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: Two Prosocial Rejection Norms, published by Emrik on April 28, 2022 on LessWrong.

When I tell people that I'm shy, the universal response I get is "wow, you don't seem shy at all!"

Problem statement

Pop culture psychology has a single concept for "debilitatingly reluctant to approach people", and it's called "shyness" or "social anxiety" and is explained by "fear of rejection". I am debilitatingly reluctant to approach people, but I am very low on fear of rejection, what gives?

I propose that my 'shyness' can be explained by very high levels of what I'll call meta-fear of rejection, and fear of fake willingness.

Friends I've talked to about this all identify as having these to different degrees, and what works out to be optimal social norms for each person depends on this 'rejection profile'.

But before we go into it, let's be clear about what the goal is: We want to maximise opportunities for mutually beneficial interactions. Popular advice is often all about benefiting the receiver of said advice, so people are encouraged to "just walk up to them!" to overcome fear of rejection with little consideration for how this might impact others. And this is particularly useless as advice for people with high levels of 3 and 4.

Another obstacle to forming optimal social norms is the fuzziness of the language we use to express invitations, requests, and offers.

Sometimes people pretend to make invitations when they're really making requests, and this lets them benefit from the other person's willingness while claiming to not be indebted to them.

"It was just an invitation, I don't owe you anything!"

Conversely, sometimes people pretend to have heard a request when the other person actually intended to invite them, and that way they can claim to be up a favour.

"But I only did that as a favour to you!"

These are all forms of social manipulation enabled by a fuzzy language. And the deception wouldn't work without a high frequency of honest mistakes in this area.

So when I walk up to someone and invite them "hey, wanna be friends?" I have to worry about 1) fear of rejecting, and 2) them mistakenly or intentionally misunderstanding my invitation as a request. I just want to be able to communicate my invitation as exactly what it is, and for them to turn me down without hesitation and pain on their part.

Two Marginally Better Social Norms

Rejection aid

When you approach someone, include some words to the effect that you'll gracefwly accept a rejection.

Hey, wanna be friends? Keep in mind that I don't want to be friends unless you honestly want the same. I'll be happy if you say 'no' if you mean it, since that will have helped us both avoid wasting time on something we didn't want. :)

If you help them feel safe rejecting you, that benefits them because they'll be less anxious in the case they have to reject you, but it also benefits you by making sure that their 'yes' is truly a 'yes'. It also reduces the likelihood that the person you're asking will (intentionally or unintentionally) misinterpret your invitation as a request.

Upfront rejection readiness

Be upfront about your willingness to give honest rejections if that's what you truly feel like.

You know, you can always ask if I want to hang out with you. If I'm too busy or don't feel like it at the moment, I'll just let you know. So you aren't harming me by asking, you are only providing me with an opportunity to do something we both happen to want, if we want it.

This prevents other people from misinterpreting your invitations as offers. And for people like me with high levels of social fear 3 and 4, hearing you commit to rejecting invitations if you mean it, will make us feel safer approaching you.

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: Effective altruism’s odd attitude to mental health, published by MichaelPlant on April 29, 2022 on The Effective Altruism Forum.

[epistemic status: shower thought]

I'm starting to wonder if there's an odd tension within EA thinking about mental health. Members of the EA community seem increasingly aware that mental health can have a really big impact on our welfare and productivity and that we should pay more attention to it - see e.g. this great post on imposter syndrome last week or that apparently the most popular 80k podcast is about the difficulties of working with poor mental health. However, when we put on our cause prioritisation hats, we seem to forget about the existence and importance of psychological suffering. If we want to help others, it's clearly got to be bednets, cash transfers, deworming, etc. It appears not to cross (many) people's minds that mental health might be a plausible global priority.

[Content warning: anecdata] What brought this out to me was a chat I had with someone who breezed up to me at the careers fair at EAG London the other day. They were talking to me about how they wanted to run a project providing more mental health support for EAs because it was such a big problem. I was encouraging but explained I wasn’t that looped into mental health for EAs because I think focus more broadly on improving global wellbeing.

