Remix.run Logo
▲ root_axis an hour ago

What value is there in just dumping something into an LLM and saying it produced errors?

▲arcastroe an hour ago | parent | next [-]

It's being hurriedly announced as fully human work. I wonder if the author had carefully avoided ever running the full argument through frontier LLMs (OpenAI/Anthropic) for fear of it being stolen or incorporated into AI training.

This may be the first time this proof is being proof-read by frontier models

(this is pure speculation, but provides an answer to your question)

▲hkmaxpro 39 minutes ago | parent [-]

I also suspect the author actively tried to avoid using AI at all.

Another possible reason: AI will correct your mistakes by using ideas of others. A proof will unlikely remain human-only, even if you just use AI to check for your errors.

▲markasoftware an hour ago | parent | prev | next [-]

Have you tried doing this recently? For my work on software at least, the latest LLMs have a very low false positive rate for bugs. They almost never hallucinate issues. My understanding is the same applies to math.

▲root_axis an hour ago | parent | next [-]

I'm not arguing that LLMs aren't useful tools for mathematicians, I'm questioning the quality of discourse that amounts to "I pasted this into the LLM and it said there are errors".

▲fasterik 44 minutes ago | parent | next [-]

There's value in knowing whether a claim is true or false. Plugging things into an LLM doesn't necessarily give 100% certainty or tell us why something is true or false in a human-comprehensible way, but it can move our credences and suggest directions for investigation.

▲whimsicalism 33 minutes ago | parent | prev [-]

classic HN isolated demand for rigor. i found it interesting

▲mahogany an hour ago | parent | prev | next [-]

I would guess that LLMs are much worse at math without having a theorem prover to validate approaches with. Since this draft is not formalized, an LLM is operating purely on the text which is more tenuous.

▲shmoil an hour ago | parent | prev | next [-]

Pre-chatgpt 5.5 it hallucinated quiet a bit on hard math. Starting from 5.6 it is incredibly good.

▲gregorygoc an hour ago | parent | prev | next [-]

I’m doing it daily and it finds a bunch of issues which are not issues in practice.

▲simplyluke 25 minutes ago | parent | prev | next [-]

I suspect LLMs are much better trained and more reliable at bug identification in react apps than draft-state proofs on the bleeding edge of math, on problems that are specifically highlighted as extremely difficult to solve without error in their training.

▲acedTrex 3 minutes ago | parent | prev | next [-]

> the latest LLMs have a very low false positive rate for bugs.

Man people live in different worlds, astra for me finds 25/6 bugs all the time.

▲SubmarineClub 21 minutes ago | parent | prev [-]

Cool story.

I have the opposite experience in my line of work.

▲chaps 19 minutes ago | parent [-]

What's your line of work?

▲augment_me 42 minutes ago | parent | prev [-]

The author should have done this ideally, since the author has been lazy and not done this, it's good that someone else has.

▲slopinthebag 33 minutes ago | parent [-]

ah yes, the author who has been working on this problem for 12 years is lazy because they didn’t feed their work into the slop machine so it could train off it and steal their work.

this place honestly sucks

▲acedTrex 2 minutes ago | parent | next [-]

It has to be new people that have come here, this place was NEVER this terrible before. Now it seems like the world has conspired to direct the worlds greatest idiots to descend on this poor site.

▲augment_me 28 minutes ago | parent | prev [-]

Well he clearly lives in the old world. It's a bit like someone not checking a search engine before cleaning some basic fast wrongfully that can be verified in 10 seconds.

▲slopinthebag 25 minutes ago | parent [-]

not everyone is a slave to the machine, man

▲augment_me 20 minutes ago | parent [-]

And see where that gets you, second place and 11 years down the drain with a half-working mess, congratulations hope the winning of not being a "slave to the machine" is feeling great.

▲acedTrex a minute ago | parent [-]

> second place

If the result holds then this is the paper that will get the medal and what history will remember. First place enough in my book.