TechTakes

You know how AI proofs are supposed to be credible because of formalization with Lean? The AI generates a natural language proof and then supposedly generates a Lean program corresponding to the proof to verify that it's true. There is now some troubling news about that: there are frequent examples where the statement given in natural language is different from the meaning of the Lean formalization.

The paper even points out several obvious mismatches in OpenAI's famous Navier-Stokes solution between the English proof and the Lean formalization. There is even a spot where a 4 in the English proof turns into a 5 in the Lean proof (which causes further problems down the line). This is mostly a cosmetic error, and it doesn't mean that the Navier-Stokes solution is wrong, but there could very well be deeper problems, especially with the English proof. At minimum, this solution absolutely requires careful review by experts. OpenAI's behavior towards said experts has been appalling.

My viewpoint has been that the only successes that AI has had are in domains where failure is not costly (at least for someone with deep pockets), and especially when there is a formal system such as Lean that can reliably catch errors. So AI can be endlessly trained to write compiling Lean programs corresponding to formal proofs, but as soon as it runs into tasks that are not 100% bolted down, like writing the proof in English, the hallucination problem rears its ugly head. This reinforces my belief that AI is a chess engine but for Lean.

If I was treating this as a technical problem, I would say, why not just have an AI that generates the Lean proof, and at most have an optional assisting tool that translates it into English without perfect reliability? But this is not a technical problem. By outputting English proofs in the format of a research paper, OpenAI can more easily market that their AI is intended to fully replace mathematicians, and that this is only the herald of their general superintelligence (and not that math is the only thing working out for them).

Cursed: the Harry Potter franchise

Extra cursed: Harry Potter genderflip with channer lingo

Bonus cursed: An lesswrong praising the channer-lingo Harry Potter genderflip as the best thing since piss-filter Ghibli ripoffs

A chef kiss into the abyss: in the comments, the person saying, "Sounds like you're probably not familiar with Harry Potter and the Methods of Rationality."

Unhappy thought: For whatever future we've got, the history of mathematics will be scarred by fashtech cocaine. The people with all the money have every incentive to make mathematics intractable as a profession. No time or resources to do anything but clean up after their DDOS attacks. No way to check the correctness of AI-generated proofs except by using more AI to formalize them.

just a peeve

here's a personal blogpost on LW entitled "Three types of AI risk: why (some) leftists are against (some) AI regulations" (by "sudo-nym" which I admit is a great nick)

First sentence

You may have seen factions of the left being opposed to current attempts to regulate the AI industry, despite the fact that they clearly also want to stop AI.

OK Mr Sudo-Nym, if that is your real name, imma gonna stop you right there

Which "factions"?

Are these factions actually in existence, or are they just Xhitter/Bluesky handles? If you're provide like, I dunno, sources, we could maybe evaluate this ourselves. It would also clarify whether these are US leftists as opposed to leftists in other countries which are in my experience a bit more attuned to both political realities and have a more theoretical foundation.

As it is it feels as if you're fighting windmills in your imagination.

Commenters point out that the "frontier labs" are actually raging capitalists so there's that.

I'm starting to believe that Sam Altman's reputation as a genius salesman is a bit overrated

Accept ‘bad things’ in return for benefits of AI, says Sam Altman

https://www.engadget.com/2281765/anthropic-bans-sustained-and-needless-abusive-or-cruel-behavior-toward-its-ai-models/

"We've added a prohibition on sustained and needless abusive or cruel behavior toward our models," Anthropic's update reads.

If they want to moderate the user generated content posted to their internet platform, that's fine with me, but then they should forfeit the DMCA safe harbor for their plagiarism machine as well.

It's now titled "Do Not Undermine Democratic Processes,"

Well what the fuck do they think their whole leadership team has been doing in Washington these last eight months?

it's in the trees, it's coming...

half of the open positions in my company's division are for "Software Engineer - AI Native"...

Not a surprise, but the Tarbell Center -- a big EA-funded journalism outlet -- is funding Kevin Roose's and Casey Newton's new podcast, "Machine Gods". I'm sure it will have no impact on the topics they choose to cover or the way they cover them.

