Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: EA Forum Prize: Winners for May-July 2021, published by Aaron Gertler on October 20, 2021 on The Effective Altruism Forum.

CEA is pleased to announce the winners of the final EA Forum Prize! This prize covers a three-month period, May-July 2021.

In first place (for a prize of $0): “All Possible Views About Humanity's Future Are Wild,” by Holden Karnofsky.

Second place ($1275): “Seven things that surprised us in our first year working in policy,” by Jack Rafferty and Lucia Coulter.

Third place ($975): “2018-2019 Long-Term Future Fund Grantees: How did they do?” by Nuño Sempere.

Fourth place ($975): “COVID: How did we do? How can we know?,” by Ghost_of_Li_Wenliang.

Fifth place ($975): “Opinion: Digital marketing is under-utilized in EA,” by JSWinchell.

Because Holden works for Open Philanthropy, one of CEA’s main funders. In keeping with our conflict-of-interest policy, he won’t receive a monetary prize, and we’ve divided that money between the other winners.

The following users were each awarded a Comment Prize ($125):

Miranda Zhang on the U.S. patient philanthropy debate

Jsevillamol on talking to civil servants about risk management

So-Low Growth on consanguinity

Jackson Wagner, explaining a downvote

sbehmer on social welfare functions and the use of willingness-to-pay

Max Daniel on key numbers that (almost) everyone in EA should know

Akash Wasil on mental health interventions

Linchuan Zhang on existential risk and Progress Studies

Louis Dixon on learning about global risks through university courses

Fin Moorhouse on free will

See here for a list of all prize announcements and winning posts.

What is the EA Forum Prize?

Certain posts and comments exemplify the kind of content we most want to see on the EA Forum. They are well-researched and well-organized; they care about informing readers, not just persuading them.

The Prize is an incentive to create content like this. But more importantly, we see it as an opportunity to showcase excellent work as an example and inspiration to the Forum's users.

The end of the Forum Prize, and a new beginning

We’ve enjoyed giving out the Forum Prize over the last three years. The winning authors deserved their rewards and recognition; I was always especially excited when I had the chance to notify a first-time winner who’d never expected that kind of response to their work.

However, the actual impact of the Prize has been hard to determine, and our best evidence (qualitative and quantitative) finds that it doesn’t seem to be a strong incentive for most authors.

We still plan to use our budget to incentivize and reward strong writing, but what we do in the future won’t be the Forum Prize — instead, we’re considering other options, including contests for the best posts on various topics, prizes specifically aimed at first-time authors, and many more.

(If you have ideas along these lines, add a comment here — though you may also want to see the list of things we’re thinking about.)

My gratitude goes out to all of our judges, who spent many hours over months or years evaluating posts for no payment and very little glory.

About the winning posts and comments

Note: I write this section in first person based on my own thoughts, rather than by attempting to summarize the views of the other judges.

All Possible Views About Humanity's Future Are Wild

“According to me, there's a decent chance that we live at the very beginning of the tiny sliver of time during which the galaxy goes from nearly lifeless to largely populated. That out of a staggering number of persons who will ever exist, we're among the first. And that out of hundreds of billions of stars in our galaxy, ours will produce the beings that fill it.”

This post was a great introduction to its series, laying out the central idea behind Holden’s work and approaching it from many differ...

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 Planning Events Early Can Be Important, published by ThomasWoodside on October 20, 2021 on The Effective Altruism Forum.

People procrastinate. The trope (and for many, a not-so-distant memory) of university students writing papers the night they are due simply reinforces that students, especially, are prone to procrastination.

For some people, procrastination works quite well. They find they have much more motivation the day before a deadline, and some procrastinators are extremely good at managing their time (in some cases, to procrastinate as much as they can). The reason that procrastination can work, especially for students, is simple: there is a clear deadline.

As a student myself, I find that this technique fails in one common circumstance, which I think is relevant to EA group organizers: planning events. The problem is that for many events, the deadline for planning is unclear. As a result, I think people frequently use simple heuristics to try to estimate how far in advance they need to plan, and often think they should start planning later than they probably should. I have attended and helped plan many student-run events, including EA events, that I think could have gone much more smoothly had they been planned earlier. In this post, I will outline some of the reasons why planning in advance can be advantageous.

