Linkpost for December
Effective Altruism
Global Poverty
Considerations for whether you should become a civil servant in a developing country [Probably Good].
Animal Advocacy
Answers to some questions about insect sentience [Rethink Priorities]. I don’t think this is exactly “frequently asked questions”, which implies that these are questions a lot of people ask. “Questions asked by people who take the possibility of insect sentience seriously and know what the type-identity theory of consciousness is”, maybe.
Particularly Good: I loved loved loved this article [Sandcastles] about welfarism vs. abolitionism and moderate vs. radical flanks in the animal advocacy movement. Just a handful of the insights: you should use more radical tactics (e.g. civil disobedience) for popular welfarist goals (e.g. ending battery cages) than for unpopular abolitionist goals (e.g. ending factory farming), because people with popular goals can get more radical before they come off as lunatics. Whether you prioritize welfarism or abolitionism is primarily a product of how long it will take before we can abolish factory farming. The division between welfarism and abolitionism is very arbitrary; there’s no principled reason that banning veal is abolitionist while banning battery cages is welfarist. Welfarist campaigns also make people aware of the evils of factory farming.
Existential Risk
How to identify AI-written fiction [Record Crash]. I was really wrong about how much backlash there would be against AI-assisted fiction writing, even in genres like webnovels which reward volume far more than prose quality. I think I’d failed to notice something this article references obliquely: human-written prose is proof of work. Before AI, an author with minimally competent prose was likely to also be minimally competent in terms of plot, character, and world. Today, an author can produce “good enough” prose without being good enough at anything else to write a story worth reading (by a definition of “worth reading” where Subnaruto counts).
If AI use were making decision-making worse, would we notice? [Dan Davies] Related: ChatGPT improves learning if people use ChatGPT as a tutor, but makes people less likely to learn if they use it to automate doing practice problems. [Arxiv]
There’s an Instagram page for a restaurant in Austin that doesn’t exist which displays AI-generated photos of its meals which, again, don’t exist [The Eater]. In spite of the article’s title being “Austin’s AI-generated restaurant explained”, it completely failed to explain it in any way.
AI systems can notice when researchers “inject” a concept into their internal thought processes [Experience Machines]. They seem to be able to tell apart injected concepts and the prompt (that is, to tell apart external stimuli from their own impulses). Models instructed to think about a particular concept (like aquariums) seem to be actually thinking about it while answering, with advanced models managing to think about a concept while not putting it incongruously into the output. Pretty cool stuff.
Anonymous experts on whether AI increases the risks of bioterrorism [80,000 Hours]. It looks like a lot of experts disagree about how much current AI models actually help potential bioterrorists: some of them believe it’s all hype, while others believe it makes biosecurity much more urgent and makes it much more likely that we’ll have to deal with a catastrophic pandemic.
Anonymous experts on how to handle infohazards in biorisk [80,000 Hours]. Some experts say that we should share information more widely so that people are more informed about the problems and have the information they need to solve problems. Others believe that we should tightly limit information that could be used by bad actors. Many experts think that information should be shared widely with stakeholders like policymakers, but that broad public access to information (e.g. in op-eds) isn’t helpful.
Other Causes
Bacteria can turn carbon dioxide in the air into protein [MIT Technology Review].
No new drugs have been approved for anxiety in the past twenty years [Our World In Data].
Meta Effective Altruism
The CEO of the Centre for Effective Altruism talks about where he sees effective altruism going in the future [Effective Altruism Forum]: building up stronger institutions; improving communications with people outside the community; and holding fast to effective altruist principles.
Recruitment for impactful organizations is very important, because the difference between the best candidate and an okay candidate can be measured in lives [Effective Altruism Forum]. However, we don’t know a lot about how to figure out which candidates will do best at a job, and not enough people are obsessed with making recruiting go well. Consider getting into recruiting, if you think it sounds like a cool sort of job!
Life Advice
Why people like your low-effort posts more than your effortposts [Eukaryote Writes Blog]. Low-effort posts tend to be short, interesting, controversial, and casually written, and not to require too much background knowledge.
How to start enjoying the company of people you don’t like [LessWrong]: generate possible sympathetic explanations for their annoying behavior; avoid situations in which they behave in an annoying way; discover positive traits about them and seek out situations where those positive traits are obvious; be kind to them; try to spend time with them (if that doesn’t make you hate them more).
Policymaking
Particularly Good: We complain about MAGAs on the right, but Democrats have their own problems with Very Online extremists [Infinite Scroll]. Many are genocide denialists and apologists for oppressive regimes, including China and Cuba. “One of the editors of Zohran Mamdani’s slick campaign videos also runs a YouTube channel where he pumps out pro-China propaganda, such as calling the idea of Uyghur repression a CIA plot.”
