I'd share my impressions as someone who has spent significant time at the EA hotel

I think this makes them a particularly easy and promising target for people who tend to abuse that kind of trust relationship and who are looking for social influence.

Most of the people at the EA hotel have been involved in the movement for a while, so they already have reasonably well-developed positions already

it's plausible that the EA Hotel could form a geographically and memetically isolated group that is predisposed for conflict with the rest of the EA community in a way that could result in a lot of negative-sum conflict.

The EA hotel has a limit of 2 years free accommodation (although it is possible exceptions might be made). Most people tend to stay only a certain number of months given that it is Blackpool and not the most desirable location. Further, there are regularly visitors and there is frequent change over in the guests. I actually feel more memetically isolated in Australia than when I was at the EA hotel; especially since visiting London is relatively easy.

Generally high-dedication cultures are more likely to cause people to overcommit to to take drastic actions that they later regret

None of the projects that I am aware of having being undertaken at the EA hotel seemed to be especially high risk. But further than this, whoever is running the checkins will have an opportunity to steer people away from high risk projects.

Thanks for continuing to write up your thoughts in so much detail, Oliver; this is super interesting and useful stuff. When you say "Note: Greg responded to this and I now think this point is mostly false", I assume that "this" refers to the previous point (1) rather than the subsequent point (2)?
Yes, that corresponds to point (1), not point (2)
Jonas V
A point I'd personally want to add to Habryka's list: I'm currently unsure whether there is sufficiently good vetting of guests. Since the EA Hotel provides valuable services (almost) for free, it kind of acts as a de facto grantmaker, and runs the risk of funding people who are accidentally doing harm. There are reasons to think that harmful projects will be overrepresented in the application pool (Habryka also made some similar points). As I understand it, the EA Hotel is currently improving their vetting, which I think will be a step in the right direction, and could potentially resolve this issue.

I am hesitant about this. I think to serve as a functional social safety net that allows people to take high-risk actions (including in the social domain, in the form of criticisms of high-status people or institutions), I think a high barrier to entry for the EA-Hotel might drastically reduce the psychological safety it could provide to many people.

Jonas V
Interesting! I agree with the points you make, but I was hoping that good vetting wouldn't suffer from these problems.
I think if we had a vetting process that people could trust would reliably cause you to identify good people, even if they made a bunch of recent critical statements of high-status institutions or something in that reference class (or had their most recent project fail dramatically, etc.), then I think that might be fine. But I think having such a vetting process and having that vetting process have a very low false negative rate and having it be transparent that that vetting process is that good are difficult enough to make it too costly.
Jonas V
There already is a basic vetting process; I'd mostly welcome fairly gradual improvements to lower downside risk. (I think my initial comment sounded more like the bar should be fairly high, similar to that of, e.g., the LTFF. This is not what I intended to say; I think it should still be considerably lower.) I think even just explicitly saying something like "we welcome criticism of high-status people or institutions" would go a long way for both shaping people's perception of the vetting process and shaping the vetters' approach. That said, your arguments did update me in the direction "small changes to the vetting process seem better than large changes."
Jonas V
My impression of how the EA Hotel crew dealt with media attention was something like "better than many did in the early stages (including myself in the early stages of EAF) but (due to lack of experience or training) considerably worse than most EA orgs would do these days." There are many counterintuitive lessons to be learnt, many of which I still don't fully understand, either. However, since the initial media interest has abated, I think this isn't really relevant for current grants anyway.
Can you say what these lessons are? Would be good to have a write up of advice and I would like to see an EA forum post on this.

CEA's semi-internal media advice contains some valuable lessons. I was going to post a write-up on the EA Forum at some point, but given that media attention has been de-emphasized as an EA priority since, I decided against pursuing that (I also have some old "EA media strategy" presentation slides but unfortunately, they're in German). If lots of people thought this would be valuable, or if we learned that EA-Hotel-type issues occur on a regular basis, I'd consider it, though. (I also think much of my experience is only relevant to global poverty and animal welfare, not to AI or other cause areas.)

Personally, I'd be interested to see this writeup, and I'd definitely chip in with some of my thoughts if you posted it.

I'm happy that people are pushing back on some of these grants, and even happier that Habryka is responding to graciously. However, I'm concerned that some comments are bordering on unhelpfully personal.

I'd suggest that, when criticising a particular project, commentors should try to explain the rule or policy that would help grant-makers avoid the same problem in the future. That should also help us avoid making things personal.

Examples I stole from other comments and reworded:

-"I'm skeptical of the grant to X because I think grantmakers should recuse themselves from granting to their friends." (I saw this criticism but don't actually know who it's referring to.)

-"I don't think any EA Funds should be given to printing books that haven't been professionally edited."

-"I think that people like Lauren should have funds available after they burn out, but I don't think the Long-Term Future Fund is the right source of post-burnout funds."

I agree with this, but also think that people should feel free to express any system-1 level reactions they have to these grants. In my experience it can often be quite hard to formalize a critique into a concrete, operationalized set of policy changes, even if the critique itself is good and valid, and I don't think I want to force all commenters to fully formalize their beliefs before they can express them here.

I do think the end goal of the conversation should be a set of policies that the LTF-Fund can implement.

I have a weird mix of feelings and guesses here.

I think it's good on the margin for people to be able to express opinions without needing to formalize them into recommendations for the reason stated here. I think the overall conversation happening here is very important.

I do still feel pretty sad looking at the comments here — some of the commenters seem to not have a model of what they're incentivizing.

They remind me of the stereotype of a parent who's kid has moved away and grown up, and doesn't call very often. And periodically the kid does call, but the first thing they hear is the parent complaining "why don't you ever call me?", which makes the kid less likely to call home.

EA is vetting constrained.

EA is network constrained.

These are actual hard problems, that we're slowly addressing by building network infrastructure. The current system is not optimal or fair, but progress won't go faster by complaining about it.

It can potentially go faster via improvements in strategy, and re-allocating resources. But each of those improvements will come in a tradeoff. You could hire more grantmakers full-time, but those grantmakers are generall... (read more)

As a potential grant recipient (not in this round) I might be biased, but I feel like there is a clear answer to this. No one is able to level up without criticism, and the quality of your decisions will often be bottlenecked by the amount of feedback you receive.

Negative feedback isn't inherently painful. This is only true if there is an alief that failure is not acceptable. Of course the truth is that failure is necessary for progress, and if you truly understand this, negative feedback feels good. Even if it's in bad faith.

Given that grantmakers are essentially at the steering wheel of EA, we can't afford for those people to not internalize this. They need to know all the criticism to make a good decision, they should cherish it.

Of course we can help them get this state of mind by celebrating their willingness to open up to scrutiny, along with the scrutiny

I think on a post with 100+ comments the quality of decisions is more likely to be bottlenecked by the quality of feedback than the quantity. Being able to explain why you think something is a bad idea usually results in higher quality feedback, which I think will result in better decisions than just getting a lot of quick intuition-based feedback.

This is a strong set of grants, much stronger than the EA community would've been able to assemble a couple of years ago, which is great to see.

When will you be accepting further applications and making more grants?

I don't know yet. My guess is in around two months.

Answer turned out to be closer to 3 months.

Summary: This is the most substantial round of grant recommendations from the EA Long-Term Future Fund to date, so it is a good opportunity to evaluate the performance of the Fund after changes to its management structure in the last year. I am measuring the performance of the EA Funds on the basis of what I am calling 'counterfactually unique' grant recommendations. I.e., grant recommendations that, without the Long-Term Future Fund, individual donors nor larger grantmakers like the Open Philanthropy Project would have identified or funded.

Based on that measure, 20 of 23, or 87%, grant recommendations, worth $673,150 of $923,150, or ~73% of the money to be disbursed, are counterfactually unique. Having read all the comments, multiple concerns with a few specific grants came up, based on uncertainty or controversy in the estimation of value of these grant recommendations. Even if we exclude those grants from the estimate of counterfactually unique grant recommendations to make a 'conservative' estimate, 16 of 23, or 69.5%, of grants, worth $535,150 of $923,150, or ~58%, of the money to be disbursed, are counterfactually unique and fit into a more conservative, risk-averse approach... (read more)

You mean it functions like a venture capital fund or angel investor?
This is great! Thank you for the care & attention you put into creating this audit.