Why plan events in advance?

People are busy

Many EA events are aimed at people who are very busy. Anecdotally, the top reason college students give for not going to events is that they have too much work to do. For better or worse, they say this much less frequently about planned events.

People often don't think about work they'll do in the future; in particular, I've noticed college students frequently underestimate how much work they'll have in two weeks. In addition, when deadlines are hanging over people's heads, college students seem to overestimate how much work they have. Both of these biases likely lead to a bias against spontaneous events.

In addition, people frequently make plans with friends and other groups they're involved with. If you ask people to come to an event two days before it's supposed to happen, they may have already planned something else. There may be a cost to cancelling those plans, and people are aware of that cost.

Lastly, planning events last minute selects for people who are not busy. This might mean people with more potential to get very involved with your group, but it also could mean people who don’t like committing to things in general.

You get first-mover advantage

In another forum post, Kuhan described how a cappella groups at Stanford got so many recruits: simply by recruiting early. If you haven't planned the logistics of your event, you can't recruit people early. As a result, groups that organize further ahead of time might have more success compared with groups that organize later. I think this is really important.

You have more venue and transportation options

If you're organizing an event that needs a venue, you might find that the best ones are already booked by the time you want to book them. This might mean you end up in a suboptimal venue. If you need transportation (e.g. a rental car), you can also have availability problems that result from booking too late.

Costs can be much lower

Venues and transportation get much more expensive the later they are booked. Here's an example, for a flight from New York City to London:

As you can see, prices almost doubled in the last month leading up to the travel date.

To give another example: a few weeks ago, I purchased a round trip Amtrak train ticket from New Haven to Philadelphia for $54. I just checked the price of a ticket on the exact same trains today (one week before the travel date), and it's $222. That means booking in advance saves more than 75%; 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: Why Planning Events Early Can Be Important, published by ThomasWoodside on October 20, 2021 on The Effective Altruism Forum.

People procrastinate. The trope (and for many, a not-so-distant memory) of university students writing papers the night they are due simply reinforces that students, especially, are prone to procrastination.

For some people, procrastination works quite well. They find they have much more motivation the day before a deadline, and some procrastinators are extremely good at managing their time (in some cases, to procrastinate as much as they can). The reason that procrastination can work, especially for students, is simple: there is a clear deadline.

As a student myself, I find that this technique fails in one common circumstance, which I think is relevant to EA group organizers: planning events. The problem is that for many events, the deadline for planning is unclear. As a result, I think people frequently use simple heuristics to try to estimate how far in advance they need to plan, and often think they should start planning later than they probably should. I have attended and helped plan many student-run events, including EA events, that I think could have gone much more smoothly had they been planned earlier. In this post, I will outline some of the reasons why planning in advance can be advantageous.

Why plan events in advance?

People are busy

Many EA events are aimed at people who are very busy. Anecdotally, the top reason college students give for not going to events is that they have too much work to do. For better or worse, they say this much less frequently about planned events.

People often don't think about work they'll do in the future; in particular, I've noticed college students frequently underestimate how much work they'll have in two weeks. In addition, when deadlines are hanging over people's heads, college students seem to overestimate how much work they have. Both of these biases likely lead to a bias against spontaneous events.

In addition, people frequently make plans with friends and other groups they're involved with. If you ask people to come to an event two days before it's supposed to happen, they may have already planned something else. There may be a cost to cancelling those plans, and people are aware of that cost.

Lastly, planning events last minute selects for people who are not busy. This might mean people with more potential to get very involved with your group, but it also could mean people who don’t like committing to things in general.

You get first-mover advantage

In another forum post, Kuhan described how a cappella groups at Stanford got so many recruits: simply by recruiting early. If you haven't planned the logistics of your event, you can't recruit people early. As a result, groups that organize further ahead of time might have more success compared with groups that organize later. I think this is really important.

You have more venue and transportation options

If you're organizing an event that needs a venue, you might find that the best ones are already booked by the time you want to book them. This might mean you end up in a suboptimal venue. If you need transportation (e.g. a rental car), you can also have availability problems that result from booking too late.