In Britain, current smokers are more likely to believe that vaping is more harmful than smoking than to believe it is less harmful [Our World in Data]. In reality, while vaping has some health risks, it is much safer than smoking. Vapes are one of the most effective tools for quitting smoking.
Education majors are vastly more likely to get As than people in other fields [American Enterprise Institute]. When they get jobs, teachers get overwhelmingly positive performance reviews, even though we know that teacher quality varies a lot. Because the standards for teachers are low, both during training and on the job, many teachers are simply not very good at their jobs. (Accuracy disputed in the comments here.)
Social Justice
In defense of the girlboss [The Argument]. Having women in leadership actually does make organizations more responsive to women’s concerns. The alternative to girlbosses isn’t not having women in leadership positions; it’s choosing unqualified women for representation purposes.
An abortion ban pushed one woman towards having an abortion [The Argument].
Many U.S. states indefinitely imprison some sex offenders after they serve their time, a process called “civil commitment” [Reason Magazine]. Some were convicted as teenagers and have lived their entire adult lives in prison; others are too sick or disabled to commit future sex offenses (e.g. they are paralyzed below the neck). In one Texas facility, residents aren’t allowed to touch each other, even to shake hands or give each other hugs. Residents are required to keep logs of when they masturbate and what they thought about. Residents aren’t allowed to speak to many of their family members, and one was even reprimanded because his wife sent him a picture of her in a tank top.
Some people have used studies comparing IVF successes and IVF failures to try to show that the gender wage gap isn’t primarily caused by having kids [Ruxandra Teslo]. However, IVF isn’t a good control: women shift into more motherhood-compatible careers when they’re planning to have kids, not just when they already have them; IVF failures themselves are very stressful in a way that might hurt women’s careers. A certain kind of intersex woman-- who are female-sexed but know from childhood that they can’t have children-- is a better control, and studies of those women show no gender wage gap.
Reality Has A Surprising Amount of Detail
Chinese Doom Scroll: Slacking off at work before smartphones: washing clothes, knitting, cooking dinner, reading novels, reading offline copies of Internet threads.
Feeding America—which provides food to food banks around the country—uses an auction system that allows food banks to bid “shares” on food shipments [Slow Boring]. Some shipments even cost negative shares (i.e. food banks are paid shares for taking them): these include potato chips and Tupperwares missing their lids.
Lots of people fall backwards into committing fraud without really meaning to. Very few of them try to commit murder to cover it up. And even fewer try the innovative “my client was too stupid to have known that her boyfriend was going to commit murder” defense [Toronto Life].
The Sin Nombre virus is a rare virus with a 40% mortality rate [Eukaryote Writes Blog]. The most interesting part of this article is that rats and mice aren’t distinct phyla and both have evolved multiple times (!).
Early humans probably got a lot of their protein from insects [Vox]. No word on whether they live in the pod.
The most ancient sense is taste [Eurydice Lives].
People who got rich using crypto can struggle to buy big assets like real estate, because the finance industry is very suspicious of crypto [New York Magazine]. Specialist brokers serve as intermediaries to help rich crypto people, like, buy anything with their money.
Racial preferences in dating are large [Vishal Prasad]. Black men, Asian men, and black women are particular losers in the dating market.
Short Stories
One Compile Man [Archive of Our Own]: the adventures of a programmer who can write any program in a single compile. I understand this to be fanfiction of One Punch Man but no understanding of One Punch Man is necessary to enjoy the story.
The New Yorker’s Isaac Chotiner Interviews Santa Claus [McSweeney’s].
Lamentation of the Aging Catboy [Tumblr]: Much prettier and sadder than you’d expect from the premise.
Excerpts from the Effective Altruist Talmud [Effective Altruism Forum].
The bit about the Sacred Stag in the header pic made me think of the use of things like tarot cards, which are heavily symbolic and may be a way of people tapping into their subconscious. Like, you could be subconsciously realizing you need to quit drinking and then you'll be more likely to interpret a Temperance card that comes up in that fashion, but also the Death card (which is supposed to technically represent change) might be read as a change of lifestyle, Justice meaning 'face the music and quit drinking', etc.
They're also often beautifully illustrated, which can help put one in a contemplative mood.
> How to start enjoying the company of people you don’t like
I think this is very often a very bad idea! My own dislike of people is an important signal for how I should act towards that person. There have been a bunch of occasions on which I disliked someone irrationally, and so I avoided interacting with them, which later resulted in me dodging a social bullet I might have been hit with had I artificially induced myself to like them and interact with them.
I would say in particular that artificially inducing liking someone is a bad idea if you're the sort of person to feel *guilty* for irrationally disliking people. Modifying your own emotional reactions because they make you feel guilty is a stereotypical mistake for a reason!! I used to feel guilty about disliking people, but eventually stopped when I realized that part of my psyche is actually incredibly well-calibrated.