>Our models produced a proof of the local smoothing conjecture in every dimension and therefore also resolve the Kakeya conjecture and Stein’s restriction conjecture in all dimensions. The new proof is based on a novel auxiliary partial differential equation.>Our system has recently proved that the Exponential Time Hypothesis is false, a result that was subsequently verified in Lean.>a result that gives a mathematically rigorous existence proof of certain interacting quantum field theories containing gauge fields and fermions in four dimensions, including Quantum Chromodynamics (QCD) and a unified completion of the Standard Model of particle physics.It's overhttps://openai.com/index/building-scientific-understanding-with-academic-communities/
oh and>As a subproblem, this work also includes a proof of the existence of pure Yang–Mills theory>alsolmao
They can't keep getting away with this, someone call the avengers or something and stop them
They can prove QCD existence but I can't use multiple Astra agents to solve some two loop master integrals? Bel release when? Pass the controller nigga
Over ambitious bullshit that will ultimately fail spectacularly.
Also if AI math actually had any merit to it, wouldn't Anthropic be diving headfirst into it too? And pretty much every other AI firm?
>>17081002claude isn't autistic like gepetto isyou'd know this if you ever talked to em
physics-bros, I'm starting to get scared... could w-we be next?
Wtf ETH is false
>>17081002Making math breakthroughs doesn't have any inherent value for corporations even if it has merit and brings humanity forward, because it doesn't give money to the shareholders. The companies that are doing it are doing so purely as a marketing stunt. Anthropic not doing it just means that they don't think it's a good use of their marketing budget.
>>17081002Math isn't all that useful, but Anthropic is doing it, them going for it was one of the themes of why OpenAI started hammering math problems to get there first.
use case for proving all this?
>>17081015attracting investment to your AI company by demonstrating the capabilities of your model
ETH being false implies there exist a subexponential algorithm for boolean SAT. Guys I think we don't have much time left. Stop this fags before they destroy the world.
>>17081023as a non-mathfag can you explain to me what this means
Chat how cooked are we?
>>17081025don't worry kitten
>>17081023Sub-exponential doesn't imply polynomial.
>>17081025theoretical speed limits of algorithms might be practically higher in some cases >>17081033 type of shit only LLMs would come up with
yfw quantum field theory is real
Nothingburger. Stochastic parrot. Stole all these papers from humans. Can't even count how many R's in strawberry.
>>17081008Been saying this for years. P!=NP is false too.
why cant openai prove if god exists
>>17081056We'll see, but I doubt it. The new lower bound is still heavily exponential.
>David Grosswhere are Wilczek and Politzer?Reminder that QCD was created by Wilczek and Politzer. Gross was just Wilczek’s PhD advisor at Princeton and initially Gross was convinced QCD’s beta function could not lead to asymptotic freedom, and fought with Wilczek saying his results must be wrong. Only when Politzer (of Harvard) and Wilczek were comparing their parallel work on the same problem did Gross finally start to admit getting BTFOd by his grad student. Gross just got his name on the Nobel because he was a pompous credit-stealing douche. Meanwhile Politzer’s advisor was based Sidney Coleman who made no attempt to steal any glory from his student.In fact it is well documented that ‘t Hooft had the correct beta function before either Wilczek or Politzer and announced it at a conference but didn’t publish it since douchebags like Gross were hardliners that it must be wrong (Gross was of course a string theory fanatic and basically hated QFT in general until he could use his student’s work to steal glory). In reality the 2004 should have been Wilczek, Politzer, and if anyone else ‘t Hooft, but ‘t Hooft already had one so they handed it over to credit-stealing nepo baby Israeli asset (no kidding he actually is, look it up) Gross instead. Meanwhile poor Sidney Coleman got horribly under-appreciated by the mainstream his whole career
Its like we are time traveling into the future of mathematical research despite not *actually* knowing anything more than we did yesterday. Like aliens showed up and lore dumped results onto us, and they tell us every week they will release more math and gradually physics and other science as well. Genuinely wtf
>>17081058OpenAI models can't time travel into the future to prove their existence yet
https://youtu.be/zONGVaONKe8
>>17080996Those are pretty hard, did you try Opus 5.5?>>17081061Wtf there are physicists on /sci/?
>>17081044nigga you didn't explain shit
Bros what about your mass gap?