I then told them about my work at the Happier Lives Institute comparing the cost-effectiveness of cash transfers to treating depression and how we'd found the latter was about 10x better (original analysis, updated analysis). They suddenly switched to sceptical mode, saying they didn't believe you could really measure people's mental health or feelings and that, even if you could, targeting poverty must still be better.

After a couple of minutes of this, I suddenly clocked how weirdly disconnected the first and second parts of the conversation were. I asked them how they could be so sceptical of mental health as a global priority when they had literally just been talking to me about it as a very serious issue for EAs. They looked puzzled - the tension seemed never to have occurred to them - and, to their credit, they replied "oh, yeah, hmm, that is weird".

Now, I don't mean to pick on this nameless person, but I was struck by how vividly it revealed something I've observed more generally but not quite been able to put my finger on. It seems like many EAs have recognised mental health is a serious problem for people like them - community members are generally young, highly educated individuals from the global north - but not yet had the further thought that it is also a serious problem for others.

Clearly, it’s not the case that being depressed, anxious, traumatised, and so on are magically not bad if you're also poor and marginalised. At the very least, the line of thinking should be "oh, wow, mental health can be bad for people like me. I guess it could be a really serious issue for anyone. Let's see how it compares to our current priorities."

I'm not claiming that everyone has one thought too few; I'm just reporting that, in my experience, this disconnect seems oddly common.

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: Prize for Alignment Research Tasks, published by Andreas Stuhlmüller on April 29, 2022 on The AI Alignment Forum.

Can AI systems substantially help with alignment research before transformative AI? People disagree.

Ought is collecting a dataset of alignment research tasks so that we can:

Make progress on the disagreement

Guide AI research towards helping with alignment

We’re offering a prize of $200-$2000 for each contribution to this dataset.

The debate: Can AI substantially help with alignment research?

Wei Dai asked the question in 2019:

[This] comparison table makes Research Assistant seem a particularly attractive scenario to aim for, as a stepping stone to a more definitive [AI Safety] success story. Is this conclusion actually justified?

Jan Leike thinks so:

My currently favored approach to solving the alignment problem: automating alignment research using sufficiently aligned AI systems. It doesn’t require humans to solve all alignment problems themselves, and can ultimately help bootstrap better alignment solutions.

Paul Christiano agrees:

Building weak AI systems that help improve alignment seems extremely important to me and is a significant part of my optimism about AI alignment. [...] Overall I think that "make sure we are able to get good alignment research out of early AI systems" is comparably important to "do alignment ourselves." Realistically I think the best case for "do alignment ourselves" is that if "do alignment" is the most important task to automate, then just working a ton on alignment is a great way to automate it. But that still means you should be investing quite a significant fraction of your time in automating alignment.

Eliezer doesn't:

"AI systems that do better alignment research" are dangerous in virtue of the lethally powerful work they are doing, not because of some particular narrow way of doing that work. If you can do it by gradient descent then that means gradient descent got to the point of doing lethally dangerous work. Asking for safely weak systems that do world-savingly strong tasks is almost everywhere a case of asking for nonwet water, and asking for AI that does alignment research is an extreme case in point.

Everyone would likely agree that AI can help a little, e.g. using next word prediction to write papers slightly faster. The debate is about whether AI can help enough with alignment specifically that it substantially changes the picture. If AI alignment is 70% easy stuff we can automate and 30% hard stuff that we can't hope to help with, the 30% is still a bottleneck in the end.

Motivation for the dataset

We’re collecting a dataset of concrete research tasks so that we can:

Make progress on the disagreement about whether AI can substantially help with alignment before TAI.

Is there even a disagreement? Maybe people aren’t talking about the same kinds of tasks and the collective term “alignment research” obscures important distinctions.

If there is a disagreement, concrete tasks will let us make progress on figuring out the correct answer.

Guide AI research towards helping with alignment.

Figure out if current language models can already be helpful now.

If they can, help Ought and others build tools that are differentially useful for alignment researchers.

