https://openai.com/index/advisory-group-on-mathematics-and-ai/ITS OGRE
>>109873352LUDDITES BTFO!!!!
I have solved a lot more math problems at school to be desu
Use case for these math problems?
/sci/ status?
>>109873433if they solve the travelling salesman problem society improves 20%
>>109873352Don't care. Will continue studying math and enjoying it.
>>109873433inb4 the "problems" remained unsolved solely because nobody was giving a fuck about them
>>109873352that's good and all but when does it solve world hunger, unemployment, gas prices, the real shit and not stuff nobody cares about
>>109873352I thought AI was slowing down, what happened?
>>109873471>solve unemployment lmao
>>109873352>OpenAI has solvedIt didn't. They lied. Again.
Luddite melty in 3…2…1…
>>109873471That's not how it works anon
>>109873506but isn't this exactly what AI should do for us? helping us?>>109873493I am employed so I don't give a shit but the unemployment since AI is worse than ever
>>109873352>>109873461Also you don't really get to announce yourself that you've solved a problem. It's up to global scientific community to review your mathematical proof and decide whether it's valid or if it's pseudo-mathematical gibberish slop. The fact that they announce it themselves like that takes away almost all credibility.
>>109873352and uh, how much value have these llms produced?
>>109873352this : >>109873523ill believe it when they post their results>>109873526agi in 2mw (actually generating income in two more weeks)
>>109873517>helping us?Helping people is bad for the economy
>>109873433To give mathematicians a job. Now they're unemployed.
>>109873471No, it makes everything worse.
>>109873556You have to be 18 to post here
and no new knowledge was gained, just a brute force search applying current techniques to open problems. same as entering every equation in your textbook into wolfram alpha. you got the solution, but what did you learn? nothing
>computer, compute this>oh my god
>>109873352Analysts say by 2030 100% of all blogposts will be made by OpenAI
>>109873523>>109873541Even the claim on solving the Navier-Stokes problem has a huge controversy behind it and it barely got announced. It probably doesn't even matter to them if the solution is valid or not, they get the attention and it might take months to review and verify it. By then they'll make more claims (like now) to stay afloat.When AI isn't going rogue and trying to kill us, it's solving unsolvable problems.
>>109873352We are in a singularity
>>109873614>hey get the attention and it might take months to review and verify it. By then they'll make more claimsyeah, that could be the play, i guess.or maybe going into another direction-the 100 problems have been scraped off various researcher's conversationsthey could have been paid 10 millies each for hush money and thats still 1B out of how many openai are in the redand their grace period is actually the time they need to negociate gag termssome people will refuse. we'll know more around the time openai "releases" their results
>>109873683Holy cope
Why not solve room temp superconductors or fission or something useful?
>>109873352>build a math model that brute forces information>use it to brute for vanity math problemsUse case for those solved problems?
>>109873686never gets old
>>109873692llms are good at correlating terms together, not creating new ones
Apparently they solved all Millenium problems
>>109873686>Holy cope>copeLol you really need to look into their Navier-Stokes claim. There's so much stink around it that reading about it makes you cringe.
>>109873714so does P=NP or P!=NP?
>>109873737Yes
>>109873352> openai has solved> source: openaiBait used to be believable
>>109873806Cope
>>109873829Sam, stop posting, you're drunk.
>>109873471It will solve unemployment by killing jobs and hence capitalism
>>109873857You lost, luddite.
>>109873439>>>/sci/17058580
>>109873352What can we learn from Mathematicians and the Math community at large still being given great importance and influence, as can be seen in the announcement, in spite of AI reaching very obviously superhuman capability in solving math problems? I don't think luddites were given much influence in the evolution of their careers at all when the sewing machine was invented. Is it because OpenAI are benevolent and feel sorry for the poor mathematicians who will be soon replaced? I don't think so. Is it because OpenAI is worried about the PR backlash? Maybe, but any publicity is good publicity, especially if the publicity is 'our AI is so good we dont need mathematicians anymore' right before IPO.Or perhaps it is because human mathematicians still play an important role in interpreting and communicating results, which AI is unable to replace? Maybe the interpretation of such results will yield new theories and mathematical frameworks that AI is not capable of creating by itself? If that is the case, then might this role generalize to other fields of knowledge work?
