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 I gave AUD$12,573 to Innovations For Poverty Action, published by Henry Howard on November 29, 2022 on The Effective Altruism Forum.

I gave 50% of my 1st-year doctor's salary to charity last year. This was mostly to GiveWell-recommended and The Life You Can Save-recommended charities. The largest share went to Innovations For Poverty Action, a global development research organisation that designs and runs trials of global development interventions around the world in order to find which interventions are effective.

We need more global development cause discovery

The main reason I favoured Innovations For Poverty Action is my feeling that the slow rate of discovery of new effective charities is a bottleneck for effective altruism. From what I can see, GiveWell has added one charity to its top recommendations in recent years (New Incentives) while it's entirely removed its list of about 10 "standout charities". I haven't noticed many new additions to The Life You Can Save's list of recommended charities in recent years. GiveWell maxxed out the funding of its top charities last year and, while they claim they now have room for hundreds of millions more dollars, this is still a drop in the pond when compared to the total amount of philanthropy and government aid money that is spent annually worldwide. Finding further effective global development causes should be top priority, so that governments and philanthropists can be advised to direct their funds more effectively.

They probably know more than us

Effective altruists do a lot of independent research looking at effective ways to make the world better. This is great. An example is the recent Open Philanthropy cause exploration prize. Most effective altruism enthusiasts aren't Nobel prize-winning economists nor do they have decades of experience in global development nor do they have extensive global networks to feed them information. This all probably puts the average effective giving enthusiast at a disadvantage when it comes to seeing and seizing on global development opportunities. When it comes to effective cause discovery I think it would be difficult for anyone to outperform established global development research organisations like Innovations for Poverty Action, The Jameel Poverty Action Lab, and the Center for Effective Global Action, each of which have established networks, experience, and track records.

They have a good record

Innovations for Poverty Action has conducted research showing that giving free bednets is more effective than charging for them, they conducted the research that led to Evidence Action's Dispensers for Safe Water program, they conducted the research around No Lean Season that first appeared to yield promising results but at scale was less promising (negative results are important too). They were recently involved in a promising trial of cash-transfers and cognitive behavioural therapy to reduce crime among at-risk young men in Liberia.

You can give to them tax-deductably in Australia

I've considered giving to The Jameel Poverty Action Lab or another global development research organisation. In Australia you can give to Innovations For Poverty Action via The Life You Can Save and the donation is tax-deductable. I don't know of a way to give to other global development research organisations tax-deductably from down here.

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: Discussing how to align Transformative AI if it’s developed very soon, published by elifland on November 28, 2022 on LessWrong.

Coauthored with equal contribution from Eli Lifland and Charlotte Siegmann. Thanks to Holden Karnofsky, Misha Yagudin, Adam Bales, Michael Aird, and Sam Clarke for feedback. All views expressed are our own.

Introduction

Background

Holden Karnofsky recently published a series on AI strategy "nearcasting”: What would it look like if we developed Transformative AI (TAI) soon? How should Magma, the hypothetical company that develops TAI, align and deploy it? In this post we focus on How might we align transformative AI if it’s developed very soon.

The nearcast setup is:

A major AI company (“Magma,” [following the setup and description of this post from Ajeya Cotra]) has good reason to think that it can develop transformative AI very soon (within a year), using what Ajeya calls “human feedback on diverse tasks” (HFDT) - and has some time (more than 6 months, but less than 2 years) to set up special measures to reduce the risks of misaligned AI before there’s much chance of someone else deploying transformative AI.

Summary

For time-constrained readers, we think the most important sections are:

Categorizing ways of limiting AIs

Clarifying advanced collusion

Disincentivizing deceptive behavior likely requires more than a small chance of catching it

We discuss Magma’s goals and strategy. The discussion should be useful for people unfamiliar or familiar with Karnofsky’s post and can be read as a summary, clarification and expansion of Karnofsky’s post.