I'd be keen to hear a bit more more about the general process used for reviewing these grants. What did the overall process look like? Were participants interviewed? Were references collected? Were there general criteria used for all applications? Reasoning behind specific decisions is great, but also risks giving the impression that the grants were made just based on the opinions of one person, and that different applications might have gone through somewhat different processes.

Here is a rough summary of the process, it's hard to explain spreadsheets in words so this might end up sounding a bit confusing:

  • We added all the applications to a big spreadsheet, with a column for each fund member and advisor (Nick Beckstead and Jonas Vollmer) in which they would be encouraged to assign a number from -5 to +5 for each application
  • Then there was a period in which everyone individually and mostly independently reviewed each grant, abstaining if they had a conflict of interest, or voting positively or negatively if they thought the grant was a good or a bad idea
  • We then had a number of video-chat meetings in which we tried to go through all the grants that had at least one person who thought the grant was a good idea and had pretty extensive discussions about those grants. During those meetings we also agreed on next actions for follows ups, scheduling meetings with some of the potential grantees, reaching out to references etc. the results of which we would then discuss at the next all-hands meeting
  • Interspersed with the all-hands meetings I also had a lot of 1-on-1 meetings (with both other fund-members and grantees) in which I worked in detail through some of
... (read more)

You received almost 100 applications as far as I'm aware, but were able to fund only 23 of them. Some other projects were promising according to you, but you didn't have time to vet them all. What other reasons did you have for rejecting applications?

Hmm, I don't think I am super sure what a good answer to this would look like. Here are some common reasons for why I think a grant was not a good idea to recommend:

  • The plan seemed good, but I had no way of assessing the applicant without investing significant amounts of time that I had not available (which is likely why you see a skew towards people the granting team had some past interactions with in the grants above)
  • The mainline outcome of the grant was good, but there were potential negative consequences that the applicant did not consider or properly account for, and I did not feel like I could cause the applicant to understand the downside risk they have to account for without investing significant effort and time
  • The grant was only tenuously EA-related and seemed to have been submitted to a lot of applications relatively indiscriminately
  • I was unable to understand the goals, implementation or other details of the grant
  • I simply expected the proposed plan to not work, for a large variety of reasons. Here are some of the most frequent:
    • The grant was trying to achieve something highly ambitious while seeming to allocate very little resources to achieving that outcome
    • The grantee had a track record of work that I did not consider to be of sufficient quality to achieve what they set out to do
  • In some cases the applicant asked for less than our minimum grant amount of $10,000

Thanks for the transparent answers.

The plan seemed good, but I had no way of assessing the applicant without investing significant amounts of time that I had not available (which is likely why you see a skew towards people the granting team had some past interactions with in the grants above)

This in particular strikes me as understandable but very unfortunate. I'd strongly prefer a fund where happening to live near or otherwise know a grantmaker is not a key part of getting a grant. Are there any plans or any way progress can be made on this issue?

In some cases the applicant asked for less than our minimum grant amount of $10,000

This also strikes me as unfortunate and may lead to inefficiently inflated grant requests in the future, though I guess I can understand why the logistics behind this may require it. It feels intuitively weird though that it is easier to get $10K than it is to get $1K.

This in particular strikes me as understandable but very unfortunate. I'd strongly prefer a fund where happening to live near or otherwise know a grantmaker is not a key part of getting a grant.

I personally have never interacted directly with the grantees of about 6 of the 14 grants that I have written up, so it it not really about knowing the grantmakers in person. What does matter a lot are the second degree connections I have to those people (and that someone on the team had for the large majority of applications), as well as whether the grantees had participated in some of the public discussions we've had over the past years and demonstrated good judgement (e.g. EA Forum & LessWrong discussions).

I don't think you should model the situation as relying on knowing a grantmaker in-person, but you should think that testimonials and referrals from people that the grantmakers trust matter a good amount. That trust can be built via a variety of indirect ways, some of which are about knowing them in person and having a trust relationship that has been built via personal contact, but a lot of the time that trust comes from the connecting person having made a variety o... (read more)

This also strikes me as unfortunate and may lead to inefficiently inflated grant requests in the future, though I guess I can understand why the logistics behind this may require it. It feels intuitively weird though that it is easier to get $10K than it is to get $1K.

A rough fermi I made a few days ago suggests that each grant we make comes with about $2000 of overhead from CEA for making the grants in terms of labor cost plus some other risks (this is my own number, not CEAs estimate). So given that overhead, it makes some amount of sense that it's hard to get $1k grants.

Wow! This is an order of magnitude larger than I expected. What's the source of the overhead here?

Here is my rough fermi:

My guess is that there is about one full-time person working on the logistics of EA Grants, together with about half of another person lost in overhead, communications, technology (EA Funds platform) and needing to manage them.

Since people's competence is generally high, I estimated the counterfactual earnings of that person at around $150k, with an additional salary from CEA of $60k that is presumably taxed at around 30%, resulting in a total loss of money going to EA-aligned people of around ($150k + 0.3 * $60k) * 1.5 = $252k per year [Edit: Updated wrong calculation]. EA Funds has made less than 100 grants a year, so a total of about $2k - $3k per grant in overhead seems reasonable.

To be clear, this is average overhead. Presumably marginal overhead is smaller than average overhead, though I am not sure by how much. I randomly guessed it would be about 50%, resulting in something around $1k to $2k overhead.

If one person-year is 2000 hours, then that implies you're valuing CEA staff time at about $85/hour. Your marginal cost estimate would then imply that a marginal grant takes about 12-24 person-hours to process, on average, all-in.

This still seems higher than I would expect given the overheads that I know about (going back and forth about bank details, moving money between banks, accounting, auditing the accounting, dealing with disbursement mistakes, managing the people doing all of the above). I'm sure there are other overheads that I don't know about, but I'm curious if you (or someone from CEA) knows what they are?

[Not trying to imply that CEA is failing to optimize here or anything—I'm mostly curious plus have a professional interest in money transfer logistics—so feel free to ignore]

I actually think the $10k grant threshold doesn't make a lot of sense even if we assume the details of this "opportunity cost" perspective are correct. Grants should fulfill the following criterion:

"Benefit of making the grant" ≥ "Financial cost of grant" + "CEA's opportunity cost from distributing a grant"

If we assume that there are large impact differences between different opportunities, as EAs generally do, a $5k grant could easily have a benefit worth $50k to the EA community, and therefore easily be worth the $2k of opportunity cost to CEA. (A potential justification of the $10k threshold could argue in terms of some sort of "market efficiency" of grantmaking opportunities, but I think this would only justify a rigid threshold of ~$2k.)

IMO, a more desirable solution would be to have the EA Fund committees factor in the opportunity cost of making a grant on a case-by-case basis, rather than having a rigid "$10k" rule. Since EA Fund committees generally consist of smart people, I think they'd be able to understand and implement this well.

This sounds pretty sensible to me. On the other hand, if people are worried about it being harder for people who are already less plugged in to networks to get funding, you might not want an additional dimension on which these harder-to-evaluate grants could lose out compared to easier to evaluate ones (where the latter end up having a lower minimum threshold).

It also might create quite a bit of extra overhead for granters having to decide the opportunity cost case by case, which could reduce the number of grants they can make, or again push towards easier to evaluate ones.