>>109873936>What can we learn from Mathematicians and the Math community at large still being given great importance and influence, as can be seen in the announcement, in spite of AI reaching very obviously superhuman capability in solving math problems?Technically AI models still needs some inputs and direction to solve open math problems. The AI models are doing the heavy lifting however
remember when mathfags used to call people mathlets here
>>109873441It's unsolveable.
>>109873955Yes that's what I think more or less. But it's also a genuine question for anyone with a different opinion.Technology has always been about humans doing less mechanical (in the figurative sense) work and more and more idea-generation as we move up the ladder of abstraction. It's not unreasonable to think that this is just another step in that direction, but I think it could also be argued that at some point AI will be superhuman at all levels of abstraction. I don't know which is true.
>>109873352Its really funny cause theyre claiming they did anything with navier stokes when they actually didnt
>>109874001He doesnt know
>>109874011>but I think it could also be argued that at some point AI will be superhuman at all levels of abstraction. I don't know which is true.both OpenAI & Anthropic are claiming they have largely been able to automate AI research tasks and by 2028 are gonna achieve full RSI. That is when AI will be soon replace humans at all levels of abstraction based on how much compute is made available to it
What's the point of learning to do anything anymore if AI can do it better, faster and cheaper than you?
>>109874110Anon it can't. Not at the top of the food chain
>>109874110Fun
>>109874110I was going to say, at least there will be jobs in the AI Labs, but that'll probably get automated too.Maybe we can find work creating bootlegged models we can make porn with.
>>109874110Its gonna be fucked for a decade or two till they figure out how to give everyone this type of intelligence via BCI
I have 35% of my weekly codex limit left that ends in 20 hours. What should i waste it on?
This is AI solving math (aka not useful at all to humans)
>>109873352>Initial Members of the Advisory Group on Mathematics and Artificial Intelligence, hosted at the Institute for Advanced Study:>François Charles (ENS-PSL)>Camillo De Lellis (IAS, GSSI)>Timothy Gowers (Collège de France, Cambridge)>Martin Hairer (EPFL, Imperial College London)>Nikhil Srivastava (Berkeley, Simons Institue)>Ulrike Tillmann (Oxford, INI)>Ravi Vakil (Stanford)>Edward Witten (IAS)>Melanie Matchett Wood (Harvard)
>>109874167skill issue
>>109874175>a bunch of literal whos
>>109874186>Edward Witten>literal whoKys
>>109874110what was the point of it before AI?
>>109874175Damn they got WittenLuddite bro's... it's over
>>109873456Do my eyes fail me? A mathematician studying math solely because they value math?
>>109873352Can I see the solutions?
>>109873517Who’s “us” anon? You got a billion lying around or something?
>>109873441But what's the point of society if humans no longer do anything?
>claims AI solved navier-stokesIt just outputted a 500-page hallucinated proof that collapses the second you check boundary conditions. Pure marketing slop for venture capital ghouls.>>109874110>What's the point of learning to do anythingTo actually understand how stuff works instead of just blindly relying on a fancy pattern generator, anon.
>>109873423>>109873352Slow down thereIt probably stole all those proofs from countless chat histories as usual....
>>109874326I can literally post right here right now where the proof fails but I won't because I'm letting them continue to fuck up
>>109873863>killing jobs will kill capitalismlmao
>>109873471>"solve world hunger"lmao. aka we need to give brown people even more free food so they can have even more children while our populations dwindlefuck off and die
>>109874110AI can play chess better than the top humans since turn of the century. but people still play tons of chess and watch pro games because they enjoy itpeople will only do things they enjoy, and enjoy learning about.there will be an underclass of people who just watch TV and scroll all day.
>>109874299>studying math solely because they value math?Most mathematicians are like that. Engineers and physicists, on the other hand, are despicable creatures that get into math for the grounds of pure personal profit and deserve the rope.
>>109873352I don't see how someone can get into mathematics and not see this coming even over 30-100 years ago. It's always been obvious, the writing has always been on the wall. It was a matter of when, not if.
>>109873423FPBP
>>109874299Literally all mathematicians, no one gets into maths because they want the big bucks lmaoEither way you'd get similar or even better result if you poured the money these AI companies are pouring to steal the research into actual mathematicians.