We describe a potential plan for Magma involving coordinating with other AI labs to deploy the most aligned AI in addition to stopping other misaligned AIs. more

We define the desirability of Magma's AIs in terms of their ability to help Magma achieve its goals while avoiding negative outcomes. We discuss desirable properties such as differential capability and value-alignment, and describe initial hypotheses regarding how Magma should think about prioritizing between desirable properties. more

We discuss Magma’s strategies to increase desirability: how Magma can make AIs more desirable by changing properties of the AIs and the context in which they’re applied, and how Magma should apply (often limited) AIs to make other AIs more desirable. more

We clarify that the chance of collusion depends on whether AIs operate on a smaller scale, have very different architectures and orthogonal goals. We outline strategies to reduce collusion conditional on whether the AIs have indexical goals and follow causal decision theory or not. more

We discuss how Magma can test the desirability of AIs via audits and threat assessments. Testing can provide evidence regarding the effectiveness of various alignment strategies and the overall level of misalignment. more

We highlight potential disagreements with Karnofsky, including:

We aren’t convinced that a small chance of catching deceptive behavior by itself might make deception much less likely. We argue that in addition to having a small chance of catching deceptive behavior, the AI’s supervisor needs to be capable enough to (a) distinguish between easy-to-catch and hard-to-catch deceptive behaviors and (b) attain a very low “false positive rate” of harshly penalizing non-deceptive behaviors. The AI may also need to be inner-aligned, i.e. intrinsically motivated by the reward. more

We are more pessimistic than Karnofsky about the promise of adjudicating AI debates. We aren’t convinced there’s much theoretical reason to believe that AI debates robustly tend toward truth, and haven't been encouraged by empirical results. more

We discuss the chance that Magma would succeed:

We discuss the promise of "hacky" solutions to alignment. If applied alignment techniques that feel br...

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 Giving What We Can recommends using expert-led charitable funds, published by Michael Townsend on November 28, 2022 on The Effective Altruism Forum.

Giving What We Can is looking to build more of an 'institutional view' on various aspects of effective giving. Something we've recently decided to more explicitly push for is donating via a fund. I'm cross-posting our page outlining why we recommend funds with hopes of getting feedback and questions from forum users :).

Funds are a relatively new way for donors to coordinate their giving to maximise their impact. This page outlines what a fund is, and why Giving What We Can generally recommends donating to funds rather than directly to individual charities.

How do funds work?

Funds allow donors to give together, as a community. Rather than each individual giving to a specific charity, donating via a fund pools their donations so that expert grantmakers and evaluators can direct those funds as cost-effectively as possible.

Using a fund is similar to using an actively managed investment fund instead of trying to pick individual stocks to invest in: in both cases, you let experts decide what to do with your money. This analogy helps explain the structure of a charitable fund, but it likely understates its benefits. This is because:

Investment funds regularly take a management fee (hedge funds, for example, typically take 1–4% of invested funds each year). Whereas the charitable funds we recommend don’t take any fees for their work.

According to the efficient-market hypothesis you should expect that any given stock costs roughly what it should given the best available information. If true, that would mean there’s not much room for experts to pick better stocks than you could (even if you picked at random!). Whereas the best charity can be at least 10 times better than a typical charity even within the same area. This means that there is substantial room for experts to make sure your donations do far more good than they otherwise would.

As we’ll see next, there are other advantages of funds — both for the donor, and the charity.

Advantages of funds

Funds make it easier to ensure that effective organisations receive the funding they need, when they need it.

Donating through an expert-led fund is often a much more effective way to support a cause than donating individually — even if you and the fund support the same organisations. This is because individual donors aren’t able to easily coordinate with each other, nor the organisations they support. Whereas if they donate together through a fund, the fund manager can:

Learn how much funding the organisation needs.

Provide funding when the organisation needs it.

Monitor how that funding is used.

Work with and incentivise organisations to be even more impactful.

As a result, donors might prefer funds because their money can often be allocated more efficiently and effectively.

But organisations also often benefit from the fund model. This is because:

Funds can provide a consistent and reliable stream of funding for effective organisations to carry out their work, whereas relying on individual donations can often be a challenge.