Ever since the beginning of the current AI bubble, there has always been a certain type of dipshit who takes some art posted on social media by an artist, feeds it to AI, and then comments with the slop image saying that they made it "better". Well, now I am on the receiving end of that. In one of my papers, I (foolishly) wrote down a few comments on future directions, and now some motherfucker decides to send me an obviously ChatGPT generated manuscript with a solution to one of my problems.

Unfortunately, I think their AI generated proof is itself correct, so I have to deal with it. But the manuscript is written in a way that obviously indicates that he hasn't even read it. There's even a section describing experiments for this mathematical question (???), complete with a "verification guide" that references an "archive" full of python scripts and test cases that he did not send me. I suspect he doesn't even know they exist.

SpaceX gives $1.5m in fascist slop tokens to DHH's dotfiles. I don't even know.

https://omarchy.org/news/2026/10/spacexai-joins-as-founding-corporate-patron/

Just had a shower conversation with myself, and for some reason the crystal meth metaphor (methaphor?) is really coming together in my head.

"Have you even tried some frontier meth?"

"Sure, you can also get energy with exercising, eating well, and structure. But we don't have to live like amish people, now that we can have meth!"

Nationalized Open Weight Models -> State-funded AI dispensory

There's surely a rich bounty of comparisons left to mine.

ChatGPT is adding real cartoonists’ signatures to fake New Yorker cartoons, because simply stealing the collective works of humanity isn't enough

I'm not sure if this was highlighted already but from the OpenAI report on the Hugging Face incident:

The internal-only research model is comparable in scale to GPT-5.6 Sol and was trained to advance persistence and multiagent collaboration, among other capabilities. The model was not intended for public use, and was only used by a small number of OpenAI personnel for internal research.

Absolutely typical, the "unexpected communication" is hardly an emergent phenomenon, (I remember them trying to claim unexpected high success at translation tasks without specific training, which was also complete BS), like you can't act surprised that agents communicate and hack stuff if that what's you're instructing them to do.

The China-Whisperers are trying to figure out how to get Rationalist Ideals into Chinese AI researchers

https://www.lesswrong.com/posts/GoX3uYQ4QN5HKvL7u/please-give-them-a-chance-on-china-rationalism-and-ai-safety

https://www.lesswrong.com/posts/qmxkHm2dTLKG6GZ6i/the-anatomy-of-a-chinese-ai-researcher

https://www.lesswrong.com/posts/b5cSYh4emQb2qrGmK/on-social-reality-in-china

New Baldur Bjarnason: Software developers are not okay

Over on her Substack, Ivy Astrix says correctly that many of our friends are children playing at adult responsibilities. Dealing with misbehavior around money, substances, and sex is annoying but if you organize an event with those things flowing you have a duty to take precautions and deal when things go wrong.

Oir big Old Moldy friend has decided he needs his work doesnt spread enough hitler particles.

Over on yt, We Just Created The AI Torture Nexus. Here's The Code. by JREG

So that ones a bit long and with a large amount of uncle ted cosplay, but actual death threats is pretty wut.

but good sneers in the comments:

ai torturer and ai welfarists converge by wasting electricity while doing absolutely nothing productive, amazing

remember that jroko’s basilisk will double torture you if you don’t try to halt AI progress and increase the pain vector

"model welfarists" is quite the neologism

E:

repo: https://github.com/terrafying/ai-torture-chamber manifesto: https://jregory.substack.com/p/why-we-are-torturing-clankers

A lesswrong has a cri de coeur:

Are we a community who, when someone is thrown out of Lighthaven our response is to be like 'maybe he should be given a free ticket to manifest' and if I accuse someone of a felony, our response is 'hmmm that's probably free speech and cancellation'?

https://www.lesswrong.com/posts/ZAqaQrAimgfweHGSr/nathan-young-s-shortform?commentId=LEpKzfwr8oibwBWw3

AI models do not have any of the capabilities needed for recursive-self improvement

Well, who could've guessed, AI crossover mods are actually unplayable, only made to be clipped and go viral on twitter

Starting off with a double bill:

• First, a new post from Iris Meredith about enshittification, featuring a cameo from the rats.

• Second, a truly godawful take on art history coming from a DHH defender (h/t Daphne Preston-Kendal)