Costs can be much lower

Venues and transportation get much more expensive the later they are booked. Here's an example, for a flight from New York City to London:

As you can see, prices almost doubled in the last month leading up to the travel date.

To give another example: a few weeks ago, I purchased a round trip Amtrak train ticket from New Haven to Philadelphia for $54. I just checked the price of a ticket on the exact same trains today (one week before the travel date), and it's $222. That means booking in advance saves more than 75%; 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: Book review: Lost Connections by Johann Hari, published by maia on October 20, 2021 on LessWrong.

Why this book is interesting

Well, it's about depression, which is generally interesting to LW readers. For instance, 34% of SSC readers said they were diagnosed or thought they had it in 2020 (source).

This book asserts that most of us are thinking about depression in a fundamentally wrong way, which would be very important if true. It also presents some interesting possible solutions for solving depression in one’s own life and solving depression as a social, collective-action problem.

It’s not really fully fleshed out and supported, nor is it a good self-help guide. The book provides a vague model pointing towards how we should think about depression differently, and even some specific causes to look at, but there’s clearly some big gaps in the supporting evidence, and there’s not much concrete advice that a depressed person won’t have considered before (“maybe you would be less depressed if you made friends, or had more money”). Many of the claims made in the book seem to rely on anecdote and individual studies. The author seems to be taking a journalistic approach -- find one or two experts and one or two random individuals whose stories support his narrative, describe a couple of studies -- rather than a more rigorous “but why should we actually believe this over alternative models?”

Another major issue is that it conflates “biopsychosocial model of depression” with a more generalized position of “antidepressants bad, all other treatments good,” which really doesn’t help the author’s case. More on this in “Critiques”.

Nonetheless, I found the book useful because it provides a more social and practical lens on the problem of depression, which we really don’t talk about much because it doesn’t fit into the standard “treatment of illness” narrative. And some of the early results in e.g. social prescribing sound very promising.

Summary of the book

High-level claims:

The biological model of depression is mostly wrong; instead, we should think of it as biopsychosocial, meaning that there are some biological and some social and some individual components. We should focus more on the psycho- and social- components, because right now we mostly treat it biologically.

Depression is a signal in your brain that tells you something is wrong, like pain or nausea; you should listen to it and try to fix the underlying causes, rather than trying to just treat the symptom. If you suppress the signal, you might be missing something important.

People are more depressed now than in the past because of society-wide trends towards people having more depressing lives. This requires societal solutions, not just individual solutions.

The author blames depression on “disconnection from” seven factors:

Meaningful work: Many people today feel that their work is “bullshit” or not important.

Social connection (loneliness): There are well-documented trends towards people having fewer friends, saying that they feel lonely more often in surveys, participating in fewer social events and clubs, etc., over time.

Meaningful values: Consumerism makes people focus on things that aren’t really meaningful to them, which makes them less happy.

Childhood trauma: People don’t talk about their trauma and it makes them sad. (I find this questionable as part of his overall argument, see “Critiques” for more detail.)

Status and respect: People don’t have control or status in the workplace, which makes them unhappy.

Natural world: People don’t see nature much and this makes them sad.

Hopeful/secure future: People have precarious lives, and can’t trust in the future of their jobs and lives. (I’m skeptical of this claim without evidence about how optimism about the future has changed over time vs. depression. The mai...

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: Book review: Lost Connections by Johann Hari, published by maia on October 20, 2021 on LessWrong.

Why this book is interesting

Well, it's about depression, which is generally interesting to LW readers. For instance, 34% of SSC readers said they were diagnosed or thought they had it in 2020 (source).

This book asserts that most of us are thinking about depression in a fundamentally wrong way, which would be very important if true. It also presents some interesting possible solutions for solving depression in one’s own life and solving depression as a social, collective-action problem.