Fund managers can provide support in addition to funding. They often give advice and share key connections that can help the organisation succeed.

In general, because a fund pools the resources of multiple donors, and because a fund manager can spend more time investigating and supporting organisations than individual donors can, funds are often a highly cost-effective donation option.

When donating to funds may not be the best option

While we think most donors should give to funds, there are some cases where it might not make sense. This would be if:

You think you can find more cost-effective donation opportunities by yourself. For example, you may have...

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: How have your views on where to give updated over the past year?, published by JulianHazell on November 28, 2022 on The Effective Altruism Forum.

I'm wondering if your views on where to give have meaningfully changed throughout 2022.

If they have, please share! I'd love to hear why.

Personally, I'd say my biggest update has been towards giving opportunities that are more speculative but potentially higher EV. My giving portfolio in the past was heavily tilted towards GiveWell, primarily because of their demonstrated track record and the strong evidenced behind their top charities.

But now, I'm increasingly feeling comfortable with shifting my portfolio more towards other speculative options.

I still think GiveWell is an excellent choice, but I have a bit of a stronger appetite now for taking risks with my giving. This was partially motivated by a talk I watched Hilary Greaves give on making a difference.

Some charities I have given to/intend to give more to going forward include:

Happier Lives Institute for their work on improving competition and intellectual diversity in the effective giving space;

The Long Term Future Fund due to my increased belief that protecting the long-term future is an especially important moral priority;

And Animal Charity Evaluator’s Recommended Charity Fund due to my increased view that funding charities working to improve animal welfare is a particularly effective way to increase present-day well-being.

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: List of past fraudsters similar to SBF, published by NunoSempere on November 28, 2022 on The Effective Altruism Forum.

To inform my forecasting around FTX events, I looked at the Wikipedia list of fraudsters and selected those I subjectively found similar—you can see a spreadsheet with my selection here. For each of the similar fraudsters, I present some common basic details below together with some notes.

My main takeaway is that many salient aspects of FTX have precedents: the incestuous relationship between an exchange and a trading house (Bernie Madoff, Richard Whitney), a philosophical or philanthropic component (Enric Duran, Tom Petters, etc.), embroiling friends and families in the scheme (Charles Ponzi), or multi-billion fraud not getting found out for years (Elizabeth Holmes, many others).

Fraud with a philosophical, philanthropic or religious component

Bernard Ebbers

Prison: Yes

Jurisdiction: US

Amount: $18B

I find the section on his faith most informative:

While CEO of WorldCom, he was a member of the Easthaven Baptist Church in Brookhaven, Mississippi. As a high-profile member of the congregation, Ebbers regularly taught Sunday school and attended the morning church service with his family. His faith was overt, and he often started corporate meetings with prayer. When the allegations of conspiracy and fraud were first brought to light in 2002, Ebbers addressed the congregation and insisted on his innocence. "I just want you to know you aren't going to church with a crook," he said. "No one will find me to have knowingly committed fraud."

Also note that eventually, $8B was restored to investors.

Enric Durán

Prison: No, life in hiding

Jurisdiction: Spain

Amount: ~$700k (all amounts are inflation approximate and adjusted using in2013dollars.com)

During the 2008 crisis, he robbed Spanish banks by taking out spurious loans, and donated the amounts to anticapitalist causes. He wrote a guide about how to do this, and widely distributed it: an online version in Spanish and English can be found here.

Personal takeaway: Stealing money for altruistic causes is not unprecedented. And if you are going to cross that line, it can be done with much more style. It will also be viewed much more sympathetically if you steal from organizations perceived to be corrupt.

Tom Petters

Prison: Yes

Jurisdiction: US

Amount: ~$5B

Some of his donations were later returned (bold emphasis my own):

Petters was appointed to the board of trustees for the College of St. Benedict in 2002; his mother had attended the school. In 2006 he gave $2 million for improvements to St. John's Abbey on the campus of adjacent Saint John's University. In light of the criminal prosecution, St. John's Abbey arranged to return the $2 million gift to the court-appointed receiver for the Petters bankruptcy. In October 2007, Petters made a $5.3 million gift to the College of St. Benedict to create the Thomas J. Petters Center for Global Education. In 2006, he served as a co-chairman of a capital campaign at his high school, Cathedral High School, and offered to match donations up to $750,000.