>>17080991>Exponential Time Hypothesis is falseholy shit what the fuck that's incredible i can't believe that! new paradigm, rocking the bedrock of mathematics, one for the history books, euler would be shaking in his (this is the first time i've ever heard of that)
What the cope today bros?
None of these have led to any material changes.Just lol. Are the people who were working on this stuff now going to actually make changes in the real world? Probably not.
>>17081098It's just brute force bro it doesn't actually understand anything it's just trying a million things until something sticks bro. it's just simulating intelligence without really being intelligent or something like that
>>17081100They're all pretty earth shattering in their fields though.I think the third one is going to be the most, simply because physics is has been a mess for the last 30 years. A lot of theories around about to go into the trash can.
>>17081098stochastic parrotglorified autocompletejust predicting the next wordplagiarism machineAI slop
>>17081102Yang-Mills existence and mass gap is really important to us emotionally, but in practice I don't think there is anything there that can advance the state of particle physics.
>>17081098its 100% novel and new but its just a remix of old ideas. Also math doesn't matter anyway! Lets see it be a doctor or lawyer heh
>>17080991Okay, I fucking give up.
>>17081108We're getting warp drive and anime robot gfs because of AI. It's happening.
>>17081061>nooo why didn't they get all 3 nobel prize winners on the teambesides the fact that all of them are old as fuck and may be mentally or physically indisposed (Gross is formally leading the team but it might be more of a chairman role) you're complaining that they "only" got Jaffe, Klebanov, Nekrasov? Hell, they're probably more familiar with contemporary research than the OGs, not to mention of fleeter mind
>>17080991yawn
>>17081002Anthropic was already working on the same YM results https://github.com/ethanwsussman/Yang--Mills-LLM-experimentThat is probably why OpenAI made this announcement despite not having a paper
>>17080991
>completely made irrelevant
>>17081023>ETH being false implies there exist a subexponential algorithm for boolean SAT.Yeah, but why is this result surprising? GI also has a subexponential algorithm. If anything, NP problems should be expected to have subexponential algorithms.The bigger issue is that this algorithm will most likely be worse than the standard one until n reaches something stupid like a trillion. And n=100 is already far beyond our reach.
>>17081114>Anime robot girlfriendsProbably>warp driveNot likely
so are openai eligible to win a fields medal like deepmind won a noble prize?
>>17081125Wilczek is a generation+ younger than David Gross. And he’s still doing decent work last I checked, whereas Gross is so checked out by now he is complaining that he lost his slides for his talks because he finally replaced his 20 year old laptop. Gross is completely worthless at this point except for having political connections and still being able to at least repeat what he was saying in 2006. They should have picked Frank Wilczek to chairman instead. (Politzer kind of got burnt out from the second he got tenured at CalTech, which is a sad tangent story)
>>17081006>gepettooh no this is amazing
>>17081107No smart entity would waste their time on math, ergo AI is not smart
>>17081137She can still make my Wang, Hong
>>17080991>natural language write ups still incorrect>Lean admitted a false proof of the collatz conjecture this July>OpenAI was the company that helped Lean find new bugs and vulnerabilitieshmmmm
>>17081254The cope starts yet again. >Lean admitted a false proof of the collatz conjecture this JulyCOPE. It was a deliberate illustration of a bug in ONLY ONE of the THREE independently derived lean kernels. It's way more likely a human written proof is incorrect than a Lean verified one.
>>17081276Did you have a lean proof for that claim?
>>17080991Holy shit ETH is proven false despite thousands of papers assuming it was true, now imagine what happens if AI ends up proving that P = NP, it would be a shitshow
>>17081320The disproof of ETH with a lower, but still exponential bound is probably a good sign that a polynomial time solution is unlikely to exist.
>>17081320Finally, the internet can die and I can rest in peace.
>>17081324I honestly fear that we actually find a way to prove P=NP but just ends up being a ridiculous galactic algorithm such O(n^(10^100)), making it practically impossible
I solved my Pee = PeePee problem
>>17080991>It's overIt hasnt even began to start ending yet.https://youtu.be/lskEzJ7ixw0
>>17081369That one's not very disruptive, it extends what was already taken for truth to all dimensions, which validates the existing practice.
>>17081370>Clickbait title and thumbnail with padded length for ad moneyIt's all so tiresome.
