[previously in series: 1, 2, 3, 4, 5, 6]

It is eerily silent in San Francisco tonight. Since Mayor Lurie's crackdown, the usual drug hawkers, catcallers, and street beggars are nowhere to be seen. Still, your luck can’t last forever, and just before you reach your destination a man with bloodshot eyes lurches towards you. You recognize him and sigh. "Go away!" you shout.

"Hey man," says Mark Zuckerberg, grabbing your wrist. "You wanna come build superintelligence at Meta? I'll give you five million, all cash."

"I said go away!"

"Ten million plus a Lambo," he counters.

"I don't even know anything about AI!" you say.

"I'll pay you fifty million to learn."

“F@$k off!”

[This is one of the finalists in the 2025 review contest, written by an ACX reader who will remain anonymous until after voting is done. I’ll be posting about one of these a week for several months. When you’ve read them all, I’ll ask you to vote for a favorite, so remember which ones you liked]

https://www.astralcodexten.com/p/your-review-islamic-geometric-patterns

I.

A thought I had throughout reading L.R. Hiatt’s Arguments About Aborigines was: What are anthropologists even doing?

The book recounts two centuries’ worth of scholarly disputes over questions like whether aboriginal tribes had chiefs. But during those centuries, many Aborigines learned English, many Westerners learned Aboriginal languages, and representatives of each side often spent years embedded in one another’s culture. What stopped some Westerner from approaching an Aborigine, asking “So, do you have chiefs?” and resolving a hundred years of bitter academic debate?

Of course the answer must be something like “categories from different cultures don’t map neatly into another, and Aboriginal hierarchies have something that matches the Western idea of ‘chief’ in some sense but not in others”. And there are other complicating factors - maybe some Aboriginal tribes have chiefs and others don’t. Or maybe Aboriginal social organization changed after Western contact, and whatever chiefs they do or don’t have are a foreign imposition. Or maybe something about chiefs is taboo, and if you ask an Aborigine directly they’ll lie or dissemble or say something that’s obviously a euphemism to them but totally meaningless to you. All of these points are well taken. It still seems weird that the West could interact with an entire continent full of Aborigines for two hundred years and remain confused about basic facts of their social lives. You can repeat the usual platitudes about why anthropology is hard as many times as you want; it still doesn’t quite seem to sink in.

https://www.astralcodexten.com/p/book-review-arguments-about-aborigines

[This is one of the finalists in the 2025 review contest, written by an ACX reader who will remain anonymous until after voting is done. I’ll be posting about one of these a week for several months. When you’ve read them all, I’ll ask you to vote for a favorite, so remember which ones you liked]

“The scientific paper is a ‘fraud’ that creates “a totally misleading narrative of the processes of thought that go into the making of scientific discoveries.”

This critique comes not from a conspiracist on the margins of science, but from Nobel laureate Sir Peter Medawar. A brilliant experimentalist whose work on immune tolerance laid the foundation for modern organ transplantation, Sir Peter understood both the power and the limitations of scientific communication.

Consider the familiar structure of a scientific paper: Introduction (background and hypothesis), Methods, Results, Discussion, Conclusion. This format implies that the work followed a clean, sequential progression: scientists identified a gap in knowledge, formulated a causal explanation, designed definitive experiments to fill the gap, evaluated compelling results, and most of the time, confirmed their hypothesis.

Real lab work rarely follows such a clear path. Biological research is filled with what Medawar describes lovingly as “messing about”: false starts, starting in the middle, unexpected results, reformulated hypotheses, and intriguing accidental findings. The published paper ignores the mess in favour of the illusion of structure and discipline. It offers an ideal version of what might have happened rather than a confession of what did.

The polish serves a purpose. It makes complex work accessible (at least if you work in the same or a similar field!). It allows researchers to build upon new findings.

But the contrived omissions can also play upon even the most well-regarded scientist’s susceptibility to the seduction of story. As Christophe Bernard, Director of Research at the Institute of Systems Neuroscience (Marseilles, Fr.) recently explained,

“when we are reading a paper, we tend to follow the reasoning and logic of the authors, and if the argumentation is nicely laid out, it is difficult to pause, take a step back, and try to get an overall picture.”

Our minds travel the narrative path laid out for us, making it harder to spot potential flaws in logic or alternative interpretations of the data, and making conclusions feel far more definitive than they often are.

Medawar’s framing is my compass when I do deep dives into major discoveries in translational neuroscience. I approach papers with a dual vision. First, what is actually presented? But second, and often more importantly, what is not shown? How was the work likely done in reality? What alternatives were tried but not reported? What assumptions guided the experimental design? What other interpretations might fit the data if the results are not as convincing or cohesive as argued?

And what are the consequences for scientific progress?

In the case of Alzheimer’s research, they appear to be stark: thirty years of prioritizing an incomplete model of the disease’s causes; billions of corporate, government, and foundation dollars spent pursuing a narrow path to drug development; the relative exclusion of alternative hypotheses from funding opportunities and attention; and little progress toward disease-modifying treatments or a cure.

https://www.astralcodexten.com/p/your-review-of-mice-mechanisms-and

Steven Byrnes is a physicist/AI researcher/amateur neuroscientist; needless to say, he blogs on Less Wrong. I finally got around to reading his 2024 series giving a predictive processing perspective on intuitive self-models. If that sounds boring, it shouldn’t: Byrnes charges head-on into some of the toughest subjects in psychology, including trance, amnesia, and multiple personalities. I found his perspective enlightening (no pun intended; meditation is another one of his topics) and thought I would share.

It all centers around this picture:

But first: some excruciatingly obvious philosophical preliminaries.

https://www.astralcodexten.com/p/practically-a-book-review-byrnes 

In June 2022, I bet a commenter $100 that AI would master image compositionality by June 2025.