Petters formed the John T. Petters Foundation to provide gifts and endowments at select universities to benefit future college students. The foundation was formed to honor his son, John Thomas Petters, who was killed on a visit in 2004 to Florence, Italy. The college student inadvertently wandered onto private property where the owner, Alfio Raugei, mistook him for an intruder and stabbed him to death." In response, in September 2004, Tom Petters pledged $10 million to his late son's college, Miami University. He later promised an additional $4 million, with the total to support two professorships and the John T. Petters Center for Leadership, Ethics and Skills Development within the Farmer School of Business. Miami Univers...

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Effective giving subforum and other updates (bonus Forum update November 2022), published by Lizka on November 28, 2022 on The Effective Altruism Forum.

TL;DR: We’ve launched an “effective giving” subforum — the banner will be up for a couple of days. We’re announcing this and a couple other updates.

More detailed summary:

We’ve launched the effective giving subforum and are working on our other test subforums ⬇️

Narrated versions of some posts will be available soon ⬇️

We’re testing targeted advertising for high-impact jobs ⬇️

We’ve fixed some bugs for pasting from a Google Document ⬇️

This is a short announcement post — we mostly wanted to give more context about the subforum. It’s also an opportunity for you to comment and share feedback, which we’d really appreciate (and you can always suggest new features here).

Effective giving subforum — and other work on subforums

We previously announced our first two pilot subforums (bioethics and software engineering). We’re now adding a third: effective giving — please feel free to join or explore it!

We hope that subforums will let users

Explore discussions and content about the subforum’s topic

Start discussions — including more casual threads — that will be visible only to people who have joined or are exploring the subforum

Keep up to date with news and ideas that are most relevant to them

We’ve made significant changes to the subforums since the first ones launched, and expect to make more (we’re aware of some bugs), so all your feedback is really valuable. You can pass that on by commenting on this post or emailing us at [email protected].

Other things to know about joining a subforum

Joining the subforum will automatically subscribe you to the related topic, meaning that posts with the relevant tag will stay on the Frontpage for longer for you (as if they had 25 extra karma points). You can change that by clicking on the bell icon on the subforum page and setting your preferences accordingly (e.g. by removing “upweight on frontpage”).

For now, joining will also add a tag to your profile. We plan to set up a way for users to opt out of this (and we might change the default settings), but it’s currently not possible. Note that you can also add more topics you’re interested in (whether or not they have associated subforums) by going to your profile and clicking “edit public profile” and finding the “My Activity” section.

Audio narrations of some Forum posts

There will likely be a proper announcement about this sometime soon, but, in addition to the Nonlinear Library, some EA Forum posts will get human narrations (that will be accessible in podcast apps and on the post pages themselves). This will be a collaboration with TYPE III AUDIO.

Testing targeted advertisements for impactful jobs

Some of you might have seen an advertisement for an open position at Metaculus (likely because you had interacted with the “software engineering” topic). We’ll be testing different types of job advertisements over the next month or so. As always, feedback is appreciated.

Bugs fixed for pasting from Google Documents

In some situations, images in posts that had been copy-pasted from Google Documents were broken. Some links were also failing. The major bugs here have been fixed.

Unfortunately, copy-pasting internal headers automatically (say, if you have a document with a table of contents that links to other sections in the document, and you want to copy-paste that into a Forum draft such that the table of contents links to sections in the Forum post) is currently not possible, and we might never make it work. The easiest way to create internal links to sections is outlined here.

Please give us feedback!

You can comment here, reach us at [email protected], or comment on the feature suggestions thread.

...

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: On the Diplomacy AI, published by Zvi on November 28, 2022 on LessWrong.