Jonas V
I tend to think that the network constraints are better addressed by solutions other than ad-hoc fixes (such as more proactive investigations of grantees), though I agree it's a concern and it updates me a bit towards this not being a good idea. I wasn't suggesting deciding the opportunity cost case by case. Instead, grant evaluators could assume a fixed cost of e.g. $2k. In terms of estimating the benefit of making the grant, I think they do that already to some extent by providing numerical ratings to grants (as Oliver explains here). Also, being aware of the $10k rule already creates a small amount of work. Overall, I think the additional amount of work seems negligibly small. ETA: Setting a lower threshold would allow us to a) avoid turning down promising grants, and b) remove an incentive to ask for too much money. That seems pretty useful to me.
It's not at all clear to me why the whole $150k of a counterfactual salary would be counted as a cost. The most reasonable (simple) model I can think of is something like: ($150k * .1 + $60k) * 1.5 = $112.5k where the $150k*.1 term is the amount of salary they might be expected to donate from some counterfactual role. This then gives you the total "EA dollars" that the positions cost whereas your model seems to combine "EA dollars" (CEA costs) and "personal dollars" (their total salary).
Hmm, I guess it depends a bit on how you view this. If you model this in terms of "total financial resources going to EA-aligned people", then the correct calculation is ($150k * 1.5) plus whatever CEA loses in taxes for 1.5 employees. If you want to model it as "money controlled directly by EA institutions" then it's closer to your number. I think the first model makes more sense, which does still suggest a lower number than what I gave above, so I will update.
I don't particularly want to try to resolve the disagreement here, but I'd think value per dollar is pretty different for dollars at EA institutions and for dollars with (many) EA-aligned people [1]. It seems like the whole filtering/selection process of granting is predicated on this assumption. Maybe you believe that people at CEA are the type of people that would make very good use of money regardless of their institutional affiliation? [1] I'd expect it to vary from person to person depending on their alignment, commitment, competence, etc.
I think you have some math errors: * $150k * 1.5 + $60k = $285k rather than $295k * Presumably, this should be ($150k + $60k) * 1.5 = $315k ?
Ah, yes. The second one. Will update.
Jonas V
(moved this comment here)
I agree this creates unfortunate incentives for EAs to burn resources living in high cost-of-living areas (perhaps even while doing independent research which could in theory be done from anywhere!) However, if I was a grantmaker, I can see why this arrangement would be preferable: Evaluating grants feels like work and costs emotional energy. Talking to people at parties feels like play and creates emotional energy. For many grantmakers, I imagine getting to know people in a casual environment is effectively costless, and re-using that knowledge in the service of grantmaking allows more grants to be made. I suspect there's low-hanging fruit in having the grantmaking team be geographically distributed. To my knowledge, at least 3 of these 4 grantmakers live in the Bay Area, which means they probably have a lot of overlap in their social network. If the goal is to select the minimum number of supernetworkers to cover as much of the EA social network as possible, I think you'd want each person to be located in a different geographic EA hub. (Perhaps you'd want supernetworkers covering disparate online communities devoted to EA as well.) This also provides an interesting reframing of all the recent EA Hotel discussion: Instead of "Fund the EA Hotel", maybe the key intervention is "Locate grantmakers in low cost-of-living locations. Where grant money goes, EAs will follow, and everyone can save on living expenses." (BTW, the EA Hotel is actually a pretty good place to be if you're an aspiring EA supernetworker. I met many more EAs during the 6 months I spent there than my previous 6 months in the Bay Area. There are always people passing through for brief stays.)
To my knowledge, at least 3 of these 4 grantmakers live in the Bay Area, which means they probably have a lot of overlap in their social network.

That is incorrect. The current grant team was actually explicitly chosen on the basis of having non-overlapping networks. Besides me nobody lives in the Bay Area (at least full time). Here is where I think everyone is living:

  • Matt Fallshaw: Australia (but also travels a lot)
  • Helen Toner: Georgetown (I think)
  • Alex Zhu: No current permanent living location, travels a lot, might live in Boulder starting a few weeks from now
  • Matt Wage: New York

I was also partially chosen because I used to live in Europe and still have pretty strong connections to a lot of european communities (plus my work on online communities making my network less geographically centralized).

Good to know!
Isn't Matt in HK?
He sure was on weird timezones during our meetings, so I think he might be both? (as in, flying between the two places)
Update: I was just wrong, Matt is indeed primarily HK
JP Addison
Boy, there are two Matts in that list.
Evaluating grants feels like work and costs emotional energy. Talking to people at parties feels like play and creates emotional energy. For many grantmakers, I imagine getting to know people in a casual environment is effectively costless, and re-using that knowledge in the service of grantmaking allows more grants to be made.

At least for me this doesn't really resonate with how I am thinking about grantmaking. The broader EA/Rationality/LTF community is in significant chunks a professional network, and so I've worked with a lot of people on a lot of projects over the years. I've discussed cause prioritization questions on the EA Forum, worked with many people at CEA, tried to develop the art of human rationality on LessWrong, worked with people at CFAR, discussed many important big picture questions with people at FHI, etc.

The vast majority of my interactions with people do not come from parties, but come from settings where people are trying to solve some kind of problem, and seeing how others solve that problem is significant evidence about whether they can solve similar problems.

It's not that I hang out with lots of people at parties, make lots of friends... (read more)

The plan seemed good, but I had no way of assessing the applicant without investing significant amounts of time that I had not available (which is likely why you see a skew towards people the granting team had some past interactions with in the grants above)

I'm pretty concerned about this. I appreciate that there will always be reasonable limits to how long someone can spend vetting grant applications, but I think EA funds should not be hiring fund managers who don't have sufficient time to vet applications from people they don't already know - being able to do this should be a requirement of the job, IMO. Seconding Peter's question below, I'd be keen to hear if there are any plans to make progress on this.

If you really don't have time to vet applicants, then maybe grant decisions should be made blind, purely on the basis of the quality of the proposal. Another option would be to have a more structured/systematic approach to vetting applicants themselves, which could be anonymous-ish: based on past achievements and some answers to questions that seem relevant and important.

but I think EA funds should not be hiring fund managers who don't have sufficient time to vet applications from people they don't already know

To be clear, we did invest time into vetting applications from people we didn't know, we just obviously have limits to how much time we can invest. I expect this will be a limiting factor for any grant body.

My guess is that if you don't have any information besides the application info, and the plan requires a significant level of skill (as the vast majority of grants do), you have to invest at least an additional 5, often 10, hours of effort into reaching out to them, performing interviews, getting testimonials, analyzing their case, etc. If you don't do this, I expect the average grant to be net negative.

Our review period lasted about one month. At 100 applications, assuming that you create an anonymous review process, this would have resulted in around 250-500 hours of additional work, which would have made this the full-time job for 2-3 of the 5 people on the grant board, plus the already existing ~80 hours of overhead this grant round required from the board. You likely would have filtered out about 50 of them ... (read more)

Thanks for your detailed response Ollie. I appreciate there are tradeoffs here, but based on what you've said I do think that more time needs to be going into these grant reviews.

It don't think it's unreasonable to suggest that it should require 2 people full time for a month to distribute nearly $1,o00,000 in grant funding, especially if the aim is to find the most effective ways of doing good/influencing the long-term future. (though I recognise that this decision isn't your responsibility personally!) Maybe it is very difficult for CEA to find people with the relevant expertise who can do that job. But if that's the case, then I think there's a bigger problem (the job isn't being paid well enough, or being valued highly enough by the community), and maybe we should question the case for EA funds distributing so much money.

I strongly agree that I would like there to be more people who have the competencies and resources necessary to assess grants like this. With the Open Philanthropy Project having access to ~10 billion dollars, the case for needing more people with that expertise is pretty clear, and my current sense is that there is a broad consensus in EA that finding more people for those roles is among, if not the, top priority.

I think giving less money to EA Funds would not clearly improve this situation from this perspective at all, since most other granting bodies that exist in the EA space have an even higher (funds-distributed)/staff ratio than this.

The Open Philanthropy Project has about 15-20 people assessing grants, and gives out at least 100 million dollars a year, and likely aims to give closer to a $1 billion dollars a year given their reserves.

