Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Why AGI Timeline Research/Discourse Might Be Overrated, published by Miles Brundage on July 3, 2022 on The Effective Altruism Forum.

TL;DR: Research and discourse on AGI timelines aren't as helpful as they may at first appear, and a lot of the low-hanging fruit (i.e. motivating AGI-this-century as a serious possibility) has already been plucked.

Introduction

A very common subject of discussion among EAs is “AGI timelines.” Roughly, AGI timelines, as a research or discussion topic, refer to the time that it will take before very general AI systems meriting the moniker “AGI” are built, deployed, etc. (one could flesh this definition out and poke at it in various ways, but I don’t think the details matter much for my thesis here—see “What this post isn’t about” below). After giving some context and scoping, I argue below that while important in absolute terms, improving the quality of AGI timelines isn’t as useful as it may first appear.

Just in the past few months, a lot of digital ink has been spilled, and countless in-person conversations have occurred, about whether recent developments in AI (e.g. DALL-E 2.0, Imagen, PALM, Minerva) suggest a need for updating one’s AGI timelines to be shorter. Interest in timelines has informed a lot of investment in surveys, research on variables which may be correlated with timelines like compute, etc. At least dozens of smart-person-years have been spent on this question; possibly the number is more like hundreds or thousands.

AGI timelines are, at least a priori, very important to reduce uncertainty about, to the extent that’s possible. Whether one’s timelines are “long” or “short” could be relevant to how one makes career investments—e.g. “exploiting” by trying to maximize influence over AI outcomes in the near-term, or “exploring” by building up skills that can be leveraged later. Timelines could also be relevant to what kinds of alignment research directions are useful, and which policy levers to consider (e.g. whether a plan that may take decades to pan out is worth seriously thinking about, or whether the “ship will have sailed” before then).

I buy those arguments to an extent, and indeed I have spent some time myself working on this topic. I’ve written or co-authored various papers and blog posts related to AI progress and its conceptualization/measurement, I’ve contributed to papers and reports that explicitly made forecasts about what capabilities were plausible on a given time horizon, and I have participated in numerous surveys/scenario exercises/workshops/conferences etc. where timelines loomed large. And being confused/intrigued by people’s widely varying timelines is part of how I first got involved in AI, so it has a special place in my heart. I’ll kcertainly eep doing some things related to timelines myself, and think some others with special knowledge and skills should also continue to do so.

But I think that, as with many research and discussion topics, there are diminishing returns on trying to understand AGI timelines better and talking widely about them. A lot of the low-hanging fruit from researching timelines has already been plucked, and even much higher levels of certainty on this question (if that were possible) wouldn’t have all the benefits that might naively be suspected.

I’m not sure exactly how much is currently being invested in timeline research, so I am deliberately vague here as to how big of a correction, if any, is actually needed compared to the current level of investment. As a result of feedback on this post, I may find out that there’s actually less work on this than I thought, that some of my arguments are weaker than I thought, etc. and update my views. But currently, while I think timelines should be valued very highly compared to a random research topic, I suspect that many reading thi...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: There will be many more effective altruist billionaires, published by Dwarkesh Patel on July 3, 2022 on The Effective Altruism Forum.

Cross posted from my blog.

Because of the generosity of a few billionaires, the effective altruism movement has recently come into a lot of money. The total amount of capital committed to the movement varies day to day with the crypto markets on which Sam Bankman-Fried’s net worth is based. But the sum was recently estimated at 46 billion1.

The movement has been trying to figure out how quickly it should give away this money. There’s lots of fascinating questions you have to resolve before you can decide on a disbursement schedule2. An especially interesting one is: how many future billionaires will be effective altruists? If a new Sam Bankman-Fried and Dustin Moskovitz will join the movement every few years, then the argument for allocating money now becomes much more compelling. Future billionaires can handle future problems, but only you can fund the causes that are neglected and important today.

Here are the three reasons I expect the number of EA billionaires to grow significantly.

Effective altruism allows thymodic natures to achieve recognition and impact that is otherwise unavailable in the modern world.

Effective altruism acts as a Schelling point for ambitious and risk taking founders.

Effective altruism creates alignment in an organization and reduces adverse selection.

