Comment by jrflo
10 hours ago
I don't think it really attacks human understanding though. You can still read and understand an AI written proof. If another person comes up with a solution to a problem, you can read their methods and understand it. It doesn't matter if a human came up with that or not. It's really only attacking the "generating new ideas" part.
That's precisely the problem though. You cannot still read and understand an AI written proof at the current skill level of the AI being applied, because they're orders of magnitude longer than human written proofs even when they don't need to be, and spend most of that length on the parts that aren't important. This has been really thoroughly documented by expert mathematicians who are engaging with AI in public like Terence Tao and showing in detail how much work it takes working alongside AI to figure out how to understand AI generated proofs. With human generated proofs that process is forced to happen before publishing the proof because the new style of AI generated proofs validated only by formal verification is supplanting the old human peer review process that forced the burden of understanding onto the publisher and not the reader.
> You cannot still read and understand an AI written proof at the current skill level of the AI being applied, because they're orders of magnitude longer than human written proofs even when they don't need to be, and spend most of that length on the parts that aren't important.
Just like how they write software, then :-)
That doesn't seem to be true. The OpenAI NS paper was 166 pages. Wiles-Taylor proof of Fermat's last theorem is 129 pages. The length is not unprecedented for a difficult unsolved problem.
To be honest, I feel like the difficulty of reading AI proofs is due to the fact that we are on the verge of being beyond human comprehension. This is a demonstrable fact as no human has figured this out despite the problem being open for almost 100 years.
> To be honest, I feel like the difficulty of reading AI proofs is due to the fact that we are on the verge of being beyond human comprehension.
I can see where that's coming from, but I really don't think it's the case. Even with Astra, the proofs you get are just off in a way that doesn't signal superhuman comprehension. As 9question1 says, a common theme is that they dwell on insignificant steps. Another one is that they'll often be full of terminology that either doesn't exist, or has this weird quality where it looks like it is trying to make some minor insight seem much greater than it is. At first glance, that'll often make it look like it knows more than you, but when it's really just doing the same thing but in a more complicated and worse fashion, that to me isn't a signal of comprehension at all. The bizarre thing is that despite all the "stochastic parrot" style nonsense you'll get in individual proof steps, they still often combine to something valid.
In either case, what all of this means is that the working mathematician still needs to go through, and generally completely rewrite, any proof output by an LLM. Otherwise you are passing the burden of unreadability onto the reader.
2 replies →
> This is a demonstrable fact as no human has figured this out despite the problem being open for almost 100 years.
That's not true. Alpoge and Buckmaster's related LLM-assisted blowup result (https://news.ycombinator.com/item?id=49605915) utilized a strategy developed recently by Cordoba and Martinez-Zoroa.
This is a very token-brained take. The length of a work has no bearing whatsoever on its comprehensibility.
Not "a" human's understanding; Humanity's understanding. Understanding the research problem, and the solution especially, is a lot more involved than simply "read their methods". That's the whole point being made.
It matters if a human came up with it because of everything mentioned in the article... A mathematician's solution is necessarily built on other's ideas that have been disseminated, internalized, pressure tested etc. Methodologies differ too. AI can abuse its compute resources and generate a true/false or counterexample statements, without laying the foundation that a decade of globalized research would have.
>You can still read and understand an AI written proof.
No you can't lol, they're multi million lines of Lean, which is already an obscure language to understand. It's an assault on your senses.
It's not only Lean code, there are English writeups too. To my understanding the pipeline for these problems is 1) solve in english 2) formalize systematically to check. No one is tackling problems purely in Lean, to my understanding.
https://cdn.openai.com/pdf/32d9f210-8b73-45e0-91bc-82a30aef8...
[dead]