BERI has maybe 2 people working full-time on grant assessment, and my current guess is that they give out about $5 million dollars of grants a year

My guess is that GiveWell also has about 10 staff assessing grants full-time, making grants of about $20 million dollars

I think at the current level of team-member-involvement, and since I do thi... (read more)

Note that BERI has only existed for a little over 2 years, and their grant-making has been pretty lumpy, so I don't think they've yet reached any equilibrium grant-making rate (one which could be believably expressed in terms of $X dollars / year).

I agree. Though I think I expect the ratio of funds-distributed/staff to roughly stay the same, at least for a bit, and probably go up a bit.

I think older and larger organizations will have smaller funds-distributed/staff ratios, but I think that's mostly because coordinating people is hard and marginal productiveness of a hire goes down a lot after the initial founders, so you need to hire a lot more people to produce the same quality of output.

I would be in favour of this fund using ~5% of its money to pay for staff costs, including a permanent secretariat. The secretariat would probably decrease pressure on grantmakers a little, and improve grant/feedback quality a little, which makes the costs seem worth it. (I know you've already considered this and I want to encourage it!)

I imagine the secretariat would:

-Handle the admin of opening and advertising a funding round

-Respond to many questions on the Forum, Facebook, and by email, and direct more difficult questions to the correct person

-Coordinate the writing of Forum posts like this

-Take notes on what additional information grantmakers would like from applicants, contact applicants with follow-up questions, and suggest iterations of the application form

-(potentially) Manage handover to new grantmakers when current members step down

-(potentially) Sift through applications and remove those which are obviously inappropriate for the Long Term Future Fund

-(potentially) Provide a couple of lines of fairly generic but private feedback for applicants

This strikes me as a great, concrete suggestion. As I tell a lot of people, great suggestions in EA only go somewhere if someone is done with them. I would strongly encourage you to develop this suggestion into its own article on the EA Forum about how the EA Funds can be improved. Please let me know if you are interested in doing so, and I can help out. If you don't think you'll have time to develop this suggestion, please let me know, as I would be interested in doing that myself if you don't have the time.
The way the management of the EA Funds is structured to me makes sense within the goals set for the EA Funds. So I think the situation in which 2 people are paid full-time for one month to evaluate EA Funds applications makes sense is one where 2 of the 4 volunteer fund managers took a month off from their other positions to evaluate the applications. Finding 2 people from out of the blue to evaluate applications for one month without continuity with how the LTF Fund has been managed seems like it'd be too difficult to effectively accomplish in the timeframe of a few months. In general, one issue the EA Funds face other granting bodies in EA don't face is the donations come from many different donors. This consequently means how much the EA Funds receive and distribute, and how it's distributed, is much more complicated than ones the CEA or a similar organization typically faces.

Thanks for the care & attention you're putting towards all of these replies!

I do think that at our current grant volume, we should invest more resources into building infrastructure for vetting grant applications.

Strong +1.

One issue with this is the fund managers are unpaid volunteers who have other full-time jobs, so being a fund manager isn't a "job" in the most typical sense. Of course a lot of people think it should be treated like one though. When this came up in past discussions regarding how the EA Funds could be structured better, suggestions like hiring a full-time fund manager came up against trade-offs against other priorities for the EA Funds, like not spending too much overheard on them, or having the diversity of perspectives that comes with multiple volunteer fund managers.

I applaud the explanations of the decisions for the grants and also the responses to the questions. Now that things have calmed down, since the EA Long Term Future Fund team suggested that requests for feedback on unsuccessful grants be made publicly, I am doing that.

My proposal was to further investigate a new cause area, namely resilience to catastrophes that could disable electricity regionally or globally, including extreme solar storm, high-altitude electromagnetic pulses (caused by nuclear detonations), or a narrow AI computer virus. Since nearly everything is dependent on electricity, including pulling fossil fuels out of the ground, industrial civilization could grind to a halt. Many people have suggested hardening the grid to these catastrophes, but this would cost tens of billions of dollars. However, getting prepared for quickly providing food, energy, and communications needs in a catastrophe would cost much less money and provide much of the present generation (lifesaving) and far future (preservation of anthropological civilization) benefits. I have made a Guesstimate model assessing the cost-effectiveness of work to improve long-term future outcomes given one of the... (read more)

(Note, I am currently more time-constrained than I had hoped to be when writing these responses, so the above was written a good bit faster and with less reflection than my other pieces of feedback. This means errors and miscommunication is more likely than usual. I apologize for that.) I ended up writing some feedback to Jeffrey Ladish, which covered a lot of my thoughts on ALLFED.  My response to Jeffrey Building off of that comment, here are some additional thoughts:  * As I mentioned in the response linked above, I currently feel relatively hesitant about civilizational collapse scenarios and so find the general cause area of most of ALLFED's work to be of comparatively lower importance than the other areas I tend to recommend grants in * Most of ALLFED's work does not seem to help me resolve the confusions I listed in the response linked above, or provide much additional evidence for any of my cruxes, but instead seems to assume that the intersection of civilizational collapse and food shortages is the key path to optimize for. At this point, I would be much more excited about work that tries to analyze civilizational collapse much more broadly, instead of assuming such a specific path.  * I have some hesitations about the structure of ALLFED as an organization. I've had relatively bad experiences interacting with some parts of your team and heard similar concerns from others. The team also appears to be partially remote, which I think is a major cost for research teams, and have its primary location be in Alaska where I expect it will be hard for you to attract talent and also engage with other researchers on this topic (some of these models are based on conversations I've had with Finan who used to work at ALLFED, but left because of it being located in Alaska).  * I generally think ALLFED's work is of decent quality, helpful to many and made with well-aligned intentions, I just don't find it's core value proposition compelling enough to be excited ab

Thank you for you recent post and your ALLFED feedback.

I have made my request for such publicly so also responding publicly, as such openness can only be beneficial to the investigation and advancing of the causes we are passionate about.

We appreciate your view of ALLFED’s work being of “decent quality, helpful to many and made with well-aligned intention”.

We also appreciate many good points raised in your feedback, and would like to comment on them as follows.

As I mentioned in the response linked above, I currently feel relatively hesitant about civilizational collapse scenarios and so find the general cause area of most of ALLFED's work to be of comparatively lower importance than the other areas I tend to recommend grants in

People’s intuition on the long-term future impact of these type of catastrophes and the tractability of reducing that impact with money varies tremendously.

One possible mechanism for extinction from nuclear winter is as follows. It is tempting to think that if there is enough stored food to keep the population alive for five years until agriculture recovers, that 10% of people will survive. However, if the food is distributed evenly, then everyone will ... (read more)

Regarding the donation to Lauren Lee:

To the extent that one thinks that funding the runways of burnt-out and/or transitioning EAs is a good idea to enable risk-neutral career decisions (which I do!), I'd note that funding (projects like) the EA Hotel seems like a promising way to do so. The marginal per-EA cost of supplying runway is probably lower with shared overhead and low COL like that.

This could also help free up a significant amount of donation money. My guess is that a central entity that could be (more) risk-neutral than individual EAs would be a more efficient insurer of EA runway needs than individual EAs. Many EAs will never use their runways, and this will mean, at best, significantly delayed donations, which is a high opportunity cost. If runway-saving EAs would otherwise donate (part of) their runways (which I would if I knew the EA community would provide one if needed), there could be net gains in EA cashflow due to the efficiency of a central insurer.

I'm not super confident in this, and I could be wrong for a lot of reasons. Obviously, runways aren't purely altruistic, so one shouldn't expect all runway money to go to donations. And it might be hard or undesirable for EA to provide certain kinds of runway due to, e.g., moral hazard. It might also be hard for EA as a community to provide runways with any reasonable assurance that the outcome will be altruistic (I take this to be one of the main objections to the EA Hotel). Still, I think the idea of insuring EA runway needs could be promising.