The latest AI development is: AI achieves human level in (blitz 5-minute-turn) full-communication anonymous online Diplomacy (paper). Why not?

I mean, aside from the obvious.

A take I saw multiple times was that AI labs, or at least Meta, were intentionally going for the scariest possible thing, which is why you create the torment nexus, or in this case teach the AI to play Diplomacy. If you had to pick a game to sound scary, you’d definitely pick Diplomacy.

The universal expectations for AI breakthroughs like this are:

The particular breakthrough was not expected, and is scary. The techniques used worked better than we expected, which is scary.

The details of the breakthrough involve someone figuring out why this particular problem configuration was easier to solve than you would expect relative to other problems and configurations, and thus makes it less scary.

We find that those details matter a lot for success, and that close variants would not be so easy. Other times we will find that those details allowed those creating the new thing to skip non-trivial but highly doable steps, that they could go back and do if necessary.

That is all exactly what we find here.

The actual AI, as I understand it, is a combination of a language model and a strategic engine.

The strategic engine, as I evaluated it based on a sample game with six bots and a human, is mediocre at tactics and lousy at strategy. Humans are bad at tactics (and often strategy) in games and Diplomacy is no exception. Diplomacy’s tactics a good match for a AI. Anticipating other players proved harder. The whole thing feels like it is ‘missing a step.’

What Makes the AI Good?

Where does the AI’s advantage come from? From my reading, which comes largely from the sample game in this video, it comes from the particulars of the format, and not making some common and costly mistakes humans make. In particular:

AI writes relatively long, detailed and explanatory communications to others.

AI does not signal its intentions via failing to communicate with its victims.

AI understands that the game ends after 1908 and modifies accordingly.

AI keeps a close eye on strategic balance in order to maximize win percentage.

AI uses its anonymity and one-shot nature to not retaliate after backstabs.

AI knows what humans are like. Humans were not adjusted to bot behaviors.

When people say the AI ‘solved’ Diplomacy, it really really didn’t. What it did, which is still impressive, is get a handle on the basics of Diplomacy, in this particular context where bots cannot be identified and are in the minority, and in particular where message detail is sufficiently limited that it can use an LLM to be able to communicate with humans reasonably and not be identified.

If this program entered the world championships, with full length turns, I would not expect it to do well in its current form, although I would not be shocked if further efforts could fix this (or if they proved surprisingly tricky).

Interestingly, this AI is programmed not to mislead the player on purpose, although it will absolutely go back on its word if it feels like it. This is closer to correct than most players think but a huge weakness in key moments and is highly exploitable if someone knows this and is willing and able to ‘check in’ every turn.

The AI is thus heavily optimized for exactly the world in which it succeeded.

Five minute turns limit human ability to think, plan and talk, whereas for a computer five minutes is an eternity. Longer time favors humans.

Anonymity of bots prevents exploitation of their weaknesses if you can’t confidently identify who they are, and the time limit kept most players too busy to try and confidently figure this out. They also hadn’t had time to learn how 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: Searching for Search, published by Nicholas Kees Dupuis on November 28, 2022 on The AI Alignment Forum.

Thanks to Dan Braun, Ze Shen Chin, Paul Colognese, Michael Ivanitskiy, Sudhanshu Kasewa, and Lucas Teixeira for feedback on drafts.

This work was carried out while at Conjecture.

This post is a loosely structured collection of thoughts and confusions about search and mesaoptimization and how to look for them in transformers. We've been thinking about this for a while and still feel confused. Hopefully this post makes others more confused so they can help.

Mesaoptimization

We can define mesaoptimization as internal optimization, where “optimization” describes the structure of computation within a system, not just its behavior. This kind of optimization seems particularly powerful, and many alignment researchers seem to think it’s one of the biggest concerns in alignment. Despite how important this is, we still understand very little about it.

For starters, it's not clear what internal optimization actually means. People have proposed several definitions of optimization which fit well when thinking about “behavioral optimization” where an agent acts to optimize its environment. One of the most clear and popular definitions comes from Alex Flint:

an optimizing system is a physical process in which the configuration of some part of the universe moves predictably towards a small set of target configurations from any point in a broad basin of optimization, despite perturbations during the optimization process.