It’s not really fully fleshed out and supported, nor is it a good self-help guide. The book provides a vague model pointing towards how we should think about depression differently, and even some specific causes to look at, but there’s clearly some big gaps in the supporting evidence, and there’s not much concrete advice that a depressed person won’t have considered before (“maybe you would be less depressed if you made friends, or had more money”). Many of the claims made in the book seem to rely on anecdote and individual studies. The author seems to be taking a journalistic approach -- find one or two experts and one or two random individuals whose stories support his narrative, describe a couple of studies -- rather than a more rigorous “but why should we actually believe this over alternative models?”

Another major issue is that it conflates “biopsychosocial model of depression” with a more generalized position of “antidepressants bad, all other treatments good,” which really doesn’t help the author’s case. More on this in “Critiques”.

Nonetheless, I found the book useful because it provides a more social and practical lens on the problem of depression, which we really don’t talk about much because it doesn’t fit into the standard “treatment of illness” narrative. And some of the early results in e.g. social prescribing sound very promising.

Summary of the book

High-level claims:

The biological model of depression is mostly wrong; instead, we should think of it as biopsychosocial, meaning that there are some biological and some social and some individual components. We should focus more on the psycho- and social- components, because right now we mostly treat it biologically.

Depression is a signal in your brain that tells you something is wrong, like pain or nausea; you should listen to it and try to fix the underlying causes, rather than trying to just treat the symptom. If you suppress the signal, you might be missing something important.

People are more depressed now than in the past because of society-wide trends towards people having more depressing lives. This requires societal solutions, not just individual solutions.

The author blames depression on “disconnection from” seven factors:

Meaningful work: Many people today feel that their work is “bullshit” or not important.

Social connection (loneliness): There are well-documented trends towards people having fewer friends, saying that they feel lonely more often in surveys, participating in fewer social events and clubs, etc., over time.

Meaningful values: Consumerism makes people focus on things that aren’t really meaningful to them, which makes them less happy.

Childhood trauma: People don’t talk about their trauma and it makes them sad. (I find this questionable as part of his overall argument, see “Critiques” for more detail.)

Status and respect: People don’t have control or status in the workplace, which makes them unhappy.

Natural world: People don’t see nature much and this makes them sad.

Hopeful/secure future: People have precarious lives, and can’t trust in the future of their jobs and lives. (I’m skeptical of this claim without evidence about how optimism about the future has changed over time vs. depression. The mai...

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 Emergent modularity and safety, published by Richard Ngo on October 21, 2021 on The AI Alignment Forum.

Our default expectation about large neural networks should be that we will understand them in roughly the same ways that we understand biological brains, except where we have specific reasons to think otherwise. How do we understand human brains? One crucial way we currently do so is via their modularity: the fact that different parts of the brain carry out different functions. Neuroscientists have mapped many different skills (such as language use, memory consolidation, and emotional responses) to specific brain regions. Note that this doesn’t always give us much direct insight into how the skills themselves work - but it does make follow-up research into those skills much easier. I’ll argue that, for the purposes of AGI safety, this type of understanding may also directly enable important safety techniques.

What might it look like to identify modules in a machine learning system? Some machine learning systems are composed of multiple networks trained on different objective functions - which I’ll call architectural modularity. But what I’m more interested in is emergent modularity, where a single network develops modularity after training. Emergent modularity requires that the weights of a network give rise to a modular structure, and that those modules correspond to particular functions. We can think about this both in terms of high-level structure (e.g. a large section of a neural network carrying out a broad role, analogous to the visual system in humans) or lower-level structure, involving a smaller module carrying out more specific functions. (Note that this is a weaker definition than the one defended in philosophy by Fodor and others - for instance, the sets of neurons don’t need to contain encapsulated information.)

In theory, the neurons which make up a module might be distributed in a complex way across the entire network with only tenuous links between them. But in practice, we should probably expect that if these modules exist, we will be able to identify them by looking at the structure of connections between artificial neurons, similar to how it’s done for biological neurons. The first criterion is captured in a definition proposed by Filan et al. (2021).: a network is modular to the extent that it can be partitioned into sets of neurons where each set is strongly internally connected, but only weakly connected to other sets. They measure this by pruning the networks, then using graph-clustering algorithms, and provide empirical evidence that multi-layer perceptrons are surprisingly modular.