Am certainly open to considering this business model for the hotel.
This is interesting, though the moral hazard / free-riding consideration seems like a big problem.
I agree that moral hazard is, but you could also imagine an excludable EA insurance scheme that reduced free-riding. E.g., pay $X/month and if you lose your job you can live here for up to a year. But since the employed EA community is not as diversified as the whole market, employed EAs may be more liable to systemic shocks that render the insurer insolvent. But of course, there's reinsurance...
The hotel did apply. It's about $7500 per person per year


I’ve gotten a sense that the staff isn’t interested in increasing the number of intro workshops, that the intro workshops don’t feel particularly exciting for the staff, and that most staff are less interested in improving the intro workshops than other parts of CFAR. This makes it less likely that those workshops will maintain their quality and impact, and I currently think that those workshops are likely one of the best ways for CFAR to have a large impact.
CFAR is struggling to attract top talent, partially because some of the best staff left, and partially due to a general sense of a lack of forward momentum for the organization. This is a bad sign, because I think CFAR in particular benefits from having highly talented individuals teach at their workshops and serve as a concrete example of the skills they’re trying to teach.

Why a large, unrestricted grant to CFAR, given these concerns? Would a smaller grant catalyze changes such that the organization becomes cash-flow positive?

By the next grant round, I plan to have spent more time learning and thinking about CFAR’s trajectory and future, and to have a more confident opinion about what the correct funding level for CFAR
... (read more)

I assume that by "cash-flow positive", you mean supported by fees from workshop participants?

I don't consider that to be a desirable goal for CFAR.

Habryka's analysis focuses on CFAR's track record. But CFAR's expected value comes mainly from possible results that aren't measured by that track record.

My main reason for donating to CFAR is the potential for improving the rationality of people who might influence x-risks. That includes mainstream AI researchers who aren't interested in the EA and rationality communities. The ability to offer them free workshops seems important to attracting the most influential people.

Yes, that's roughly what I mean. I'm gesturing towards "getting to a business structure where it's straightforward to go into survival mode, wherein CFAR maintains core staff & operations via workshop fees." Seems like in that configuration, the org wouldn't be as buffeted by the travails of a 6-month or 12-month fundraising cycle. I agree that being entirely supported by workshop fees wouldn't be a desirable goal-state for CFAR. But having a "survival mode" option at the ready for contingencies seems good.
Why a large, unrestricted grant to CFAR, given these concerns? Would a smaller grant catalyze changes such that the organization becomes cash-flow positive?

I have two interpretations of what your potential concerns here might be, so might be good to clarify first. Which of these two interpretations is closer to what you mean?

1. "Why give CFAR such a large grant at all, given that you seem to have a lot of concerns about their future"

2. "Why not give CFAR a grant that is conditional on some kind of change in the organization?"

I'm curious about both (1) and (2), as they both seem like plausible alternatives that you may have considered.

Seems good.

1. "Why give CFAR such a large grant at all, given that you seem to have a lot of concerns about their future"

I am overall still quite positive on CFAR. I have significant concerns, but the total impact CFAR had over the course of its existence strikes me as very large and easily worth the resources it has taken up so far.

I don't think it's the correct choice for CFAR to take irreversible action right now because they correctly decided to not run a fall fundraiser, and I still assign significant probability to CFAR actually being on the right track to continue having a large impact. My model here is mostly that whatever allowed CFAR to have a historical impact did not break, and so will continue producing value of the same type.

2. "Why not give CFAR a grant that is conditional on some kind of change in the organization?"

I considered this for quite a while, but ultimately decided against it. I think grantmakers should generally be very hesitant to make earmarked or conditional grants to organizations, without knowing the way that organization operates in close detail. Some things that might seem easy to change from the outside often turn o... (read more)

This is super helpful, thanks! Perhaps a crux here is whether whatever mechanism historically drove CFAR's impact has already broken or not. (Just flagging, doesn't seem important to resolve this now.)
Yeah, that's what I intended to say. "In the world where I come to the above opinion, I expect my crux will have been that whatever made CFAR historically work, is still working"

Was wondering if you can explain more about the reasoning for funding Connor Flexman. Right now, the write-up doesn't explain much and makes me curious what "independent research" means. Also would be interested in learning what past projects Connor has worked on that led to this grant.

The primary thing I expect him to do with this grant is to work together with John Salvatier on doing research on skill transfer between experts (which I am partially excited about because that's the kind of thing that I see a lot of world-scale model building and associated grant-making being bottlenecked on).

However, as I mentioned in the review, if he finds that he can't contribute to that as effectively as he thought, I want him to feel comfortable pursuing other research avenues. I don't currently have a short-list of what those would be, but would probably just talk with him about what research directions I would be excited about, if he decides to not collaborate with John. One of the research projects he suggested was related to studying historical social movements and some broader issues around societal coordination mechanisms that seemed decent.

I primarily know about the work he has so far produced with John Salvatier, and also know that he demonstrated general competence in a variety of other projects, including making money managing a small independent hedge fund, running a research project for the Democracy Defense Fund, doing some research at Brown university, and participating in some forecasting tournaments and scoring well.

I'd like to challenge the downside estimate re: HPMoR distribution funding.

So I felt comfortable recommending this grant, especially given its relatively limited downside

I think that funding this project comes with potentially significant PR and reputational risk, especially considering the goals for the fund. It seems like it might be a much better fit for the Meta fund, rather than for the fund that aims to: "support organizations that work on improving long-term outcomes for humanity".

Could you say a bit more about what kind of PR and reputational risks you are imagining? Given that the grant is done in collaboration with the IMO and EGMO organizers, who seem to have read the book themselves and seem to be excited about giving it out as a prize, I don't think I understand what kind of reputational risks you are worried about.

I am not OP but as someone who also has (minor) concerns under this heading:

  • Some people judge HPMoR to be of little artistic merit/low aesthetic quality
  • Some people find the subcultural affiliations of HPMoR off-putting (fanfiction in general, copious references to other arguably low-status fandoms)

If the recipients have negative impressions of HPMoR for reasons like the above, that could result in (unnecessarily) negative impressions of rationality/EA.

Clearly, there also many people that like HPMoR and don't have the above concerns. The key question is probably what fraction of recipients will have positive, neutral and negative reactions.

Hmm, so my model is that the books are given out without significant EA affiliation, together with a pamphlet for SPARC and ESPR. I also know that HPMoR is already relatively widely known among math olympiad participants. Those together suggest that it's unlikely this would cause much reputational damage to the EA community, given that none of this contains an explicit reference to the EA community (and shouldn't, as I have argued below).

The outcome might be that some people might start disliking HPMoR, but that doesn't seem super bad and of relatively little downside. Maybe some people will start disliking CFAR, though I think CFAR on net benefits a lot more from having additional people who are highly enthusiastic about it, than it suffers from people who kind-of dislike it.

I have some vague feeling that there might be some more weird downstream effects of this, but I don't think I have any concrete models of how they might happen, and would be interested in hearing more of people's concerns.

Not the book giveaway itself, but posting grant information like this can be very bad PR.
I think I agree, but why do you think so?
(Responding to the second point about which fund is a better fit for this, will respond to the first point separately) I am broadly confused how to deal with the "which fund is a better fit?" question. Since it's hard to influence the long-term future I expect a lot of good interventions to go via the path of first introducing people to the community, building institutions that can improve our decision-making, and generally opting for building positive feedback loops and resources that we can deploy as soon as concrete opportunities show up. My current guess is that we should check in with the Meta fund and their grants to make sure that we don't make overlapping grants and that we communicate any concerns, but that as soon as there is an application that we think is worth it from the perspective of the long-term-future that the Meta fund is not covering, that we should feel comfortable filling it, independently of whether it looks a bit like EA-Meta. But I am open to changing my mind on this.
Could this be straightforwardly simplified by bracketing out far future meta work as within the remit of the Long Term Future Fund, and all other meta work (e.g. animal welfare institution-building, global development institution-building) as within the remit of the Meta Fund? Not sure if that would cleave reality at the joints, but seems like it might.

I actually think that as long as you communicate potential downside risks, there is a lot of value in having independent granting bodies look over the same pool of applications.