In this framing, an agent and its environment together form an optimization process, where the agent acts upon the environment such that the system as a whole robustly converges to a set of target states.

But we run into problems when we try to map this behavioral definition to a definition about the structure of a process's internal computation. When we say mesaoptimization, we seem to mean something different than just that the computation converges to a smaller target. For example, an image classifier takes a large set of initial configurations of images including a lot of noise and irrelevant details, and layer by layer narrows it down to a probability distribution concentrated on a single class prediction. There seems to be a sense that this is not doing the kind of optimization we are concerned about when we talk about mesaoptimization.

Mesaoptimization was originally defined in Risks from Learned Optimization as internal search:

We will say that a system is an optimizer if it is internally searching through a search space (consisting of possible outputs, policies, plans, strategies, or similar) looking for those elements that score high according to some objective function that is explicitly represented within the system.

An advantage of this framing is that we do have some idea what we mean by “search” and have concrete examples of things which unambiguously qualify. Ideally, we’d like to point to the more general class of computation we’re worried about, but once you start thinking about what this general class of computation might look like, it quickly becomes clear that we don’t even know what “search” is. The space of search algorithms also seems much larger and more flexible than implied in the examples we usually think of.

At the moment we have very little idea what kind of algorithms we should expect neural networks to learn, and neither do we have a good picture of what kind of algorithms in particular we should be concerned about when we think of misalignment. If the intuition that “search” is somehow central to internal optimization holds validity, then becoming less confused about what learned search looks like should be central to making risks from internal optimization more concrete.

What is Search?

We have examples of processes which most woul...

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: My take on Jacob Cannell’s take on AGI safety, published by Steve Byrnes on November 28, 2022 on The AI Alignment Forum.

Jacob Cannell wrote some blog posts about AGI safety / alignment and neuroscience between 2010 and 2015, which I read and enjoyed quite early on when I was first getting interested in the same topics a few years ago. So I was delighted to see him reappear on Lesswrong a year ago where he has been a prolific and thought-provoking blogger and commenter (in his free time while running a startup!). See complete list of Jacob’s blog posts and comments.

Having read a bunch of his writings, and talked to him in various blog comments sections, I thought it would be worth trying to write up the places where he and I seem to agree and disagree.

This exercise will definitely be helpful for me, hopefully helpful for Jacob, and maybe helpful for people who are already pretty familiar with at least one of our two perspectives. (My perspective is here.) I’m not sure how helpful it will be for everyone else. In particular, I’m probably skipping over, without explanation, important areas where Jacob & I already agree—of which there are many!

(Before publishing I shared this post with Jacob, and he kindly left some responses / clarifications / counterarguments, which I have interspersed in the text, in gray boxes. I might reply back to some of those—check the comments section in the near future.)

1. How to think about the human brain

1.1 “Evolved modularity” versus “Universal learning machine”

Pause for background:

A. “Evolved modularity”: This is a school of thought wherein the human brain is a mishmosh of individual specific evolved capabilities, including a specifically-evolved language algorithm, a specifically-evolved “intuitive biology” algorithm, a specifically-evolved “intuitive physics” algorithm, an “intuitive human social relations” algorithm, a vision-processing algorithm, etc., all somewhat intermingled for sure, but all innate. Famous advocates of “evolved modularity” these days include Steven Pinker (see How the Mind Works) and Gary Marcus. I’m unfamiliar with the history but Jacob mentions early work by Cosmides & Tooby.

B. “Universal learning machine”: Jacob made up this term in his 2015 post “The Brain as a Universal Learning Machine”, to express the diametrically-opposite school of thought, wherein the brain has one extremely powerful and versatile within-lifetime learning algorithm, and this one algorithm learns language and biology and physics and social relations etc. This school of thought is popular among machine learning people, and it tends to be emphasized by computational neuroscientists, particularly in the “connectionist” tradition.