If they can’t, guide future language model work towards supporting those tasks.

As an important special case of step two, the dataset will guide the plan for Elicit.

Limitations

Ideally, we’d come up with tasks and automation together, iterating quickly on how to set up the tasks so that they are within reach of language models. If tasks are constructed in isolation, they are likely to be a worse fit for automation. In practice, we expect that language models won’t be applied end-to-end to tasks like this, mapping inputs to outputs, but will be part of compositi...

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: Virtue signaling is sometimes the best or the only metric we have, published by Holly Elmore on April 28, 2022 on The Effective Altruism Forum.

Subtitle: Costly virtue signaling is an irreplaceable source of empirical information about character.The following is cross-posted from my blog, which is written for a more general audience:

We all hate virtue signaling, right? Even “virtue” itself has taken on a negative connotation. When we’re too preoccupied with how we appear to others, or even too preoccupied with being virtuous, it makes us inflexible and puts us out of touch with our real values and goals.

But I believe the pendulum has swung too far against virtue signaling. A quality virtue signal shows that a person follows through with their best understanding of the right thing to do, and is still one of the only insights we have into others’ characters and our own. I don’t care to defend empty “cheap talk” signals, but the best virtue signals offer some proof of their claim by being difficult to fake. Maybe, like being vegan, they take a great deal of forethought, awareness, and require regular social sacrifices. Being vegan proves dedication to a cause like animal rights or environmentalism proportional to the level of sacrifice required. The virtuous sacrifice of being vegan isn’t what makes veganism good for the animals or the environment, but it is a costly signal of character traits associated with the ability to make such a sacrifice.

So the virtue signal of veganism doesn’t mean you are necessarily having a positive impact or that veganism is the best choice, but it does show that you as a person are committed, conscientious, gentle, or deeply bought into the cause such that the sacrifice becomes easier for you than it would be for other people. It shows character and acting out your values.

Out of your commitment to doing the most good possible, you may notice that you start to think veganism isn’t actually the best way to help animals for a lot of people.1 I believe this represents a step forward for helping animals, but one problem is that now it’s much easier to hide lack of virtuous character traits from measurement.2 It’s harder to know where the lines are or how to track the character of the people you may one day have to decide to trust or not to trust, it’s harder to support virtuous norms that make it easier for the community to act out its values, and it’s harder to be accountable to yourself.Many will think that it is good when a person stops virtue signaling, or that ostentatiously refusing to virtue signal is a greater sign of virtue. But is it really better when we stop offering others proof of positive qualities that are otherwise hard to directly assess? Is it better to give others no reason to trust us?

Virtue signals are a proxy for what actually matters— what we are likely to do and the goals that are likely to guide our behavior in the future. There is much fear about goodharting (when you take the proxy measure as an end in itself, rather than the thing it was imperfectly measuring) and losing track of what really matters, but we cannot throw out the baby with the bathwater. All measures are proxy measures, and using proxies is the only way to ask empirical questions. Goodharting is always a risk when you measure things, but that doesn't mean we shouldn't try to measure character.

The cost of virtue signals can be high, and sometimes not worth it, but I submit that most people undervalue quality virtue signals. Imagine if Nabisco took the stance that it didn’t have anything to prove about the safety and quality of its food, and that food safety testing is just a virtue signal that wastes a bunch of product. They could be sincere, and somehow keep product quality and safety acceptably high, but they are taking away your way of knowing...

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: Increasing Demandingness in EA, published by Jeff Kaufman on April 29, 2022 on The Effective Altruism Forum.

In thinking about what it means to lead a good life, people often struggle with the question of how much is enough: how much does our morality demand of us? People have given a wide range of answers to this question, but effective altruism has historically used "giving 10%". Yes, it's better if you donate a larger fraction, switch to a job where you can earn more, or

put your career to use directly, but if you're giving 10% to effective charity you're doing your share, you've met the bar to consider yourself an EA, and we're happy to have you on board.