I think a single granting body is likely to end up missing a large number of good opportunities, and general intuitions arounds hits-based giving make me think that encouraging independence here is better than splitting up every grant into only one domain (this does rely on those granting bodies being able to communicate clearly around downside risk, which I think we can achieve).

Rohin Shah
Is this different from having more people on a single granting body? Possibly with more people on a single granting body, everyone talks to each other more and so can all get stuck thinking the same thing, whereas they would have come up with more / different considerations had they been separate. But this would suggest that granting bodies would benefit from splitting into halves, going over grants individually, and then merging at the end. Would you endorse that suggestion?
I don't think you want to go below three people for a granting body, to make sure that you can catch all the potential negative downsides of a grant. My guess is that if you have 6 or more people it would be better to split it into two independent grant teams.
Peter Wildeford
Yes, this is a great idea to help reduce bias in grantmaking.

Forgive me if you've written it up elsewhere, but do you have a plan for follow-ups? In particular what success looks like in each case.

Thanks for the detailed writeups and for investigating so many grants.

I would quite like us to do follow-ups, but the LTF-Fund is primarily time-constrained and solid follow-ups require a level of continuous engagement that I think currently would be quite costly for any of the current fund members.

I do think we might want to look into adding some additional structure to the fund where we maybe employ someone for half-time to follow up with grantees, perform research, help with the writeups, etc. But I haven't thought that through yet.

For now, I expect to perform follow-up evaluations when the same people re-apply for a new grant, in which case I will want to look in detail into how the past grants we gave them performed. I expect a lot of our grantees to reapply, so I do expect this to result in a good amount of coverage. This way there are also real stakes to the re-evaluation, which overall makes me think that I would be more likely to do a good job at them (as well as anyone else who might take them on).

Myself and the other organizers of Catalyst (the eventual name of "A biorisk summit for the Bay Area biotech industry, DIY biologists, and biosecurity researchers") recently wrote up a retrospective on the project, which may be of interest for people trying to understand how our LTFF funding was put to use.

Mr. Habryka,

I do not believe the $28,000 grant to buy copies of HPMOR meets the evidential standard demanded by effective altruism. “Effective altruism is about answering one simple question: how can we use our resources to help others the most? Rather than just doing what feels right, we use evidence and careful analysis to find the very best causes to work on.” With all due respect, it seems to me that this grant feels right but lacks evidence and careful analysis.

The Effective Altruism Funds are "for maximizing the effectiveness of your donations" according to the homepage. This grant's claim that buying copies of HPMOR is among the most effective ways to donate $28,000 by way of improving the long-term future rightly demands a high standard of evidence.

You make two principal arguments in justifying the grant. First, the books will encourage the Math Olympiad winners to join the EA community. Second, the book swill teach the Math Olympiad winners important reasoning skills.

If the goal is to encourage Math Olympiad winners to join the Effective Altruism community, why are they being given a book that has little explicitly to do with Effective Altruism? The Life You Can Save,... (read more)

Dear Morgan,

In this comment I want to address the following paragraph (#3).

I also want to point out that the fact that EA Russia has made oral agreements to give copies of the book before securing funding is deeply unsettling, if I understand the situation correctly. Why are promises being made in advance of having funding secured? This is not how a well-run organization or movement operates. If EA Russia did have funding to buy the books and this grant is displacing that funding, then what will EA Russia spend the original $28,000 on? This information is necessary to evaluate the effectiveness of this grant and should not be absent.

I think that it is a miscommunication on my side.

EA Russia has the oral agreements with [the organizers of math olympiads]...

We contacted organizers of math olympiads and asked them whether they would like to have HPMoRs as a prize (conditioned on us finding a sponsor). We didn't promise anything to them, and they do not expect anything from us. Also, I would like to say that we hadn't approached them as the EAs (as I am mindful of the reputational risks).

Dear Morgan,

In this comment I want to address the following paragraph (related to #2).

If the goal is to encourage Math Olympiad winners to join the Effective Altruism community, why are they being given a book that has little explicitly to do with Effective Altruism? The Life You Can Save, Doing Good Better, and _80,000 Hours_are three books much more relevant to Effective Altruism than Harry Potter and the Methods of Rationality. Furthermore, they are much cheaper than the $43 per copy of HPMOR. Even if one is to make the argument that HPMOR is more effective at encouraging Effective Altruism — which I doubt and is substantiated nowhere — one also has to go further and provide evidence that the difference in cost of each copy of HPMOR relative to any of the other books I mentioned is justified. It is quite possible that sending the Math Olympiad winners a link to Peter Singer’s TED Talk, “The why and how of effective altruism”, is more effective than HPMOR in encouraging effective altruism. It is also free!

a. While I agree that the books you've mentioned are more directly related to EA than HPMoR. I think it would not be possible to give them as a prize. I think the fact that the organizers whom we contacted had read HPMoR significantly contributed to the possibility to give anything at all.

b. I share your concern about HPMoR not being EA enough. We hope to mitigate it via leaflet + SPARC/ESPR.

I think this comment suggests there's a wide inferential gap here. Let me see if I can help bridge it a little.

If the goal is to teach Math Olympiad winners important reasoning skills, then I question this goal. They just won the Math Olympiad. If any group of people already had well developed logic and reasoning skills, it would be them. I don’t doubt that they already have a strong grasp of Bayes’ rule.

I feel fairly strongly that this goal is still important. I think that the most valuable resource that the EA/rationality/LTF community has is the ability to think clearly about important questions. Nick Bostrom advises politicians, tech billionaires, and the founders of the leading AI companies, and it's not because he has the reasoning skills of a typical math olympiad. There are many levels of skill, and Nick Bostrom's is much higher[1].

It seems to me that these higher level skills are not easily taught, even to the brightest minds. Notice how society's massive increase in the number of scientists has failed to produce anything like linearly more deep insights. I have seen this for myself at Oxford University, where many of my fellow students could compute ve... (read more)

Thanks for your long critique! I will try to respond to as much of it as I can.

As I see it, there are four separate claims in your comment, each of which warrants a separate response:

1. The Long-Term Future Fund should make all of its giving based on a high standard of externally transparent evidence

2. Receiving HPMoRs is unlikely to cause the math olympiad participants to start working on the long-term future, or engage with the existing EA community

3. EA Russia has made an oral promise of delivering HPMoRs without having secured external funding first

4. If the Long-Term Future Fund is making grants that are this risky, they should not be advertised as the go-to vehicle for donations

I will start responding to some of them now, but please let me know if the above summary of your claims seems wrong.

I don't think that 2) really captures the objection the way I read it. It seems that on margin, there are much more cost effective ways of engaging math olympiad participants, and that the content distributed could be much more directly EA/AI related at lower cost than distributing 2000 pages of hard copy HPMoR.

I don't think anyone should be trying to persuade IMO participants to join the EA community, and I also don't think giving them "much more directly EA content" is a good idea.

I would prefer Math Olympiad winners to think about long-term, think better, and think independently, than to "join the EA community". HPMoR seems ok because it is not a book trying to convince you to join a community, but mostly a book about ways how to think, and a good read.

(If they readers eventually become EAs after reasoning independently, it's likely good; if they for example come to the conclusion there are mayor flaws in EA and it's better to engage with the movement critically, it's also good.)

Agree with this. I do think there is value in showing them that there exists a community that cares a lot about the long-term-future, and do think there is some value in them collaborating with that community instead of going off and doing their own thing, but the first priority should be to help them think better and about the long-term at all. I think none of the other proposed books achieve this very well.

Hello, first of all, thank you for engaging with my critique. I have some clarifications for your summary of my claims.

  1. Ideally, yes. If there is a lack of externally transparent evidence, there should be strong reasoning in favor of the grant.

  2. I think that there is no evidence that using $28k to purchase copies of HPMOR is the most cost-effective way to encourage Math Olympiad participants to work on the long-term future or engage with the existing community. I don't make the claim that it won't be effective at all. Simply that there is little reason to believe it will be more effective, either in an absolute sense or in a cost-effectiveness sense, than other resources.

