>A Severe Misalignment of AI in MathematicsSigned by 25 fields medalists.https://mathandai.org/
>>17052244Who?
>>17052247CEO of math
>>17052244These assholes have been shilling Lean and AI math for years now. You reap what you sow. And let's be real, it's probably just an excuse to perform a regulatory capture. More laws and barriers so no one else can follow openai's footsteps.
so farartists: meltdowncoders: submit to the machine god, preserve dignity by being humblemathematicians: meltdownphysicists are next. let's see how they will fare
>>17052250>artists: meltdownlol no. let me know when chatgpt can put oil on canvas or carve stone. until then art is safe.
>>17052251yeah bud. we'll let you know soon as. be safe. sleep tight
>>17052250>physicists are next. let's see how they will farethey're already actively using it in hep pheno. just ask jesse thaler
>>17052264sure thing, like terry was using it a year ago and writing how wonderful life is and future is brightlet's see when it starts making huge progress autonomously with cheerleaders just prompting "go on, you can do it!"
>>17052266>https://mathandai.org/stringfags might cry but they're mathematicians in disguise.
>>17052244Pure math has no value outside of human intellectual curiosity. Thus having an AI do pure math has no value at all. It is pretty simple. The only reason they are currently using it to solve math problems is for clout.
>>17052244lmao they are actually whinging. This is the institutional math. They should just set up disproofs like anything you see here. An AI company shits up the filed with unverified claims and they are expected to be taken seriously. The AI prompter is literally a schizophrenic with no understanding of the field in question. If he can't defend his AI proof, then he can fuck off.This whinging about the importance of their ivory tower is disgusting. Better critiques have been stated in a lolicon imageboard. An AI proof is quite literally making up a language nobody can read and declaring you have found the proof in that language and that your banker should therefor give you another 100 gorillion dollars. The larger issue the mathematicians should just close rank and build their own AIs. THe sweetest revenge is bankrupting pseud companies. But they can't close ranks because of the openness of academia. Everything these AI companies have done is the logical end of the institutional structure. The farming of the masses for the benefit of the few. Copyright be damned. IP be damned. Privacy be damned.
>>17052268you contradict yourselfif it had no value they wouldn't throw millions at it
>>17052273Like I said, they are throwing millions of dollars at it for clout. To make their systems look intelligent and one up each other. The math itself has no value.
>>17052276that's the newfound value of math then - as benchmarks for advertisement
>>17052244>mathematicians realize they'll be made irrelevant by AI>comes out against AIYeah, let's hold back all of society just so mathematicians can still feel like they're important. Turns out "academics" are just as self-interested as everybody else.
>>17052276people don't like their culture being ruined for advertisements and headlines
>>17052250hilarious seeing ai dismantle the ego of pseuds and academics as they come to terms with their wasted youth. looksmaxxing and soft skills maxxing was always the way to go. hopefully it'll teach them to be more humble and kind towards each other but i know how insecure "intellectual" types are since they have nothing else going for them. they'll just come up with more ways to cope
>>17052279>hold back societyHow has AI done anything but regress society ? LLM usage literally and provably destroys your brain, there is no positive outcome for anyone except the epstein class.
>>17052244>However, the push by AI companies to solve mathematical problems as a benchmark is detrimental to the science of mathematics, and to the mathematical community.>detrimental to the mathematical communityJesus christ, I feel second-hand embarrassment for them. Imagine thinking putting this out makes you look good.
>>17052244>hey chatgpt how many black field medalists signed the declaration on Math and AIWhy isn't math listening to the black voices /sci/
>>17052284epstein did nothing wrong
>>17052285human culture is the vehicle of progress. the chain of scientific and mathematical culture has brought civilization to where it is today. human success comes from both discovering and also communicating ideas to others in order to train the future generation to follow in our footsteps. severing the chain of mathematical culture for advertising clout is a net negative for humanity.