I say "historically", because it feels like this is changing; I think EAs would generally still agree with my paragraph above, but while in 2014 it would have been uncontroversial now I think some would disagree and others would have to think for a while.

EA started out as a funding-constrained movement. Whether you looked at global poverty, existential risk, animal advocacy, or movement building, many excellent people were working as volunteers or well below what they could earn because there just wasn't the money to offer competitive pay. Every year GiveWell's total room for more funding was a multiple of their money moved. In this environment, the importance of donations was clear.

EA has been pretty successful in raising money, however, and the primary constraint has shifted from money to people. In 2015, 80k made a strong case for focusing on what people can do directly, not mediated by donations, and this case is even stronger today. Personally, I've found this pretty convincing, though in 2017 I decided to return to earning to give because it still seemed like the best fit for me.

What this means, however, is that we are now trying to build a different sort of movement than we were ten years ago. While people who've dedicated their careers toward the most critical things have made up the core of the movement all along, the ratio of impact has changed.

Imagine you have a group of people donating 10% to the typical mix of

EA causes. You are given the option to convince one of them to start working on one of 80k's priority areas, but in doing so N others will get discouraged and stop donating. This is a bit of a false dilemma, since ideally these would not be in conflict, but let's stick with this for a bit because I think it is illustrative. In 2012 I would have put a pretty low number for N, perhaps ~3, partly because we were low on money, but also because we were starting a movement. In 2015 I would have put N at ~30: a factor of 6 because of the difference between 10% and the most that people in typical earning to give roles can generally donate (~60%) and a factor of 5 because of the considerations in Why you should focus more on talent gaps, not funding gaps. With the large recent increases in EA-influenced spending I'd roughly put N at

~300 [1], though I'd be interested in better estimates.

Unfortunately, a norm of "10% and you're doing your part" combines very poorly with the reality of 100% of someone's career having ~300x more impact than 10%. This makes EA feel much more demanding than it used to: instead of saying "look at the impact you can have by donating 10%", we're now generally saying "look at the impact you can have by building your entire career around work on an important problem."

(This has not applied evenly. People who were already planning to make EA central to their career are generally experiencing EA as less demanding: pay in EA organizations has gone up, there is less stress around fundraising, and there is less of a focus on frugality or other forms of personal sacrifice. In some cases these changes mean that if someone does decide to shift their career it is less of a sacrifice...

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: Project: bioengineering an all-female breed of chicken to end chick culling, published by Cienna on April 28, 2022 on The Effective Altruism Forum.

In this post, "I" refers to Cienna's point of view, with Q&A answers and accuracy checking by Zane.

I'm backing the research crowdfunding campaign Can bioengineering be used to create an all-female breed of chickens? which asks, what does it take to get male chickens to lay eggs so the egg industry won't cull them? This experiment is a step toward a possible solution to ending the cruel and wasteful practice of male chick culling in the commercial egg industry. Mr. Zane Oberholzer, PhD candidate at the University of the Witwatersrand in South Africa, will be testing regulator sequences for their ability to selectively suppress the expression of the gene DMRT1 in gonadal tissue, resulting in the development of ovaries instead of testes. We’re posting here about this project and our conversation in case others are interested in funding this research experiment too.

This all-or-nothing raise is open until EOD Pacific Time on May 4th, 2022 (unless it gets extended, which sometimes happens).

The problem

Commercial chicken meat and chicken eggs come from different specialized breeds. For the egg industry, laying hens are needed, but only enough males to maintain the breeding population. Chickens have a sex ratio of roughly 50/50, so the male chicks are regarded as a wasted investment and culled.

One attempted partial solution to this problem is egg-sorting machines, which separate the eggs by sex before hatching. This saves part of the time investment, but despite having been around for a while, this tech is still disappointingly cumbersome.

An on-paper partial solution would be to transfer the unwanted male chicks to the meat industry, where they would live several weeks longer and maybe die a less painful death. But in practicality the laying breeds are physiologically very different from the "broiler" meat chickens and would not be wanted. Broilers have been selected for rapid growth and efficient conversion of high-protein feed into muscle mass. Layer breeds take longer to mature and have different physical proportions. A marketing premium would be needed to compensate for the higher cost of raising smaller, slower-growing birds.