  3. I'm not sure about this, but this was the impression the forum post gave me. If this is not the case, then, as I said, this grant displaces some other $28k in funding. What will that other $28k go to?

  4. Not necessarily that risky funds shouldn't be recommended as go-to, although that would be one way of resolving the issue. My main problem is that it is not abundantly clear that the Funds often make risky grants, so there is a lack of transparency for an EA newcomer. And while this particularly applies to the Long Term fund, given it is harder to have evidence concerning the Long Term, it does apply to all the other funds.

Sorry for the delay, others seem to have given a lot of good responses in the meantime, but here is my current summary of those concerns:

1. Ideally, yes. If there is a lack of externally transparent evidence, there should be strong reasoning in favor of the grant.

By word-count the HPMOR writeup is (I think) among the three longest writeups that I produced for this round of grant proposals. I think my reasoning is sufficiently strong, though it is obviously difficult for me to comprehensively explain all of my background models and reasoning in a way that allows you to verify that.

The core arguments that I provided in the writeup above seem sufficiently strong to me, not necessarily to convince a completely independent observer, but I think for someone with context about community building and general work done on the long-term future, I expect it to successfully communicate the actual reasons for why I think the grant is a good idea.

I generally think grantmakers should give grants to whatever interventions they think are likely to be most effective, while not constraining themselves to only account for evidence that is easily communicable to other people. They then should also i... (read more)

Thanks for the response. I don’t have the time to draft a reply this week but I’ll get back to you next week.

A small correction:

Facilitating conversations between top people in AI alignment (I’ve in particular heard very good things about the 3-day conversation between Eric Drexler and Scott Garrabrant that Eli facilitated)

I do indeed facilitate conversations between high level people in AI alignment. I have a standing offer to help with difficult conversations / intractable disagreements, between people working on x-risk or other EA causes.

(I'm aiming to develop methods for resolving the most intractable disagreements in the space. The more direct experience I have trying my existing methods against hard, "real" conversations, the faster that development process can go. So, at least for the moment, it actively helps me when people request my facilitation. And also, a number of people, including Eric and Scott, have found it to be helpful for the immediate conversation.)

However, I co-facilitated that particular conversation between Eric and Scott. The other facilitators were, Eliana Lorch, Anna Salamon, and Owen Cotton Barratt.

Will update to say "help facilitate". Thanks for the correction!
Is there any resource (eg blogpost) for people curious about what "facilitating conversations" involves?

At the moment, not really.

There's the classic Double Crux post. Also, here's a post I wrote, that touches on one sub-skill (out of something like 50 to 70 sub-skills that I currently know). Maybe it helps give the flavor.

If I were to say what I'm trying to do in a sentence: "Help the participants actually understand eachother." Most people generally underestimate how hard this is, which is a large part of the problem.

The good thing that I'm aiming for in a conversation is when "that absurd / confused thing that X-person was saying, clicks into place, and it doesn't just seem reasonable, it seems like a natural way to think about the situation".

Another frame is, "Everything you need to do to make Double Crux actually work."

A quick list of things conversational facilitation, as I do it, involves:

  • Tracking the state of mind of the participants. Tracking what's at stake for each person.
  • Noticing when Double Illusion of Transparency, or talking past eachother, is happening, and having the participants paraphrase or operationalize. Or in the harder cases, getting each view myself, and then acting as an intermediary.
  • Identifying
... (read more)
Yes, that helps, thanks. "Mediating" might be a word which would convey the idea better.
[Are there ways to delete a comment? I started to write a comment here, and then added a bit to the top-level instead. Now I can't make this comment go away?]

Feedback that I sent to Jeffrey Ladish about his application:

Excerpts from the application

I would like to spend five months conducting a feasibility analysis for a new project that has the potential to be built into an organization. The goal of the project would be to increase civilizational resilience to collapse in the event of a major catastrophe -- that is, to preserve essential knowledge, skills, and social technology necessary for functional human civilization.

The concrete results of this work would include an argument for why or why not a project aimed at rebuilding after collapse would be feasible, and at what scale.

Several scholars and EAs have investigated this question before, so I plan to build off existing work to avoid reinventing the wheel. In particular, [Beckstead 2014](https://www.fhi.ox.ac.uk/wp-content/uploads/1-s2.0-S0016328714001888-main.pdf) investigates whether bunkers or shelters might help civilization recover from a major catastrophe. He enumerates many scenarios in which shelters would *not* be helpful, but concludes with two scenarios worthy of deeper analysis: “global food crisis” and “social collapse”. I plan to focus on “social collapse”, noting tha... (read more)

Jeffrey Ladish
I want to give a brief update on this topic. I spent a couple months researching civilizational collapse scenarios and come to some tentative conclusions. At some point I may write a longer post on this, but I think some of my other upcoming posts will address some of my reasoning here. My conclusion after investigating potential collapse scenarios: 1) There are a number of plausible (>1% probability) scenarios in the next hundred years that would result in a "civilizational collapse", where an unprecedented number of people die and key technologies are (temporarily) lost. 2) Most of these collapse scenarios would be temporary, with complete recovery likely on the scale of decades to a couple hundred years. 3) The highest leverage point for intervention in a potential post-collapse environment would be at the state level. Individuals, even wealthy individuals, lack the infrastructure and human resources at the scale necessary to rebuild effectively. There are some decent mitigations possible in the space of information archival, such as seed banks and internet archives, but these are far less likely to have long term impacts compared to state efforts. Based on these conclusions, I decided to focus my efforts on other global risk analysis areas, because I felt I didn't have the relevant skills or resources to embark on a state-level project. If I did have those skills & resources, I believe (low to medium confidence) it would be worthwhile project, and if I found a person or group who did possess those skills / resources, I would strongly consider offering my assistance.
Josh Jacobson
Are you saying here that you believe the scenarios add up to a greater than 1% probability of collapse in the next hundred years, or that you believe there are multiple scenarios that each have greater than 1% probability?
Jeffrey Ladish
Some quick answers to your questions based on my current beliefs: * Is there a high chance that human population completely collapses as a result of less than 90% of the population being wiped out in a global catastrophe? I think the answer in the short term is no, if "completely collapses" means something like "is unable to get back to at least 1950's level technology in 500 years". I think think there are a number of things that could reduce humanity's "technological carrying capacity". I'm currently working on explicating some of these factors, but some examples would be drastic climate change, long-lived radionuclides, increase in persistent pathogens. * Can we build any reasonable models about what our bottlenecks will be for recovery after a significant global catastrophe? (This is likely dependent on an analysis of what specific catastrophes are most likely and what state they leave humanity in) I think we can. I'm not sure we can get very confident about exactly which potential bottlenecks will prove most significant, but I think we can narrow the search space and put forth some good hypotheses, both by reasoning from the best reference class examples we have and by thinking through the economics of potential scenarios. * Are there major risks that have a chance to wipe out more than 90% of the population, but not all of it? My models of biorisk suggests it's quite hard to get to 90% mortality, I think most nuclear winter scenarios also have less than a 90% food reduction impact I'm not sure about this one. I can think of some scenarios that would wipe out 90%+ of the population but none of them seem very likely. Engineered pandemics seem like one candidate (I agree with Denkenberger here), and the worst-case nuclear winter scenarios might also do it, though I haven't read the nuclear winter papers in a while, and there has been several new papers and comments in the last year, including real disagreement in the field (yay, finally!) * Are there n
You say no to "Is there a high chance that human population completely collapses as a result of less than 90% of the population being wiped out in a global catastrophe?" and say "2) Most of these collapse scenarios would be temporary, with complete recovery likely on the scale of decades to a couple hundred years." I feel like I'd much better understand what you mean if you were up for giving some probabilities here even if there's a range or they're imprecise or unstable. There's a really big range within "likely" and I'd like some sense of where you are on that range.
This is very helpful to see your reasoning and cruxes. I reply to the ALLFED related issues above, but I thought I would reply to the pandemic issue here. Here is one mechanism that could result in greater than 90% mortality from a pandemic: multiple diseases at the same time: multipandemic.