Thymos

In The End of History and the Last Man, Fukuyama argues that the leading contender for the final form of government is capitalist liberal democracy. Capitalism is peerless in satisfying people’s desires and democracy is so far the best method of affording them recognition. His greatest hesitation about the sustainability of liberal democracies is whether societies where everyone has comfortable lives and no one gets special recognition can appease the appetites of the most ambitious personalities. As he puts it:

[T]he virtues and ambitions called forth by war are unlikely to find expression in liberal democracies. There will be plenty of metaphorical wars—corporate lawyers specializing in hostile takeovers who will think of themselves as sharks or gunslingers, and bond traders who imagine . that they are “masters of the universe.” . But as they sink into the soft leather of their BMWs, they will know somewhere in the back of their minds that there have been real gunslingers and masters in the world, who would feel contempt for the petty virtues required to become rich or famous in modern America. How long megalothymia will be satisfied with metaphorical wars and symbolic victories is an open question.

You can always have a great late night conversation by asking, “What would Napoleon or Ceasar do if he was born in modern America?” Surely the unique combinations of genes which make up the will and capacities of such men have not disappeared. But today their ambition cannot be exercised through great conquests and wars. So how is their energy redirected?

The first place to look for modern Caesars or Alexanders is Silicon Valley. Napoleon’s Grande Armée was basically run like a startup - extremely efficient and flexible, with quick promotion and delegation given to the most able and even quicker terminations provided to the least. Say what you will, they certainly knew how to capture bigger markets. When Napoleon told the Austrian statesman Metternich, “You cannot stop me, I can spend 30,000 men a month,” he was simply expressing the principle of blitzscaling. It would be much as if the Uber CEO told the Lyft CEO, “You can’t catch up to us, I’m willing to burn 1 billion dollars of VC cash a quarter.”

There are only a few startups whose mission is so intrinsically motivating that it would satisfy the most ambitious individuals of history. Sure, building a ...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Trigger-Action Planning, published by CFAR!Duncan on July 3, 2022 on LessWrong.

Epistemic status: Established and confirmed

There has been a tremendous amount of research on “implementation intentions” since their development by psychologist Peter Gollwitzer in the late 1990’s. A meta- analysis of 94 studies involving 8461 participants found that interventions using implementation intentions were an average of .65 standard deviations more effective than control interventions. Similar effect sizes were found in the 34 studies which looked at behavioral change on personal or health goals (average of .59 standard deviations more effective). Trigger-action planning—our version of implementation intentions—draws directly on this research and has proven useful to the majority of our alumni for a wide range of problems, tasks, and goals.

In previous sections of this book, we’ve looked at the differences between System 1 and System 2, talked about the process of turning goals into plans, and learned to distinguish useful and relevant practice from irrelevant or unproductive practice. In this section, we will combine those insights and their implications into a single, robust technique for building awareness and supporting behavioral change.

Complex chains: The parable of the Sphex

Sphexes are a genus of wasps, and for many years, a story about their behavior has been a major touchstone in cognitive science. Typically, when it comes time for egg laying, a sphex will build a burrow and fill it with paralyzed insects for her future larvae to eat. When hunting, she will sting her prey, wait for the venom to take effect, drag the prey back to the burrow entrance, leave it outside while she goes in and reconnoiters (presumably confirming the absence of predators or structural problems), and finally come back out to drag her victim inside.

This sequence of actions is elaborate, organized, and complex, and on the surface seems to indicate an impressive level of mental sophistication for an insect whose brain weighs less than a milligram. However, in 1879, French entomologist Jean Henri Fabre decided to dig deeper:

I will mention an experiment...at the moment when the Sphex is making her domiciliary visit, I take the cricket left at the entrance to the dwelling and place her a few inches farther away. The Sphex comes up, utters her usual cry, and...comes out of her hole to seize it and bring it back to its right place. Having done this, she goes down again, but alone [once more leaving the cricket outside]. I play the same trick upon her, and the Sphex has the same disappointment on her return to the surface. The victim is once more dragged back to the edge of the hole, but the wasp always goes down alone. . . forty times over, did I repeat the same experiment on the same wasp; her persistence vanquished mine and her tactics never varied.