The next question is whether those modules correspond to internal functions. Although it’s an appealing and intuitive hypothesis, the evidence for this is currently mixed. On one hand, Olah et al.’s (2020) investigations find circuits which implement human-comprehensible functions. And insofar as we expect artificial neural networks to be similar to biological neural networks, the evidence from brain lesions in humans and other animals is compelling. On the other hand, they also find evidence for polysemanticity in artificial neural networks: some neurons fire for multiple reasons, rather than having a single well-defined role.

If it does turn out to be the case that structural modules implement functional modules, though, that has important implications for safety research: if we know what types of cognition we’d like our agents to avoid, then we might be able to identify and remove the regions responsible for them. In particular, we could try to find modules responsible for goal-directed agency, or perhaps even ones which are used for deception. This seems like a much more achievable goal for interpretability research than the goal of “reading off” specific thou...

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 Emergent modularity and safety, published by Richard Ngo on October 21, 2021 on The AI Alignment Forum.

Our default expectation about large neural networks should be that we will understand them in roughly the same ways that we understand biological brains, except where we have specific reasons to think otherwise. How do we understand human brains? One crucial way we currently do so is via their modularity: the fact that different parts of the brain carry out different functions. Neuroscientists have mapped many different skills (such as language use, memory consolidation, and emotional responses) to specific brain regions. Note that this doesn’t always give us much direct insight into how the skills themselves work - but it does make follow-up research into those skills much easier. I’ll argue that, for the purposes of AGI safety, this type of understanding may also directly enable important safety techniques.

What might it look like to identify modules in a machine learning system? Some machine learning systems are composed of multiple networks trained on different objective functions - which I’ll call architectural modularity. But what I’m more interested in is emergent modularity, where a single network develops modularity after training. Emergent modularity requires that the weights of a network give rise to a modular structure, and that those modules correspond to particular functions. We can think about this both in terms of high-level structure (e.g. a large section of a neural network carrying out a broad role, analogous to the visual system in humans) or lower-level structure, involving a smaller module carrying out more specific functions. (Note that this is a weaker definition than the one defended in philosophy by Fodor and others - for instance, the sets of neurons don’t need to contain encapsulated information.)

In theory, the neurons which make up a module might be distributed in a complex way across the entire network with only tenuous links between them. But in practice, we should probably expect that if these modules exist, we will be able to identify them by looking at the structure of connections between artificial neurons, similar to how it’s done for biological neurons. The first criterion is captured in a definition proposed by Filan et al. (2021).: a network is modular to the extent that it can be partitioned into sets of neurons where each set is strongly internally connected, but only weakly connected to other sets. They measure this by pruning the networks, then using graph-clustering algorithms, and provide empirical evidence that multi-layer perceptrons are surprisingly modular.

The next question is whether those modules correspond to internal functions. Although it’s an appealing and intuitive hypothesis, the evidence for this is currently mixed. On one hand, Olah et al.’s (2020) investigations find circuits which implement human-comprehensible functions. And insofar as we expect artificial neural networks to be similar to biological neural networks, the evidence from brain lesions in humans and other animals is compelling. On the other hand, they also find evidence for polysemanticity in artificial neural networks: some neurons fire for multiple reasons, rather than having a single well-defined role.

If it does turn out to be the case that structural modules implement functional modules, though, that has important implications for safety research: if we know what types of cognition we’d like our agents to avoid, then we might be able to identify and remove the regions responsible for them. In particular, we could try to find modules responsible for goal-directed agency, or perhaps even ones which are used for deception. This seems like a much more achievable goal for interpretability research than the goal of “reading off” specific thou...

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 AGI Safety Fundamentals curriculum and application, published by richard ngo on October 20, 2021 on The Effective Altruism Forum.

Over the last year EA Cambridge has been designing and running an online program aimed at effectively introducing the field of AGI safety; the most recent cohort included around 150 participants and 25 facilitators from around the world. Dewi Erwan runs the program; I designed the curriculum, the latest version of which appears in the linked document. We expect the program to be most useful to people with technical backgrounds (e.g. maths, CS, or ML), although the curriculum is intended to be accessible for those who aren't familiar with machine learning, and participants will be put in groups with others from similar backgrounds. If you're interested in joining the next version of the course (taking place January - March 2022) apply here to be a participant or here to be a facilitator. Applications are open to anyone and close 15 December. (We expect to be able to pay facilitators, but are still waiting to confirm the details.)

