Omg
Omg
Explain in Valorant terms
Unbannable cheats invented
probably closer to people learning how to counter strafe rather than unbannable cheats, it's a massive step for the math world to improve on the global scale but nothing universe-breaking (a lot of proofs already assume the hypothesis is true), unbannable cheats would be closer to something like proving P = NP
I meant it more in the sense of AI advancing super super fast.
They're solving a new millenium problem every month T_T
if you're talking about the implications to AI it would probably be closer to changing the VCT rules such that coaches can now talk at any time, and teams then build crazy superteams and the game becomes played at a much higher level
I meant implications to math. Normal players are human mathematicians. And AI has unbannable cheats. Regular human mathematicians are cooked
Humans working alongside AI are better than AI without humans, and better than humans without AI
This RH proof is 250k pages long in natural language (before it eventually gets simplified). Im sure the human mathematicians poked and prodded claude a bit, telling it what they want. Cant see that being the case for too long. Im just a chud masters student, who really knows?
I doubt it'll be verified and accepted for at least two years
where is da proof
oh wow yea someone said them and Claude made a ton of progress, let's wait to see if it gets verified (i doubt it will)
if it turns out a VLR thread is how i found out they proved the Riemann hypothesis that would be crazy
how would a generative ai chatbot be trustworthy in this case
what do you mean by this? a lot of recent top mathematical efforts have been by humans working alongside AI and then have been fact checked by either other humans or Lean
and im wondering how a generative ai model can be trustworthy in mathematics when it's notorious for making shit up and hallucinating as opposed to reasoning models
like i understand the use of ai in mathematics and other fields im just wondering how you can justify and trust use of a chatbot with a generative ai model like claude
ur smarter than me thats why im asking
so basically there is this thing called Lean which lets you deterministically verify mathematics proof, think of it similar to how function pythoning code would always compile on a normal compiler, or smth that will always deterministically tell you whether youre right or wrong. and this proof passed the lean verification, so as long as there are no kernel level bugs within Lean (the verifier) itself, the proof is correct. and the chances of Lean having a kernel bug allowing this is generally thought to be very low
tbh w u the whole reason behind the advances in maths by frontier labs is due to Lean, they have an easily verifiable goal to give models who can then hillclimb and loop for days with endless compute and always have a way to check if theyre right
alright that makes a lot of sense ty
Lean, it is not an AI program and has been the mathematical standard for years, #14 basically got it
it does need to be verified by humans, just to make sure no key assumptions were made that haven't been proven
i mean the authors arent stupid
they definitely ran it within the lean kernel and it verified
now if humans will be able to get anything useful from the result is a different question
but yeah this is like the biggest proof in maths ever
Lean4 certificates aren't absolute sources of truth !
https://www.arxiv.org/pdf/2610.08144
from what i can recall and my understanding in general, this paper is more about the fact that the natural language paper made by OAI during the whole NV-stokes drama was in fact not matching the lean proof. this doesnt mean that the lean proof was incorrect. just that humans/LLMs couldnt turn what was in the proof into writing
well it proves more than that. The whole discrepancy about lean is that you can technically prove false positives if you're ardent enough/ "soundness bugs" . A wrong/lax definition that doesn't correspond to the existent corollary/lemma and you can end up proving completely wrong things. Basically human audit is necessary, and we can't claim anything until that happens
fairs i guess, idk man i just dont think anthropic would let an incorrect proof of RH out in the wild right as theyre IPO is happening and i trust the Lean kernel it has positive aura and seems like a project with good intentions
well I can't speak on that. A proof of RH whether correct or not , would need a lot of time to be verified by the community, so "risking an incorrect proof" is of no concern here, as it'll all be done long after IPO anyway. I also like Lean, but it's just another prover. We've had Rocq for years now, I wouldn't say it's any more or less a better project.
but yeah, cool stuff
Yeah the lean code is 17 million lines
who cares? nothing ever happens. you and me still both chuds.
Me chud math student, AI taking away my job
have you watched the 3blue1brown video about AI in math
I've watched a bit, our opinions dont really match
i would drop out and become a dj
Im gonna flip burgers
lets goon together.
u gotta go into ai or finance before its too late good luck buddy
alternatively hope the bubble pops, or that ai takes over literally all the white collar jobs so everyone else is in the same boat
AI is going to achieve the same progress in STEM in the next 20 years that we humans did in the last 2000.
we've done more in the previous 20 years than the 2000 before that as well
Damn then with AI we might achieve it in like 5 years.
I mean the past 5 years were crazy.
I played a small part in it.
Wait so just as their IPO is around the corner...they have proved the hypothesis...damn crazy coincidence