Fabre’s own experiments on other wasps (from the same colony, from the same species but other colonies, and from other species) showed that this was not the only possible result—many wasps eventually break the pattern and drag their prey straight into the burrow. But even the quickest tend to repeat themselves four or five times, implying that the overall process is less a single, coherent strategy and more a series of disconnected if-then actions:

By “chaining together” a series of simple, atomic responses (e.g if I come out of my burrow and there’s a paralyzed cricket, drag it inside immediately), the sphex is able to execute complex, multi-step behaviors as if it were capable of thinking and planning ahead—even though it largely isn’t. The “intelligence” lies in the algorithm, rather than in active cognition.

The trigger-action pattern

There is another species that is capable of chaining together a series of atomic reflex actions into complex and...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: The Tree of Life: Stanford AI Alignment Theory of Change, published by Gabriel Mukobi on July 2, 2022 on The Effective Altruism Forum.

Crossposted on Less Wrong.

0. Preface

Summary: AI alignment needs many great people working on research, engineering, governance, social science, security, operations, and more. I think AI alignment-focused university groups are uniquely underexploited sources of that future talent. Here are my plans for running such a group at Stanford, broadly involving mass outreach to a wide range of students, efficient fit-testing and funneling to a core group of promising students, and robust pipelining of those students into a connected network of impactful opportunities.

Confidence: Moderate. This theory of change is heavily inspired by similar plans from Michael Chen (Effective Altruism at Georgia Tech) and Alexander Davies (Harvard AI Safety Team) as well as EA field-building learnings from Stanford Effective Altruism. While parts of this have been tested in those communities, parts are new or different. That said, I’m fairly confident this is at least a significant improvement over our current local AI alignment field-building operations. The purpose of posting this is both to refine these plans and to share ideas that might be useful for groups with similar goals.

Terminology: I use the term “AI alignment” as an umbrella term throughout this to refer not only to technical AI alignment/existential AI safety, but also to AI governance, AI strategy, and other work that directly reduces existential risk from AI. The specific choice of this term (as opposed to e.g. “AI safety”) is to practically distinguish the long-term/existential risk focus of our new group from the near-term-leaning Stanford Center for AI Safety.

1. Introduction

1.1 Context

Stanford Effective Altruism has gradually been increasing its focus on AI alignment over the last year. We invited more AI-related guests to our open Monday dinners, ran events like an ELK weekend and a Rob Miles binge party, organized one and a half alignment-focused retreats, and ran a local AGI Safety Fundamentals Technical Alignment reading group in the spring. From these activities, we saw some promising signs, like:

Many people we’d never met before attended our Monday dinners because of cold email list messages advertising an AI alignment speaker.

Among our group of core EAs, many got more interested in AI alignment and asked for more activities like these.

However, we also found some potential issues:

Many of those students we’d never met before, suggesting our EA outreach was failing to reach and capture them (e.g. some people interested in AI alignment who attended our Monday dinners had heard about but decided not to apply to our intro EA fellowship).

For our core EAs who did get significantly interested in AI alignment, there was a big gap between “I have read and discussed AI enough to think this is the most pressing problem” and “I feel like I have the skills and opportunities to have an impact.”

It could feel at times like Stanford EA was just about AI alignment and that there wasn’t a space for other cause areas.

These initial activities showed a lot of promise for expanding and formalizing AI alignment opportunities at Stanford, but they also made us feel like we should try a different approach.

1.2 SAIA: A New Organization

One day in spring 2022, the Stanford EA organizers decided to split off a sub-group solely dedicated to AI alignment field-building. The initial idea was simple: Run most of our existing AI-focused activities under this new AI-focused group in order to keep Stanford EA more EA/cause-prioritization focused and outreach to students who could be good fits for AI alignment without the direct EA association. And thus, Stanford AI Alignment (SAIA, pronounced "sigh-uhh") was...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Naive Hypotheses on AI Alignment, published by Shos Tekofsky on July 2, 2022 on LessWrong.