This is the (very slightly edited) feedback that I sent to GCRI based on their application (caveat that GCR-policy is not my expertise and I only had relatively weak opinions in the discussion around this grant, so this should definitely not be seen as representative of the broader opinion of the fund):

I was actually quite positive on this grant, so the primary commentary I can provide is a summary of what would have been sufficient to move me to be very excited about the grant.
Overall, I have to say that I was quite positively surprised after reading a bunch of GCRI's papers, which I had not done before (in particular the paper that lists and analyzes all the nuclear weapon close-calls).
I think the biggest thing that made me hesitant about strongly recommending GCRI, is that I don't have a great model of who GCRI is trying to reach. I am broadly not super excited about reaching out to policy makers at this stage of the GCR community's strategic understanding, and am confused enough about policy capacity-building that I feel uncomfortable making strong recommendations based on my models there. I do have some models of capacity-building that suggest some concrete act
... (read more)

Oliver Habryka's comments raise some important issues, concerns, and ideas for future directions. I elaborate on these below. First, I would like to express my appreciation for his writing these comments and making them available for public discussion. Doing this on top of the reviews themselves strikes me as quite a lot of work, but also very valuable for advancing grant-making and activity on the long-term future.

My understanding of Oliver's comments is that while he found GCRI's research to be of a high intellectual quality, he did not have confidence that the research is having sufficient positive impact. There seem to be four issues at play: GCRI’s audience, the value of policy outreach on global catastrophic risk (GCR), the review of proposals on unfamiliar topics, and the extent to which GCRI’s research addresses fundamental issues in GCR.

(1) GCRI’s audience

I would certainly agree that it is important for research to have a positive impact on the issues at hand and not just be an intellectual exercise. To have an impact, it needs an audience.

Oliver's stated impression is that GCRI's audience is primarily policy makers, and not the EA long-term future... (read more)

I want to make sure that there isn't any confusion about this: When I do a grant writeup like the one above, I am definitely only intending to summarize where I am personally coming from. The LTF-Fund had 5 voting members last round (and will have 4 in the coming rounds), and so my assessment is necessarily only a fraction of the total assessment of the fund. I don't currently know whether the question of the target audience would have been super valuable for the other fund members, and given that I already gave a positive recommendation, their cruxes and uncertainties would have actually been more important to address than my own.

On the question of whether we should have an iterative process: I do view this publishing of the LTF-responses as part of an iterative process. Given that we are planning to review applications every few months, you responding to what I wrote allows us to update on your responses for next round, which will be relatively soon.

That makes sense. I might suggest making this clear to other applicants. It was not obvious to me.
Thanks, this is good to know.
(Breaking things up into multiple replies, to make things easier to follow, vote on, and reply to) Of those, I had read "Long-term trajectories of human civilization" and "The far future argument for confronting catastrophic threats to humanity: Practical significance and alternatives" before I made my recommendation (which I want to clarify was a broadly positive recommendation, just not a very-positive recommendation). I actually had a sense that these broad overviews were significantly less valuable to me than some of the other GCRI papers that I've read and I predict that other people who have thought about global catastrophic risks for a while would feel the same. I had a sense that they were mostly retreading and summarizing old ground, while being more difficult to read and of lower quality than most of the writing that already exists on this topic (a lot of it published by FHI, and a lot of it written on LessWrong and the EA Forum). I also generally found the arguments in them not particularly compelling (in particular I found the arguments in "The far future argument for confronting catastrophic threats to humanity: Practical significance and alternatives" relatively weak, and thought that it failed to really make a case for significant convergent benefits of long-term and short-term concerns. The argument seemed to mostly consists of a few concrete examples, most of which seemed relatively tenuous to me. Happy to go into more depth on that). I highlighted the "A model for the probability of nuclear war" not because it was the only paper I read (I read about 6 GCRI papers when doing the review and two more since then), but because it was the paper that did actually feel to me like it was helping me build a better model of the world, and something that I expect to be a valuable reference for quite a while. I actually don't think that applies to any of the three papers you linked above. I don't currently have a great operationalization of what I mean by
That is interesting to hear. Some aspects of the overviews are of course going to be more familiar to domain experts. The integrated assessment paper in particular describes an agenda and is not intended to have much in the way of original conclusions. I would be quite interested in further thoughts you have on this. I’ve actually found that the central ideas of the far future argument paper have held up quite well, possibly even better than I had originally expected. Ditto for the primary follow-up to this paper, “Reconciliation between factions focused on near-term and long-term artificial intelligence”, which is a deeper dive on this theme in the context of AI. Some examples of work that is in this spirit: · Open Philanthropy Project’s grant for the new Georgetown CSET group, which pursues “opportunities to inform current and future policies that could affect long-term outcomes” (link) · The study The Malicious Use of Artificial Intelligence, which, despite being led by FHI and CSER, is focused on near-term and sub-existential risks from AI · The paper Bridging near- and long-term concerns about AI by Stephen Cave and Seán S. ÓhÉigeartaigh of CSER/CFI All of these are more recent than the GCRI papers, though I don’t actually know how influential GCRI’s work was in any of the above. The Cave and ÓhÉigeartaigh paper is the only one that cites our work, and I know that some other people have independently reached the same conclusion about synergies between near-term and long-term AI. Even if GCRI’s work was not causative in these cases, these data points show that the underlying ideas have wider currency, and that GCRI may have been (probably was?) ahead of the curve. That’s fine, but note that those organizations have much larger budgets than GCRI. Of them, GCRI has closest ties to FHI. Indeed, two FHI researchers were co-authors on the long-term trajectories paper. Also, if GCRI was to be funded specifically for research to improve the decision-making of peo
Just wanted to make a quick note that I also felt the "overview" style posts aren't very useful to me (since they mostly encapsulate things I already had thought about) At some point I was researching some aspects of nuclear war, and reading up on a GCRI paper that was relevant, and what I found myself really wishing was that the paper had just drilled deep into whatever object level, empirical data was available, rather than being a high level summary.
Thanks, that makes sense. This is one aspect in which audience is an important factor. Our two recent nuclear war model papers (on the probability and impacts) were written to be accessible to wider audiences, including audiences less familiar with risk analysis. This is of course a factor for all research groups that work on topics of interest to multiple audiences, not just GCRI.
Thanks for posting the response! Some short clarifications: My perspective only played a partial role in the discussion of the GCRI grant, since I am indeed not the person with the most policy expertise on the fund. It only so happens that I am also the person who had the most resources available for writing things up for public consumption, so I wouldn't update too much on my specific feedback. Though my perspective might still be useful for understanding the experience of people closer to my level of expertise, of which there are many, and I do obviously think there is important truth to it (and obviously as a way to help me build better models of the policy space, which I do think is valuable). I strongly agree with this, and also think that a lot of the best work is cross-cutting and interdisciplinary. I think the degree to which things are interdisciplinary is part of the reason for why there is some shortage for EA grantmaking expertize. Part of my hope with facilitating public discussion like this is to help me and other people in grantmaking positions build better models of domains where we have less expertize.
All good to know, thanks. I'll briefly note that I am currently working on a more extended discussion of policy outreach suitable for posting online, possibly on this site, that is oriented toward improving the understanding of people in the EA-LTF-GCR community. It's not certain I'll have the chance to complete given my other responsibilities it but hopefully I will. Also if it would help I can provide suggestions of people at other organizations who can give perspectives on various aspects of GCRI's work. We could follow up privately about that.

Question: How funding constrained do you feel like the Long-Term Future Fund is? Do you feel like you get to make essentially every grant you think you'd reasonably want to make or are there more awesome grants you would've made if only the fund had raised more money?

Stefan Torges from REG recently asked me for our room for funding, and I sent him the following response back: The value of these marginal grants doesn't feel like it would go down more than 20% than our current worst grants, since in every round I feel like there is a large number of grants that are highly competitive with the lowest-ranked grants we do make. In other words, I think we have significant room for funding at about the quality level of grants we are currently making.