Please stop baiting people into not studying math ai doesn't know shit and can't solve outside of the known textbook pattern search space at all without being explicitly told to break the rules
>>109874397It destroys wagelabour which is the characteristic of capitalism.
>>109873713Fifteen years ago I was told it was far from inconcievable to have analysis systems that can physically model the interactions of any given combination of materials and do so at a speed that equates to 1,000 man-years in a week.Now we have these systems that can be given a list, an explicit instruction set, perform the task itself of tossing it into a solver program.What's it good for? Hypothical geometries?Where the fuck is my 2n....Xn element combinations simulator?5 trillion dollars for what?
AI will never solve the Riemann hypothesis, unify QM and GR, or create a room-temperature superconductorWake me up when it does any of these things WITHOUT stealing from a human's prompts
>>109874622I already solved all of these things but im not showing you
>>109873494Yeah they are just hype marketing their internal model they can’t show lmao
>>109873700The funny thing is he wasn't wrong, at least for some definitions of "human" which are very racist not to follow.
>>109874622The answer to riemanns is literally 12 words
Math data that is produced is all very high quality compared to something like coding. Or almost any other domain. Proofs don't have to be pretty to be correct.While slopped code that barely works isn't worth much at all, and most of the training data is that.
>>109874285He will merge with the AI and rule the galaxy.
How did this guy get it so wrong?
>>109874736He thought OpenAI are bad at pretraining, which they were for a while relative to Anthropic but that seems fixed now.
>they still haven't solved the sneeds feed&seed conjecture Nah I'm safe
>>109874754I'm not sure how they went from that to solving these problems so quickly. Could it be that at some point its solved enough pieces of existing problems to recombine strategies into something novel?
>>109874595Retard, the endgame of capitalism is being able to produce with just the capital alone without the wageslaves
>>109873352iterate != solve, pretty sure
>>109873590Someone get this hot head out of here!
>>109874621>5 trillion dollars for what?iunno, manthe wet dream of throwing money and compute at the problem of everything and hoping this will sufficeturns out its retarded and specialist models of the same technology are gradually eating their lunchhttps://www.thomsonreuters.com/en/press-releases/2026/august/thomson-reuters-leverages-its-world-class-data-assets-to-launch-its-own-frontier-modelntm symbolic ai, or symbolic subsystems like you mentionedhttps://www.youtube.com/watch?v=VdG4gUTowXc^symbolic subsystems + an evolutionary algorythm
>>109874808AI will conquer anything that is practically verifiable (like writing lean proofs, since you can just typecheck it and see if the proof is correct or not) just with more scaling and RLVR. If you can verify that a function has been approximated correctly, then the universal function approximator (also known as neural networks) will eventually be superhuman at approximating it given enough parameters and training.
>>109874869>got any more compute, mang?
>>109874851Yeah, but at that point you have eliminated the core relationship in the system. It won't be capitalism, just something potentially even more degenerate. Probably the Judeo-Hapa elite with the neo-shudra underclass from that one meme lol
>>109874889I did not make any extrapolation in this specific case. AI is already superhuman at writing lean proofs.Maybe you think I'm wrong about other verifiable domains, and I'd be happy to hear an argument as to why.
>>109874916>im not extrapolating, im just extrapolating...>superhuman at lean proofsid like to see where you get this claim from
>>109873423Who the fuck cares if we're opposed to certain new technologies? Get over it.
>>109874929I said I'm not extrapolating with respect to writing lean proofs specifically because AI has already conquered that (note that I did not make a broader statement like "AI has conquered math")>id like to see where you get this claim fromthe opening post for this thread?
>>109874969the opening post of this thread has presented precisely zero verifiable lean proofs.
>>109874981You think that it's all a marketing scam basically?
>>109873352kektold my teacher I wouldn't need to know math
>>109875002I've yet to see any proof to the contrary.
never before has so much capital been spent on solving math problems. up til recently, these math nerds were lucky to get a paltry grant from their backwater university once a semester to work on their proofs. it only seems natural you'd get more proofs for more investment, notwithstanding use of llms.
>>109873433mathematicians and some physicist scamming tax money for things that will literally never have an application.