Apparently doominess works for my brain, cause Eliezer Yudkowsky’s AGI Ruin: A List of Lethalities convinced me to look in to AI safety. Either I’d find out he’s wrong, and there is no problem. Or he’s right, and I need to reevaluate my life priorities.

After a month of sporadic reading, I've learned the field is considered to be in a state of preparadigmicity. In other words, we don’t know how to think about the problem yet, and thus novelty comes at a premium. The best way to generate novel ideas is to pull in people from other disciplines. In my case that's computational psychology: modeling people like agents. And I've mostly applied this to video games. My Pareto frontier is "modeling people like agents based on their behavior logs in constructed games created to trigger reward signals + ITT'ing the hell out of all the new people I love to constantly meet". I have no idea if this background makes me more or less likely to generate a new idea that's useful to solving AI alignment, but the way I understand the problem now: everyone should at least try.

So I started studying AI alignment, but quickly realized there is a trade-off: The more I learn, the harder it is to think of anything new. At first I had a lot of naive ideas on how to solve the alignment problem. As I learned more about the field, my ideas all crumbled. At the same time, I can't really assess yet if there is a useful level of novelty in my naive hypotheses. I'm still currently generating ideas low on "contamination" by existing thought (cause I'm new), but also low on quality (cause I'm new). As I learn more, I'll start generating higher quality hypotheses, but these are likely to become increasingly constrained to the existing schools of thought, because of cognitive contamination from everyone reading the same material and thinking in similar ways. Which is exactly the thing we want to avoid at this stage.

Therefore, to get the best of both worlds, I figured I'd write down my naive hypotheses as I have them, and keep studying at the same time. Maybe an ostensibly "stupid" idea on my end, inspires someone with more experience to a workable idea on their end. Even if the probability of that is <0.1%, it's still worth it. Cause, you know, .... I prefer we don't all die.

So here goes:

H1 - Emotional Empathy

If you give a human absolute power, there is a small subset of humans that actually cares and will try to make everyone’s life better according to their own wishes. This is a trait in a subset of humans. What is this trait, and can we integrate it in to the reward function of an AGI?

Does the trait rely on lack of meta-cognition? Does this trait show up equally at various IQ levels or does it peak at certain IQ levels? If the trait is less common at higher IQ levels, then this is probably a dead end. If the trait is more common at higher IQ levels, then there might be something to it.

First candidate for this trait is “emotional empathy”, a trait that hitches one’s reward system to that of another organism. Emotional empathy that we wire in to the AGI would need to be universal to all humanity, and not biased, like the human implementation.

H2 - Silo AI

Silo the hardware and functionality of AGI to particular tasks. Like governments are run in trifecta to avoid corruption. Like humans need to collaborate to make things greater than themselves. Similarly, limit AGI to functions and physicalities that force it to work together with multiple other, independent AGI’s to achieve any change in the world.

Counterargument: Silo’ed AI is effectively Tool AI, to which Gwern has written a counterargument that people won’t develop Tool AI cause it will always be worse than Agent AI.

Maybe that’s what we need to police?...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: AI safety university groups: a promising opportunity to reduce existential risk, published by mic on June 30, 2022 on The Effective Altruism Forum.

Summary: AI safety university groups are a promising way to grow the talent pool working to reduce existential risk from AI and can be fairly straightforward to set up successfully. Based on the experience of several student groups last semester, running AGI Safety Fundamentals or a similar AI safety reading group can attract dozens of participants and boost their likelihood to work on AI safety, while engaging many that wouldn’t have been reached by cause-neutral effective altruism community building. If you don’t have the time or experience to run an AI safety reading group, it can be valuable to broadly publicize a virtual program facilitated by another group (e.g., EA Cambridge).

Overall, AI safety field-building is a new, exciting area with much low-hanging fruit. At the same time, there are risks from poorly executed field-building which should be thoughtfully handled, such as inducing a poor impression of AI safety or increasing interest in accelerating the development of (potentially unsafe) artificial general intelligence (AGI).

If you’re interested in AI safety field-building, either at a university or in another setting (e.g., technical workplaces), fill out our interest form and let’s have a call to plan out next steps! Even if you think you’d do fine without a call, I’d strongly recommend taking a few seconds to fill this out, so that we can connect you with resources and other group organizers. We’d be excited to chat with not only people interested in hands-on group organizing, but also people excited to support groups remotely (perhaps by developing content).