Here are two other things that are kinda related:

“Evolutionary psychology” is the basic idea of getting insight into psychological phenomena by thinking about evolution. In principle, “evolutionary psychology” and “evolved modularity” are different things, but unfortunately people seem to conflate them sometimes. For example, I read a 2018 book entitled Beyond Evolutionary Psychology, and it was entirely devoted to (a criticism of) evolved modularity, as opposed to evolutionary psychology per se. Well, I for one think that evolved modularity is basically wrong (as usually conceived; see next subsection), but I also think that doing evolutionary psychology (i.e., getting insight into psychological phenomena by thinking about evolution) is both possible and an excellent idea. Not only that, but I also think that actual evolutionary psychologists have in fact produced lots of good insights, as long as you’re able to sift them out from a giant pile of crap, just like in every field.

“Cortical uniformity” is the idea—due originally to Vernon Mountcastle in the 1970s and popularized by Jeff Hawkins in On Intellig...

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: 2022 ALLFED highlights, published by Ross Tieman on November 28, 2022 on The Effective Altruism Forum.

Executive Summary

Like many others, we are reeling from recent FTX news, and what it means for ALLFED and the whole EA Community. Alliance to Feed the Earth in Disasters was an FTX Future Fund grantee. Taken the current landscape, we were debating whether we should include it in these highlights. We have decided to do so in the interest of transparency and integrity, so as to accurately report on our January-November 2022 position.

We would like to start with massive thanks to Jaan Tallinn, whose generous support last year through the Survival and Flourishing Fund ($1,154,000) is a major reason why we are able to weather this storm (also a huge thank you to all our other donors, we appreciate each and every one).

2022 marks ALLFED’s 5th anniversary (2017-2022). Being a fully remote team, we now have team members on all continents except Antarctica. By the end of the year, we will have a presence in New Zealand due to David Denkenberger accepting a professor position at the University of Canterbury in Christchurch.

In these 2022 highlights:

To start with, we give updates on ALLFED’s 2022 research, including our papers and Abrupt Sun Reduction Scenario (ASRS) preparedness and response plans, including a recent proposal for the US government.

Next, we talk about financial mechanisms for food system interventions, including superpests, climate food finance nexus, pandemic preparedness, and our policy work.

We then move to operations and communications highlights, including our media mentions.

We next talk about events, workshops and presentations we have delivered this year.

We then dive into some major changes to our team (including at the management level), ALLFED’s internships and our volunteering program, and also give key statistics from this spring’s research associate recruitment (you will also find there imminent PhD opportunities as well as a temporary researcher position with David in New Zealand).

Finally, we thank those whose support we wish to especially recognize this year and talk about our funding needs for 2023, which range from dedicated funding to establish an ALLFED UK charity, to resilient food pilots, to support to continue key priority research projects on the topic of resilient foods for nuclear winter-level shocks, and to support preparedness and response plans (essential if we are to be able to present to decision makers within the current policy window).

There is no escaping the fact that, rather unexpectedly, our funding situation has worsened due to the FTX developments. We will therefore be especially grateful for your donations and support this giving season (please visit our donation webpage or contact [email protected] if you are interested in donating appreciated stock).

Since our inception, we have been contributing annual updates to the EA Forum. You can find last year’s ALLFED Highlights here, and here is our last EA Forum post EA Resilience & ALLFED's Case Study.

Research

It’s been a good year for research at ALLFED.

Papers

We have submitted 4 papers to peer review, one of which has now been accepted and published.

Authors: David Denkenberger, Anders Sandberg, Ross John Tieman, Joshua M. Pearce

Status: Published (peer reviewed)

Journal: The International Journal of Disaster Risk Reduction

This paper estimates the long-term cost-effectiveness of resilient foods for preventing starvation in the face of a global agricultural collapse caused by a long-lasting sunlight reduction, and compares it with that of investing in artificial general intelligence (AGI) safety. Using two versions of a probabilistic model, the researchers find that investing in resilient foods is more cost-effective than investing in AGI safety, with a confidence of ...