The project

Well, what if the male chicks could grow up to lay eggs too? Then all the chicks would be wanted. No sex sorting would be needed. The same size breeding flock with the same resource inputs could produce twice as many egg layers.

Here's a background bit of clarification: hens lay eggs regardless of whether those eggs are fertilized. Ovulation results in egg laying. No roosters are needed for hens to lay eggs, only for those eggs to have the potential to develop into chicks. So a chicken doesn't have to be genetically fertile in order to lay eggs. The question is, does a chicken have to have the female combination of sex chromosomes in order to ovulate?

Creating a breed of chickens where all the adults lay eggs is beyond the scope of this small experiment; luckily, other researchers in the same lab are working on synergistic projects toward the shared goal of engineering such chickens. I'll let you read about this experiment's scope, details, methods, budget, timeline, ethics documents, et cetera on the project page, since Zane has already covered those.

Our conversation

At first the idea of an all-female breed of chickens sounded pretty "out there" to me. I took a coffee walk to sort my thoughts, and found I could imagine that it was doable, just not how it could be done. I returned with a bunch of poorly formed questions about sex determination in birds, and a curiosity to learn what I didn't know that made this seem doable to someone else. I reached out to Zane, who agreed to a video c...

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: 'Dropping out' isn't a Plan, published by Mark Xu on April 28, 2022 on The Effective Altruism Forum.

The following career-choice diagram looks silly:

"Go into finance" is a career plan; "Don't go into finance" is the absence of a plan, so it shouldn't be a choice-node. It makes more sense for the diagram to look like:

Similarly, many people seem to implicitly use something like the following career-choice diagram when thinking about school:

However, "don't go to school" isn't a career plan, so it shouldn't be a choice node. The diagram should look more like:

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: Virtue signaling is sometimes the best or the only metric we have, published by Holly Elmore on April 28, 2022 on LessWrong.

Subtitle: Costly virtue signaling is an irreplaceable source of empirical information about character.The following is cross-posted from my blog, which is written for a more general audience, but I think the topic is most important to discuss here on LW.

We all hate virtue signaling, right? Even “virtue” itself has taken on a negative connotation. When we’re too preoccupied with how we appear to others, or even too preoccupied with being virtuous, it makes us inflexible and puts us out of touch with our real values and goals.

But I believe the pendulum has swung too far against virtue signaling. A quality virtue signal shows that a person follows through with their best understanding of the right thing to do, and is still one of the only insights we have into others’ characters and our own. I don’t care to defend empty “cheap talk” signals, but the best virtue signals offer some proof of their claim by being difficult to fake. Maybe, like being vegan, they take a great deal of forethought, awareness, and require regular social sacrifices. Being vegan proves dedication to a cause like animal rights or environmentalism proportional to the level of sacrifice required. The virtuous sacrifice of being vegan isn’t what makes veganism good for the animals or the environment, but it is a costly signal of character traits associated with the ability to make such a sacrifice.

So the virtue signal of veganism doesn’t mean you are necessarily having a positive impact or that veganism is the best choice, but it does show that you as a person are committed, conscientious, gentle, or deeply bought into the cause such that the sacrifice becomes easier for you than it would be for other people. It shows character and acting out your values.

Out of your commitment to doing the most good possible, you may notice that you start to think veganism isn’t actually the best way to help animals for a lot of people.1 I believe this represents a step forward for helping animals, but one problem is that now it’s much easier to hide lack of virtuous character traits from measurement.2 It’s harder to know where the lines are or how to track the character of the people you may one day have to decide to trust or not to trust, it’s harder to support virtuous norms that make it easier for the community to act out its values, and it’s harder to be accountable to yourself.Many will think that it is good when a person stops virtue signaling, or that ostentatiously refusing to virtue signal is a greater sign of virtue. But is it really better when we stop offering others proof of positive qualities that are otherwise hard to directly assess? Is it better to give others no reason to trust us?