Transformative AI might pose considerable existential risk within the coming decades or later this century. Trained to maximize an objective, AI systems often find creative and undesirable ways to do so. YouTube's recommendation algorithm, trained to suggest content to maximize user engagement, ended up recommending increasingly extreme content and proliferating misinformation. Language models such as GPT-3, trained to predict the next word in a piece of text, can be prompted to produce offensive content and leak personal information from the training data. As AI capabilities advance, we can expect the stakes of these risks from misaligned systems to dramatically increase. Aligning deep learning systems with human values might be extremely challenging, and approaches that work for making a simpler model safe may fail for more advanced models.

At the moment, leading AI companies such as Google Brain and Meta AI are actively working on developing AI with increasingly general capabilities (see A path to autonomous machine intelligence and Pathways), while not investing much in research to ensure that future systems are aligned and safe. (For a more thorough introduction to existential risk from AI, see Introduction to AI Safety - Robert Miles, Why AI alignment could be hard with modern deep learning, AI alignment - Wikipedia, and Power-seeking AI and X-risk mitigation.)

To make progress on the problem of existential risk from AI, it’s extremely valuable to have additional talented people working on it. That includes people in a variety of roles, such as machine learning safety research, conceptual alignment research, policy research, software engineering, operations, machine learning engineering, security, social science, and communications. Universities are a major source of potential talent, especially as students are actively exploring different possible career options. However, the vast majority of students have not heard about AI safety, even among students interested in AI, and fewer still have considered how they could work on AI safety.

Despite t...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Strategic Perspectives on Long-term AI Governance: Introduction, published by MMMaas on July 2, 2022 on The Effective Altruism Forum.

This is the introduction of an in-progress report called Strategic Perspectives on Long-term AI Governance (see sequence).

Aim: to contribute to greater strategic clarity, consensus, and coherence in the Long-term AI Governance field, by disentangling current and possible approaches, and providing insight into their assumptions and cruxes. Practically, the aim is for this to enable a more informed discussion about what near-term actions we can or should pursue to ensure the beneficial development and impact of Transformative AI in the long-term.

Epistemic status: attempted theory-building, with all the strengths and pitfalls of map-drawing . Draws on years of research in the space and >2 year of focused conversations on this topic, but views are tentative, presented in order to enable further discussion.

Status: This is a project that I will be producing over the coming months. The breadth of topics and perspectives covered means that some sections are hastily sketched: as I continue developing the report, I welcome feedback and commentary. I am grateful to the many people who have already given feedback and encouragement so far.

Primer

Tl;dr: look at this Airtable.

The Long-term AI Governance community aims to shape the development and deployment of advanced AI--whether understood as Transformative AI (TAI) or as Artificial General Intelligence (AGI)--in beneficial ways.

However, there is currently a lack of strategic clarity, with disagreement over relevant background assumptions; what actions to take in the near-term, the strengths and risks of each approach, and where different approaches might strengthen or trade off against one another.

This sequence will explore 15 different Strategic Perspectives on long-term AI governance, exploring their distinct:

assumptions about key strategic parameters that shape transformative AI, in terms of technical landscape and governance landscape;

Theory of victory and rough impact story;

internal tensions, cruxes, disagreements or tradeoffs within the perspective;

historical analogies and counter-examples;

recommended actions, including intermediate goals and concrete near-term interventions;

suitability, in terms of outside-view strengths and drawbacks.

Project Roadmap

(Links to be added as entries come out)

I introduce and sketch this sequence (you are here);

I give a background and rationale for this project, by discussing some of the challenges faced by the Long-term AI Governance field, comparing strategic clarity, consensus, and coherence, and mapping how other work has contributed to questions around strategic clarity.