This post contains an overview of the course and an abbreviated version of the curriculum; the full version (which also contains optional readings, exercises, notes, discussion prompts, and project ideas) can be found here. Comments and feedback are very welcome, either on this post or in the full curriculum document; suggestions of new exercises, prompts or readings would be particularly helpful. I'll continue to make updates until shortly before the next cohort starts.

Course overview

The course consists of 8 weeks of readings, plus a final project. Participants are divided into groups of 4-6 people, matched based on their prior knowledge about ML and safety. Each week (apart from week 0) each group and their discussion facilitator will meet for 1.5 hours to discuss the readings and exercises. Broadly speaking, the first half of the course explores the motivations and arguments underpinning the field of AGI safety, while the second half focuses on proposals for technical solutions. After week 7, participants will have several weeks to work on projects of their choice, to present at the final session.

Each week's curriculum contains:

Key ideas for that week

Core readings

Optional readings

Two exercises (participants should pick one to do each week)

Further notes on the readings

Discussion prompts for the weekly session

Week 0 replaces the small group discussions with a lecture plus live group exercises, since it's aimed at getting people with little ML knowledge up to speed quickly.

The topics for each week are:

Week 8 (several weeks later): Projects

Abbreviated curriculum (only key ideas and core readings)

Week 0 (optional): introduction to machine learning

This week mainly involves learning about foundational concepts in machine learning, for those who are less familiar with them, or want to revise the basics. If you’re not already familiar with basic concepts in statistics (like regressions), it will take a bit longer than most weeks; and instead of the group discussions from most weeks, there will be a lecture and group exercises. If you’d like to learn ML in more detail, see the further resources section at the end of this curriculum.

Otherwise, start with Ngo (2021), which provides a framework for thinking about machine learning, and in particular the two key components of deep learning: neural networks and optimisation. For more details and intuitions about neural networks, watch 3Blue1Brown (2017a); for more details and intuitions about optimisation, watch 3Blue1Brown (2017b). Lastly, see Simonini (2020) for an introduction to how deep learning can be used to solve reinforcement learning tasks.

If you’re not familiar with the basics of statistics, like linear regression and classification:

Introduction: lin...

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 AGI Safety Fundamentals curriculum and application, published by richard ngo on October 20, 2021 on The Effective Altruism Forum.

Over the last year EA Cambridge has been designing and running an online program aimed at effectively introducing the field of AGI safety; the most recent cohort included around 150 participants and 25 facilitators from around the world. Dewi Erwan runs the program; I designed the curriculum, the latest version of which appears in the linked document. We expect the program to be most useful to people with technical backgrounds (e.g. maths, CS, or ML), although the curriculum is intended to be accessible for those who aren't familiar with machine learning, and participants will be put in groups with others from similar backgrounds. If you're interested in joining the next version of the course (taking place January - March 2022) apply here to be a participant or here to be a facilitator. Applications are open to anyone and close 15 December. (We expect to be able to pay facilitators, but are still waiting to confirm the details.)

This post contains an overview of the course and an abbreviated version of the curriculum; the full version (which also contains optional readings, exercises, notes, discussion prompts, and project ideas) can be found here. Comments and feedback are very welcome, either on this post or in the full curriculum document; suggestions of new exercises, prompts or readings would be particularly helpful. I'll continue to make updates until shortly before the next cohort starts.

Course overview

The course consists of 8 weeks of readings, plus a final project. Participants are divided into groups of 4-6 people, matched based on their prior knowledge about ML and safety. Each week (apart from week 0) each group and their discussion facilitator will meet for 1.5 hours to discuss the readings and exercises. Broadly speaking, the first half of the course explores the motivations and arguments underpinning the field of AGI safety, while the second half focuses on proposals for technical solutions. After week 7, participants will have several weeks to work on projects of their choice, to present at the final session.