Virtue signals are a proxy for what actually matters— what we are likely to do and the goals that are likely to guide our behavior in the future. There is much fear about goodharting (when you take the proxy measure as an end in itself, rather than the thing it was imperfectly measuring) and losing track of what really matters, but we cannot throw out the baby with the bathwater. All measures are proxy measures, and using proxies is the only way to ask empirical questions. Goodharting is always a risk when you measure things, but that doesn't mean we shouldn't try to measure character.

The cost of virtue signals can be high, and sometimes not worth it, but I submit that most people undervalue quality virtue signals. Imagine if Nabisco took the stance that it didn’t have anything to prove about the safety and quality of its food, and that food safety testing is just a virtue signal that wastes a bunch of product. They could be sincere, and somehow keep product quality and safety acceptably high, ...

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 Might an Alignment Attractor Look like?, published by shminux on April 28, 2022 on LessWrong.

It is a generally accepted view in MIRI and MIRI-adjacent circles that the odds are high that an eventual self-improving AGI will not be nice to humans, to put it mildly. It is actually becoming a bit of a zeitgeist, gaining more and more widespread support. There are plenty of arguments supporting this view, including some observational data (like the infamous Boeing 737 MAX MCAS). It is easy to fall in with the prevailing wisdom in this particular bubble, given that really smart people give really persuasive reasons for it. So I would like to ask people to imagine alternatives: a state of the world where a self-improving general AI is naturally aligned with humanity.

Now, what does it mean to be aligned? We sort of have an intuitive understanding of this: a strawberry-picking robot should not rip off people's noses by accident, an AI that finds humans no longer worth their attention would disengage and leave, a planetary defense mechanism should not try to exterminate humans even if they want it shut down, though it can plausibly resist getting shut down, by non-violent means. Where, given some slack instead of a relentless drive to optimize something at any price, it would choose actions that are human-compatible, rather than treating humanity like any other collection of atoms. Where it would help us feel better without wireheading those who don't want to be wireheaded, and would be careful doing it to those who want to. Without manipulating our feeble easily hackable minds into believing or acting in a way that we would not consent to beforehand. There are plenty of other reasonably intuitive examples, as well.

Humans are not a great example of an animal-aligned intelligence, of course. Our influence on other lifeforms is so far a huge net-negative, with the diversity of life on Earth plummeting badly. On the other hand, there are plenty of examples of symbiotic relationships between organisms of widely different intelligence levels, so maybe that is a possibility. For example, maybe at some point in the AI development it will be a logical step for it to harness some capabilities that humans possess and form a cyborg of sorts, or a collective consciousness, or uploaded human minds...

It seems that, when one actually tries to think up potential non-catastrophic possibilities, the space of them is rather large, and it is not inconceivable, given how little we still know about human and non-human intelligence, that some of those possibilities are not that remote. There are plenty of fictional examples to draw inspiration from, The Culture being one of the most prominent. The Prime Intellect is another one, with a completely different bent.

So, if we were to imagine a world where there is a human-friendly attractor of sorts that a self-improving AI would settle into, how would that world look?

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: Jaan Tallinn's 2021 Philanthropy Overview, published by jaan on April 28, 2022 on LessWrong.

to follow up my philantropic pledge from 2020, i've updated my philanthropy page with 2021 results.

in 2021 i made $22M worth of endpoint grants — exceeding my commitment of $14.4M (20k times $718.11 — the minimum price of ETH in 2021). notes:

this number includes $1.9M to orgs that do re-granting (LTFF, EAIF, impetus grants, and PPF) — so it's likely that some of that $1.9M should not be included in the "endpoint grants in 2021" total. regardless, i'm comfortably above my commitment level for that not to matter;

i have an ongoing substantial charitable project that's not reflected in the 2021 numbers — it's possible (and likely if ETH price holds, as SFF's s-process alone can't handle such amount) that i will report it retroactively next year or in 2024.

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