empath75

The selzer poll was not a fake poll. It was an outlier. She was famously one of the best most reliable pollster in the country. Sometimes polls are just wrong just from pure bad luck and you still publish them if you trust your methodology.

It isn’t necessarily a crime to lie or do a hoax, and a fake poll by itself is probably well within the lines of free speech. The fraud question comes into it for two possible cases:

1) someone might have paid them to do a real poll and they faked it, which is just straightforward fraud with a clear victim.

2) (the fun one) they might have done it to manipulate political prediction markets, which gets into securities fraud questions.

Assuming that AI-driven game development ever becomes a real thing, they could literally lose every single developer they have now and still be fine.

Everything is political. Just people on the early internet tended to have a particular personality type that shared certain political views.

It is always incredible to me how quickly we go from "this entire technology is pure science fiction" to "nitpicking over details"

Yes you can look at the entire history of the stock. From IPO to today, there was never a time that you could buy it at a meaningful discount to the IPO price.

The point is simulation for autonomous vehicles and robots.

How do you know when the stock 'settles down'. Google stock _never_ settled down.

Anthropic's valuation has nothing to do with AGI and everything to do with claude code.

It isn’t but the borrow checker has a formal proof that does need mathlib.

Deporting someone that arrived here 20 years ago, who has a family, whose spouse might be an American citizen, who might have American citizen kids, grandkids, friends, employers, they might own a company and contribute to the economy, etc absolutely is a moral atrocity. Violating border crossing laws is not a crime that is worth upending so many lives. It’s a misdemeanor and the fact that you have to elevate it to a bank robbery to make it seem reasonable tells you everything you need to know. Not all crimes are worth ruining lives over.

I am in favor of deporting people that have not been here long, but there has to be a way to fine people and get them documented after a certain amount of time.

I am currently working on a language compiler in lean and it seems to be fine, but compilation times are horrific and having a mathlib dependency just to build a programming language is a big ask for people.

OTOH, my borrow checker is provably sound, which is why I am using it.

[flagged]

Deporting illegal aliens might be popular as an abstract concept but as soon as the question is about deporting actual people that they know, their opinion changes. It’s great politics to say you’re going to do it, but bad politics and a moral atrocity to _actually_ do it.

The only ethical way to do it is better security at the border and making legal immigration and visits much easier. You can’t just snatch people who are established in communities and put them into camps without leading to a thoroughly justified moral outrage.

Why do people say they built it, when LLVM built it. They didn't write any assembly code themselves.

If you're working with an AI to solve a mathematical problem, presumably it is contributing ideas. I think the scientists might be confusing the LLM's ideas with their own.

I think people are confusing two ideas.

for all n there is some natural number x such that BB(n) = x

x like all natural numbers is trivially computable.

But BB(n) as a function that you could run and _find_ x, is not computable in general, because it would entail solving the halting problem.

The nightmare scenario is an AWS or Azure zero-day given to a swarm of openweight agents.

I think eventually there will be sort of a centralized more or less automated repository for ingesting and sorting ai-generated lean proofs and making them searchable and re-usable. On some level it kind of doesn't matter if mathematicians can ingest the results, if coding agents can just search for them online and use them in their own proofs.

I actually think it would be very smart for the big AI labs to get together to fund an independent organization to manage such a thing, and hire mathematicians to run it.

What is happening now is that some aspects of mathematics are turning into essentially an exercise in software engineering. It is well known that proofs and computer programs have an isomorphism, and I think the eventual merger is more or less inevitable.

That's not to say that there isn't an infinite amount of work remaining for mathematicians to do. There are only so many problems that are going to be amenable to this approach.

LLMs find flaws in human-published math papers _all the time_, usually in the process of formalizing them in lean.

IME as claude works through a problem it develops its own internal vocabulary to describe things, and then the end result is often totally unreadable.

If you have ever worked with claude code and lean, it goes back and forth between lean and NL reasoning. I have done a few proofs with claude and it almost _never_ gets the argument right in prose. It usually has to go into lean and grind through proof obligations and then it finds problems, work arounds etc.

Because they have hundreds of potential proofs they were sitting on, and lack the internal expertise to judge them. It seems like they're working to formalize all of the in Lean, but that takes time.

It is wild to me how much stock mathematicians put into the social aspects of math. The whole social framework you are describing _did not exist_ for almost the entire history of math. It might very well be seen as a historical anomaly, like renaissance mathematicians challenging each other to solve cubic equations.

Claude could not fix them without a lot of help, so i did not relinquish sorting through those details in general. Just the drudgery of grinding through proof obligations.

> If the lean proof doesn't match the natural language one (which is the one the AI generated to solve the problem), it sounds like the lean proof isn't verifying the intended claim?

No, the other way around. The natural language proof was derived from the lean code, badly. This is my experience with using claude and lean to prove things. Its natural language explanations drift a lot from the lean, both before and after. But the lean code is the lean code.

Yes, exactly. There's no real pressure on AI to get the natural language version of the proof correct, and no way to really judge it automatically.

I recently spent 3 weeks with claude formalizing a CS paper about a borrow checker in lean, for a personal project.

The formalization went through, but there were _several_ mistakes in the original paper that it uncovered, from type setting errors to (many) formulas that quantified over all resources as printed, but actually applied to only arising resources in the calculus..

So the formalization did give me a formally verified borrow checker that I could use to build a programming language on top of, but it was _not_ exactly the borrow calculus that was printed in the paper.

I expect this is the most common experience when mechanizing a printed paper. There are a lot of skipped steps and handwaving.

Russia has bioweapons labs and specifically did research on aerosolized anti-biotic resistant plague.

PHD inference doesn't scale, though.