DALL-E2 had just come out, showcasing the potential of AI art. But it couldn’t follow complex instructions; its images only matched the “vibe” of the prompt. For example, here were some of its attempts at “a red sphere on a blue cube, with a yellow pyramid on the right, all on top of a green table”.

At the time, I wrote:

I’m not going to make the mistake of saying these problems are inherent to AI art. My guess is a slightly better language model would solve most of them…for all I know, some of the larger image models have already fixed these issues. These are the sorts of problems I expect to go away with a few months of future research.

Commenters objected that this was overly optimistic. AI was just a pattern-matching “stochastic parrot”. It would take a deep understanding of grammar to get a prompt exactly right, and that would require some entirely new paradigm beyond LLMs. For example, from Vitor:

Why are you so confident in this? The inability of systems like DALL-E to understand semantics in ways requiring an actual internal world model strikes me as the very heart of the issue. We can also see this exact failure mode in the language models themselves. They only produce good results when the human asks for something vague with lots of room for interpretation, like poetry or fanciful stories without much internal logic or continuity.

Not to toot my own horn, but two years ago you were naively saying we'd have GPT-like models scaled up several orders of magnitude (100T parameters) right about now (https://readscottalexander.com/posts/ssc-the-obligatory-gpt-3-post#comment-912798).

I'm registering my prediction that you're being equally naive now. Truly solving this issue seems AI-complete to me. I'm willing to bet on this (ideas on operationalization welcome).

So we made a bet!

All right. My proposed operationalization of this is that on June 1, 2025, if either if us can get access to the best image generating model at that time (I get to decide which), or convince someone else who has access to help us, we'll give it the following prompts:

1. A stained glass picture of a woman in a library with a raven on her shoulder with a key in its mouth

2. An oil painting of a man in a factory looking at a cat wearing a top hat

3. A digital art picture of a child riding a llama with a bell on its tail through a desert

4. A 3D render of an astronaut in space holding a fox wearing lipstick

5. Pixel art of a farmer in a cathedral holding a red basketball

We generate 10 images for each prompt, just like DALL-E2 does. If at least one of the ten images has the scene correct in every particular on 3/5 prompts, I win, otherwise you do. Loser pays winner $100, and whatever the result is I announce it on the blog (probably an open thread). If we disagree, Gwern is the judge.

Some image models of the time refused to draw humans, so we agreed that robots could stand in for humans in pictures that required them.

In September 2022, I got some good results from Google Imagen and announced I had won the three-year bet in three months. Commenters yelled at me, saying that Imagen still hadn’t gotten them quite right and my victory declaration was premature. The argument blew up enough that Edwin Chen of Surge, an “RLHF and human LLM evaluation platform”, stepped in and asked his professional AI data labelling team. Their verdict was clear: the AI was bad and I was wrong. Rather than embarrass myself further, I agreed to wait out the full length of the bet and re-evaluate in June 2025.

The bet is now over, and official judge Gwern agrees I’ve won. Before I gloat, let’s look at the images that got us here.

https://www.astralcodexten.com/p/now-i-really-won-that-ai-bet

[This is one of the finalists in the 2025 review contest, written by an ACX reader who will remain anonymous until after voting is done. It was originally given an Honorable Mention, but since last week’s piece was about an exciting new experimental school, I decided to promote this more conservative review as a counterpoint.]

“Democracy is the worst form of Government except for all those other forms that have been tried from time to time.” - Winston Churchill

“There exists in such a case a certain institution or law; let us say, for the sake of simplicity, a fence or gate erected across a road. The more modern type of reformer goes gaily up to it and says, “I don’t see the use of this; let us clear it away.” To which the more intelligent type of reformer will do well to answer: “If you don’t see the use of it, I certainly won’t let you clear it away. Go away and think. Then, when you can come back and tell me that you do see the use of it, I may allow you to destroy it.” - G.K. Chesterton

https://www.astralcodexten.com/p/your-review-school

[Original thread here: Missing Heritability: Much More Than You Wanted To Know]

1: Comments From People Named In The Post 2: Very Long Comments From Other Very Knowledgeable People 3: Small But Important Corrections 4: Other Comments

https://www.astralcodexten.com/p/highlights-from-the-comments-on-missing-ed5

[I haven’t independently verified each link. On average, commenters will end up spotting evidence that around two or three of the links in each links post are wrong or misleading. I correct these as I see them, and will highlight important corrections later, but I can’t guarantee I will have caught them all by the time you read this.]

https://www.astralcodexten.com/p/links-for-july-2025

Stephen Skolnick is a gut microbiome expert blogging at Eat Shit And Prosper. His most recent post argues that contra the psychiatric consensus, schizophrenia isn’t genetic at all - it’s caused by a gut microbe. He argues:

1. Scientists think schizophrenia is genetic because it obviously runs in families
2. But the twin concordance rates are pretty low - if your identical twin has schizophrenia, there’s only about a 30%-40% chance that you get it too. Is that really what we would expect from a genetic disease?
3. Also, scientists have looked for schizophrenia genes, and can only find about 1-2% as many as they were expecting.
4. So maybe we should ask how a disease can run in families without being genetic. Gut microbiota provide an answer: most people “catch” their gut microbiome from their parents.
5. Studies find that schizophrenics have very high levels of a gut bacterium called Ruminococcus gnavus.
6. This bacterium secretes psychoactive chemicals. Constant exposure to these chemicals might be the cause of schizophrenia.

I disagree with all of this. Going in order:

https://www.astralcodexten.com/p/contra-skolnick-on-schizophrenia