I frame the goals and scope of this project--what it does, and does not aim to do;

I sketch the components of a strategic perspective (i.e. the space of arguments or cruxes along which perspectives can differ or disagree);

I set out 15 distinct Strategic Perspectives in Long-term AI Governance, clustering different views on the challenges of AI governance (see below);

I reflect on the distribution of these perspectives in the existing Long-term AI Governance community; and discuss four approaches to navigating between them;

In four Appendices, I cover:

(1). Definitions of key terms around advanced AI, AI governance, long-term(ist) AI governance, theories of impact, and strategic perspectives;

(2). A review of existing work in long-term AI governance, and how this project relates to it;

(3). scope, limits and risks of the project;

(4). a taxonomy of relevant strategic parameters in the technical and governance landscapes around TAI, and possible assumptions across them.

Summary of perspectives

Perspective

In (oversimplified) slogan form

Exploratory

We remain too uncertain to ...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: [Linkpost] Existential Risk Analysis in Empirical Research Papers, published by Dan Hendrycks on July 2, 2022 on The AI Alignment Forum.

I’ve noted before that it can be difficult to separate progress in safety from progress in capabilities. However, doing so is important, as we want to ensure that we are making differential progress on safety, rather than just advancing safety as a consequence of advancing capabilities. In particular, I think that research should rigorously evaluate trade-offs between improving safety and advancing capabilities.

This paper introduces the concept of “capabilities externalities,” and provides concrete steps for how to analyze the risk of different research papers. It also introduces “x-risk sheets,” which can be added to papers and help researchers discuss their impact on AI x-risk. Since discussion of AGI is becoming increasingly normalized in mainstream ML, I think careful discussions of x-risk are increasingly feasible within mainstream ML. The paper is largely aimed at mainstream ML researchers, and it is one step towards normalizing the discussion of existential risks.

Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: AXRP Episode 16 - Preparing for Debate AI with Geoffrey Irving, published by DanielFilan on July 1, 2022 on The AI Alignment Forum.

Google Podcasts link

Many people in the AI alignment space have heard of AI safety via debate - check out episode 6 of AXRP if you need a primer. But how do we get language models to the stage where they can usefully implement debate? In this episode, I talk to Geoffrey Irving about the role of language models in AI safety, as well as three projects he’s done that get us closer to making debate happen: using language models to find flaws in themselves, getting language models to back up claims they make with citations, and figuring out how uncertain language models should be about the quality of various answers.

Topics we discuss:

Status update on AI safety via debate

Language models and AI safety

Red teaming language models with language models

GopherCite

Uncertainty estimation for language reward models

Following Geoffrey’s work, and working with him

Daniel Filan: Hello, everybody. Today I’ll be speaking with Geoffrey Irving. Geoffrey is a safety researcher at DeepMind who leads the scalable alignment team. We’ll be speaking about three papers he’s co-authored: Red Teaming Language Models with Language Models, whose first author is Ethan Perez; Teaching Language Models to Support Answers with Verified Quotes, aka the GopherCite paper, whose first authors are Jacob Menick, Maja Trebacz and Vladimir Mikulik; and Uncertainty Estimation for Language Reward Models, whose first author is Adam Gleave. For links to what we’re discussing, you can check the description of this episode and you can read the transcript at axrp.net. Welcome to the show, Geoffrey.

Geoffrey Irving: Thank you.

Status update on AI safety via debate

Daniel Filan: So I guess my first question is, what happened to AI Safety via Debate?

Geoffrey Irving: It is the most important question. So I think the thing that happened is I’m still doing that stuff. I’m building up towards it. And the overall research agenda here is figure out and implement the protocol for humans and machines discussing problems that gives us good answers that we endorse after reflection. So that’s debate, but then broadened: there’s a lot of different versions of debate. There’s different things you can add onto it. You could do debate plus evidence or debate plus more aggressive red teaming. There’s a large space of protocols there.

Geoffrey Irving: And these papers are kind of pieces of that story, but we were also kind of working on just pushing on the main thing. I think that will take some time to put together because there’s still pieces to build up. But that is still very much the agenda, or a piece of the larger agenda. An example of a thing which I’ve updated on since then is the human needs to talk. The debate paper only had two machines talking, and then at the end a human judges. And that’s just not the right thing to do, obviously, because humans need to ask clarifying questions, say what they currently think in case they’ve misunderstood what’s been said so far.