Each week's curriculum contains:

Key ideas for that week

Core readings

Optional readings

Two exercises (participants should pick one to do each week)

Further notes on the readings

Discussion prompts for the weekly session

Week 0 replaces the small group discussions with a lecture plus live group exercises, since it's aimed at getting people with little ML knowledge up to speed quickly.

The topics for each week are:

Week 8 (several weeks later): Projects

Abbreviated curriculum (only key ideas and core readings)

Week 0 (optional): introduction to machine learning

This week mainly involves learning about foundational concepts in machine learning, for those who are less familiar with them, or want to revise the basics. If you’re not already familiar with basic concepts in statistics (like regressions), it will take a bit longer than most weeks; and instead of the group discussions from most weeks, there will be a lecture and group exercises. If you’d like to learn ML in more detail, see the further resources section at the end of this curriculum.

Otherwise, start with Ngo (2021), which provides a framework for thinking about machine learning, and in particular the two key components of deep learning: neural networks and optimisation. For more details and intuitions about neural networks, watch 3Blue1Brown (2017a); for more details and intuitions about optimisation, watch 3Blue1Brown (2017b). Lastly, see Simonini (2020) for an introduction to how deep learning can be used to solve reinforcement learning tasks.

If you’re not familiar with the basics of statistics, like linear regression and classification:

Introduction: lin...

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 Coordination within EA: community & ecosystems, published by CristinaSchmidtIbáñez on October 20, 2021 on The Effective Altruism Forum.

Coordination problems in EA seem somewhat pressing, but we are lacking a structured way to think through different kinds of coordination challenges and bottlenecks.

In this post we

Introduce a way to categorize and understand coordination services within the EA community: ecosystem & community member services

Present different potential services in these categories

Present a way to categorize coordination bottlenecks into three functional areas: knowledge, infrastructure and support

Our intention with this post is mainly to open up a discussion about this topic so that everyone can contribute to it and understand it better.

Introduction

The importance of coordination & coordination challenges has been periodically discussed for several years in the EA movement. In recent years, it seems that these challenges are becoming more acute - in November 2020 Benjamin Todd mentioned that currently, “the main bottlenecks might be organizational capacity, infrastructure, and management to help train people up, as well as specialist skills that people can put to work now”. (80,000 Hours episode). Under the assumption that these are the current bottlenecks of the community, coordination should help alleviate the effects of these bottlenecks.

The EA community is fairly highly collaborative and seems to focus on coordination more than other social movements. As such it does not appear that a lack of coordination related to building capacity and infrastructure has been the cause of negative outcomes (there are risks associated with info hazards or the unilateralist’s curse where a lack of coordination can be negative). Nevertheless, it seems that improving coordination would allow our community to do even more good. We think that there is significant room for improvement to build systems, or infrastructure, that could enable better coordination within the next 3-5 years and we would expect them (given the current trajectory of the community) to pay off for > 5 years after they’ve been refined.

In this post, we aim to map the meta landscape of internal coordination bottlenecks within the community. We hope this is useful for those thinking of or currently working in this space.

While we believe there are also significant opportunities to improve or develop coordination between EA and the communities adjacent to EA (“the outside world”), this post will focus more on internal coordination issues.

Note: “Organizations” refer to those which actively engage with the EA community, regardless of identification as EA-aligned or not.

Why internal coordination services?

We want the EA community to do not just some good but the most good possible. If we think that pursuing EA goals are very impactful, then even small marginal gains should be very impactful. The case for coordination in general has been previously made by Owen Cotton Barratt & Benjamin Todd; we have little to add to that.

Internal coordination services include those that help organizations & community members work together efficiently by sharing knowledge, or reducing resources required for undertaking specific actions. We think most of these are quite tractable.

We think internal coordination is particularly tractable (given adequate resources) because 1) Resources are fairly fungible within the EA community, 2) EA is a high trust community, 3) There is general consensus that (most) EA goals are worth pursuing. (See: The almighty Hive will for more)

However, there are forces that lead to suboptimal coordination including conflicting priorities for individual agents to invest time in coordination, principal-agent problems, lack of appeal of coordination work (e.g. coordination is “not sexy”), and...