Well deserved and well done to them on this proof. It is quite interesting to see that the AI bros continue to hype and worship hallucinating sophists like ChatGPT and GPT 4.
By now we should have already expected that AIs like LLMs are able to create unique proofs and new solutions to existing unsolved mathematical problems. They still haven't after years of hype and not even one single mention of buzzwords like 'LLMs', 'AI', 'GPT', etc in this thread. I'll tell you why:
The difference is those teenagers were able to clearly explain the process of deriving this proof transparently with the proof itself being (and still is) subject to intense scrutiny even by experienced mathematicians, going against what was thought to have been 'impossible'. Unlike the finest of LLMs and AI models which just repeat the same nonsense it has been trained on and confidently outputs more nonsense, whilst many celebrate this sophistry as a so-called 'breakthrough' even when it cannot transparently reason with its own decisions.
It goes without saying that these teenagers are very intelligent in mathematics to create this proof as it is not straightforward to just 'generate' it, given that it requires an amount of creativity AND originality that not even ChatGPT or LLMs in general can bullshit it's way around and will still tell you that it is impossible.
That proof is the true breakthrough; not magic AI black-boxes that spit out nonsense.
That doesn’t counter what I have said. The AI *did not* come up with the proof. It just summarized your own explanation based on the text you have given it.
Summarization of existing text is not the same thing as creating a proof from scratch.
It actually rewrote my explanation and made a substantial change to part of it, using a completely different chain of logic than I did. But you're right, it struggles to do original mathematics. When I asked it to write an explanation from scratch, it made a mistake in the proof.
I just wanted to point out that GPT 4 can be quite useful despite its shortcomings.
Why are being downvoted?, there have been crazy comment on this website, hyping to the max ChatGPT and how all the humans now are useless, I was just ignoring HN for the past weeks because it have been ridiculous
> ‘Probably because it has nothing to do with the article.’
Assuming you have read the comment, it is totally relevant to the article with AI (in this case LLMs or GPTs like ChatGPT) still not being able to transparently prove unsolved mathematical problems or even generate such solutions with and without supervision, since even if it was supervised, it will still generate it incorrectly and as with its black-box nature, it cannot reason or explain transparently.
The fact the those teenagers were able to create and derive this proof without regurgitation and withstood the scrutiny of experienced mathematicians tells us that it requires creative thought with transparent reasoning in the field to go against the books written by experts that once said it was ‘impossible’ until proven otherwise.
By now we should have already expected that AIs like LLMs are able to create unique proofs and new solutions to existing unsolved mathematical problems. They still haven't after years of hype and not even one single mention of buzzwords like 'LLMs', 'AI', 'GPT', etc in this thread. I'll tell you why:
The difference is those teenagers were able to clearly explain the process of deriving this proof transparently with the proof itself being (and still is) subject to intense scrutiny even by experienced mathematicians, going against what was thought to have been 'impossible'. Unlike the finest of LLMs and AI models which just repeat the same nonsense it has been trained on and confidently outputs more nonsense, whilst many celebrate this sophistry as a so-called 'breakthrough' even when it cannot transparently reason with its own decisions.
It goes without saying that these teenagers are very intelligent in mathematics to create this proof as it is not straightforward to just 'generate' it, given that it requires an amount of creativity AND originality that not even ChatGPT or LLMs in general can bullshit it's way around and will still tell you that it is impossible.
That proof is the true breakthrough; not magic AI black-boxes that spit out nonsense.