Geoffrey Irving: And so there’s a broad space of different interaction protocols we’d like to explore that fix the holes in debate. And this is kind of building up towards that. One reason for doing GopherCite is if you want to do general debate, again, taking into account a bunch of leaf evidence. So debate is about machines and humans discussing some argument for something being true. But at the ends of that tree are leaves where you have checkable facts of some form. And directly, you can just have facts that every human knows, but that just isn’t that large a space of facts. And so this is extending the space of leaves so that we can implement more practical versions of debate on more interesting tasks. So I think these do fit...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Monthly Overload of EA - July 2022, published by DavidNash on July 1, 2022 on The Effective Altruism Forum.

Link post for 2022 July EA Updates

Top News

Lizka Vaintrob, Fin Moorhouse and Joshua Teperowski Monrad have launched an EA Criticism and Red Teaming contest with $100,000 in prizes. There has already been criticism of how this competition has been announced

The Future Fund with an update on their recent grants

Julia Wise - Power dynamics between people in EA

Abraham Rowe - Critiques of EA that I want to read

Ben Garfinkel - On Deference and Yudkowsky's AI Risk Estimates

Holden Karnofsky - AI Could Defeat All Of Us Combined

James Ozden & Neil Dullaghan with a list of megaproject ideas for animals

Nuño Sempere - A Critical Review of Open Philanthropy’s Bet On Criminal Justice Reform

Kurzgesagt's recent video invites viewers to consider the potential for a vast number of future people, it already has over 2 million views

Events

8th-10th July - EAGxAustralia

29th-31st July - EA Global: San Francisco

2nd-4th September - EAGxSingapore (Applications now open)

23rd-25th September - EA Global: Washington, D.C.

16th-18th September - EAGxBerlin

21st-23rd October - EAGx Virtual

Meta Updates

Open Philanthropy with a post looking at how accurate their own predictions are

Luke Muehlhauser - Effective altruism as I see it

Owen Cotton-Barratt - Global health is important for the epistemic foundations of EA, even for longtermists

A post looking for 'Examples of someone admitting an error or changing a key conclusion'

Ozy Brennan on the benefits of categorising people who are more or less dedicated in EA

80,000 Hours would be happy to see more projects in the careers space

Rémi T with the idea of taking an 'EA Break'

Jan Kulveit - Ways money can make things worse

Ofer & Owen Cotton-Barratt - Impact markets may incentivize predictably net-negative projects

Mathias KB asking what the overhead costs of grantmaking are

A post from me on how we should see effective altruism as more about coordination & field incubation

Thomas Woodside - You Don't Need To Justify Everything

Richard Y Chappell asking what an alternative EA movement might look like

Joey Savoie - 'Why EAs should normalize using Glassdoor'

Chris Freiman - You Don’t Have to be a Moral Saint to be an Effective Altruist

A post with 'unflattering' reasons the author is attracted to EA

Effektiv Spenden: A meta opportunity in Germany & Switzerland

Derek Shiller - Fanatical EAs should support very weird projects

A post titled 'How To Prevent EA From Ever Turning Into a Cult'

New Projects

Clearer Thinking have a new regranting program for projects that could have a big positive impact on the future of the world, applications close 15th July

Brian Jabarian has launched a community for economists

An intro post to User-Friendly, an EA-aligned marketing agency

Giving What We Can have a dashboard with their growth metrics

Historea Club is a 'A peer-led exploration club for history enthusiasts in the EA community'

Condor Camp is a 10 day program for talented students in Brazil to learn about the worlds most pressing problems

Giveffektivt - a Danish regranting site for effective charities

Community building

Emma Williamson - Community Builders Spend Too Much Time Community Building

Owen Cotton Barratt - There is no separate "community building" track

Habryka - On funding, trust relationships, and scaling our community

S.E. Montgomery - Community builders should focus more on supporting friendships within their group

Patrick Gruban & Lukas Trötzmüller - Mastermind Groups: A new Peer Support Format to help EAs aim higher

Critiques

Michael Nielsen - Notes on effective altruism

Joey Savoie on deference culture in EA

Archana Ahlawat with replies to the post 'Towards Ineffective Altruism'

José Luis Ricón responding to 'Ineffective Altruism'

Tr...