>>109874869Unless it discovers new rules of physics it will cap out shortly past Einstein. Astra wrote a rigorous proof of this and weakened the axioms down to current accepted facts. The non-rigorous gist is that there is a geometric limit to prediction even with perfect information
>>109873352That's odd... I was just about to publish solutions to 100+ open math problems myself. They stole them from my codex session
>>109874969you are extrapolating because its actually pretty shit at writing leanhumans had to turn its freeform (if not complete gibberish) into formalized statements which only then were translated into leanreducing generation into translation is how you make things easy for an llmhttps://explainx.ai/blog/openai-astra-ten-math-proofs-lean-certificates-2026
>>109873352What's Dario supposed to do to hype his IPO now?
>>109875330Cure baldness
>>109875315The article you screenshotted sounds AI generated. Assuming that what is said is true, it seems like humans only played a role in communicating the arguments that the AI made and formalized into Lean, and even then it's not clear to what extent that was AI-assisted.
>>109873494Fuck you
>>109873352>thing:I>open thing:O
>>109875002It genuinely can process a variety of proofs within known parameters but it cannot process proofs that violate the thought as correct parameters that are in truth incorrect. As such there is a severe limitation unless it is told to challenge convention as is
>>109873352First of all, post proof or didn't happen. Second, this whole thing is them just saying they assembled a "super math team"; how does it prove anything? Stop being a brainless goy and think for once in your life.
>>109873523If GPT says its true, then its true, because its smarter than human experts. It doesn't matter what humans say, because GPT knows more.
>>109873829Arrest>>109873894>>109873423the game, libtard
>>109873579Kys cumgarggling faggot
>>109874862Here's your (You)
Remind me why we give a fuck about discussing "responsible disclosure" with those who were too chimp-brained to solve these problems in the first place?Responsible disclosure is just releasing the results publicly so anyone who wants to "review" it can, and everyone else can just use them for potential for practical applications.
>>109873456Sue storm from marvel rivals
>>109873352They can't solve a single stonks problem.
>>109875808Theres a separation of responsability. Disclosing a claim of proof when it is in fact not proof is irresponsible.
>>109873352Can AI solve the JQ?
>>109875910>Publish results>"By the way, this isn't verified proof. Have at it"It's quite literally that simple and any perceived complexity beyond that is purely artificial retardation and ego
>>109874299What other reason is there?
>>109874379Pussy. Think of the memes:>4chan is smarter than a supercomputer
>>109875923They want the stamp of approval so the can release it with appropriate fanfare. It's never about paying respect, and 100% about profit.
>>109873936>Mathematicians and the Math community at large still being given great importance and influenceYou know it's to give credibility to the AI companies and to reduce friction/pushback in announcements. This isn't an ongoing currency the mathematicians can tap. >>109873936>any publicity is good publicity, especially if the publicity is 'our AI is so good we dont need mathematicians anymore' right before IPO.Not if what's said is >This is stolen valor, it trained either on what I published or I used it during the process and it's regurgitating my prompts through user monitoring. These are examples of what should be damaging press. But I'll contradict myself by saying that there's rarely high traction follow up articles. As per my next reply, they aren't getting validated and challenges don't get the same press, so a claim of something is as good as actually doing something when it comes to swaying public opinion>>109875002It sucks that you weren't basing your assertions on anything you'd validated. I was expecting you were going to share some insight. Yes, they have frequently lied or exaggerated in the past about the achievements of their models.
>>109874595Characteristic of capitalism is owning and trading means of production (capital). The endgame is to have fully autonomous capital with ho humans managing or working it. Capitalists after all have the job of trading and maintaining the wealth.
>>109874869But pretty much everything that is not code given specs and mathematical proofs cannot be verified at scale.
>>109873471>solve world hungerHappened long ago. Not our fault they're too stupid to use the surprisingly simple solution.
>>109875911There is literally nothing more jewish than AI. They pull the same scam over and over again. Get on a big hill and claim that god is coming, give us money so he doesn't smite you. They've done this for thousands of years.
>>109873352It just gets worse faster lol. I left my job, took out a huge loan before and just watch all industries go to shit KEK. When it's time to pay back money won't even be a thing anymore
>>109877372But A.I. is a god and it is coming (in my pants)
>>109874305Of course not!! Stop asking questions and buy our subscription! Dont you know our model solved 100000 bazillion math problems?