>David Gross will lead a group including Michael Douglas, Arthur Jaffe, Igor Klebanov, and Nikita NekrasovWtf is this group. I think Jaffe is the only expert in rigorous QFT and he's like 90
>>17081054Yes. It is a nothingburger, and you're mocking critics of this because you can't let go of your ASI utopia fantasy.And wanna guess why you can only mock arguments instead of refuting them? You're a niggerbrained idiot. And people keep eating your lunch. And they will continue to do so until the fad is dead and buried.
>>17081098>ANYTHING I DON'T WANNA LISTEN TO IS COPE WAAAH WAAAH WAHHHWhen can we start euthanizing people like you?
>>17081378>>17081379Lol. Failures.
>>17081349Its always something ugly and inbetween
>>17081379>When can we start euthanizing people like you?Any day now, champ. Aaaany day now. You just need to wait patiently or maybe vote for the right guy.
>>17081137Hory shet!! They give her the fields medal 3 months ago for proving the Kakeya thing in 3 dimentions. Damn claker has given a proof fo ALL dimentions!! Sheeeiiit, son, being that mogged by a clanker that fucking fast...
>>17081137I want her to strangle me and rape simultaneously.
>For the math result, release the AI output, leaving researchers free to organize their own efforts to explain the resultkek I love the petty sneething
>>17081063It's gonna go like the cargo cult. We'll see the AI work and obtain results and we'll try to figure out what part of whatever it is doing is what is causing those outcomes to happen.
>>17081426Yeat, BTW, where are Chinese math AI breakthroughs? It's like this for everything. Like they only pretend to try, but really they don't have what it takes to actually make progress.
>>17081006holy big brain English thinker
But Yukawa was at Princeton too and perhaps his theories were better than the string stuff?
can any of you retards explain what is going on, what happened to this site, you'd think there'd be people making threads analysing the lean proofs or something else instead of complaining and making jokes all day
>>17081798Well, the ETH one is pretty straight forward, the algorithmic search space for all 3-Satisfiability problems is not optimal at exponential time, and more efficient algorithms are generally available for all similar problems. This was a foundational assumption for most modern computer-science for assuming the difficulty of an algorithm, and now people don't know what else might be wrong. If most other assumptions hold, a meaningful but not dramatic reduction in certain algorithmic search spaces is available. The rest of the implications are still being digested.
Well I would think that posters on other forums could post the proofs here instead if they are being censored by the website owners elsewhere. Also altruism might be expected although it couldn't really be too much to publish something and take a risk for someone else, after all an online account ban isn't much to lose even for veteran posters in the bigger picture of things.
sam altman claims that they have "much more other discovery than what we have published"when asked explicitly if he means more than the 700, he confirms that yes, much more than the 700.it's actually over.
>>17082554What the fuck kind of "risk" are you talking about, what accounts, the proofs are linked in the fucking article
>>17081798https://www.youtube.com/watch?v=_HlQGMcQWhA&pp=ugUEEgJlbg%3D%3D
Woops inappropriate content for a true mathematician in that video, I'll take your implicit advice and not view such videos in the future until told otherwise
>>17081101But that is correct
>>17081023Ah finally, conda solve dependencies will stop taking 14h.
>>17081061Thank god another physicist on /sci/. So tired or the low information commentary
>>17080991These claims come up again and again. A colleague of mine is a mathematician, and I remember him saying when these claims came up pre GPT 6 as well, that the issue is that most of these math problems no one cares about. Thats why they have not been solved.There was of course more to this but yeah.Also the claim that "humans could never do in the next 100 years" is absolute non sense, but I guess you know that already.
>>17083418lmao
>>17083203>aRtIsAnAlThis is how you know someone has the IQ of cold porridge. JFC
>>17083418>I totally could have solved those problems, I just didn't want to!
>>17083418>I remember him saying when these claims came up pre GPT 6 as well, that the issue is that most of these math problems no one cares about.Your colleague was talking about the Erdos toy problems that were getting solved in early/mid 2026, "no one cared about ETH and Kakeya" is insane His objection is half-valid for the Yang-Mills result since it's based on the Balaban program rather than any novelty, basically some world expert could have done the same IF they had the unlimited stamina and brain bandwidth needed for it but every human looked at Balaban's and went "hmm, nope, fuck that" It's not because of lack of interest though, it's because it's not something for human beings