>>17052244More and more I'm starting to realise that "humans understanding deeper and deeper mathematics" is a fools errand. Mathematics can get infinitely complex and obtuse and waiting for humans to catch up and slowly digest it is a waste of time. The future is to give in completely to the machine god, let the machine god understand deep mathematics and let the machine god progress it. Humans merely apply the results to the real world if possible. t. former pure math PhD student
>>17052289This is outdated. Mexico and US are based on the state. Japan isn’t 13 anymore. Saudi Arabia, Yemen, and Oman are based on marriage, and a lot of Europe has Romeo and Juliet laws. China is surprisingly still 14.
>>17052292sounds like a whole lot of cope to me. i'd probably tell myself similar things too though if my whole sense of self-worth was tied to making mathematical discoveries that are now being made trivial by machines. all sorts of laborers and artists went through similar phases of cope in the past decades.
>>17052297>Mexico and US are based on the stateI don't know about Mexico, but the US state laws are irrelevant unless you plan on never crossing state lines. Just see what they did to matt gaetz for having sex with a 17 year old.
>>17052300>don't worry about hiring new employees. AI can replace them>don't worry about supporting/training artists. AI can replace them>don't worry about math. AI will become smart enough to replace them>lol let's just bet all of civilization against AI becoming intelligent fast enough to catch us when the bottom falls out
>>17052292>train the future generation>mathematical culture lol tf is mathematical culture and how do mathematicians ever train anybody beyond whats essentially the method of revelation
>>17052306>on science and math board>doesn't understand science or mathnot worth replying to you anymore
>>17052309uh oh meltdown. youre talking to a different person
>>17052244yes yesHOWEVER
>>17052244oh that surely changes things
>>17052284it wrecks havoc in normie brains because they offload retarded shit to it. those "studies" are no a blanket statement.
Why do a bunch of Fields Medalists care? They have tenure and are past their prime anyway.
Honestly, what did this nigga think would happen? How can you not extrapolate out? He was literally doing ads for OpenAI a few months agohttps://x.com/OpenAI/status/2060451757818601808
>>17052359>accepts openAI money >advertises for them>openAI uses your work>REEEEEEEEEEEEEEEE>openAI and its consequences for mankind....seethe_soyjak.png
>>17052251it did that last week
>>17052359he's obviously high iq, but people refuse to understand that being expert in 1 field doesn't make them an expert in every other fieldhe unironically thought ai will always be his dumb coding buddy, even oai ad money aside, because he simply is pretty shit at predicting the future ai progress, it's that simple
>>17052244Same as artist denounce ai art desu
>>17052374>Expecting some wisdom from a man with "Flatten the curve" "infographic"pathetic
>>17052262Human made art will always have people willing to pay for it, doesn't matter how much you shill and shit your pants. Not even an artist myself
>>17052408so how much you pay for the latest art piece you bought bro?
>>17052374Predicting the future requires a certain degree of creativity and imagination which is often difficult for autists
>>17052244What do they want to happen though? The only call to action in the letter is>These issues must be addressed urgently
>>17052435He wants guaranteed credit for mathematicians for work that AI does
>>17052374my future prediction is that chinese models are going to eat american companies alive, and we're all going to hell because we bet everything on this shit
>>17052451so he wants everyone whose work is included in the dataset for the LLM to be credited on any LLM-generated result?
>>17052459Yep. Universal Basic Credit.
>>17052455Chinese models just distill claude. They'll always be a step behind.
How did Tao become the only famous mathematician under 60? Scholze mogs him.
>2000 endorsersthis is just getting sad
>>17052468i'm endooooorsing
>>17052292Causality implies everything led to today. No matter what it was. Bring on the rape and cancer.
So what's the point? Why waste years of grad school working and a problem that could be solved by a few thousands of compute? Solved by a few bucks of compute a couple years down the line?
>>17052250>coders: submit to the machine god, preserve dignity by being humbleInsane cope. It's over buddy
>>17052359That’s false Tao literally said yesterday that he was tricked to do an ad for oai
>>17052482This is a good thing. Even pre-AI, tons of want to be mathematicians were convinced they're going to solve some great problem and become famous. Math has wasted many lives.
>>17052482The point is understanding now. There's no point if you care only about prestige or money in the field. Those who really wanna learn will still stay but others won't. We are more of students now than teachers.
>>17052244funny how they activated only after their usefulness started to decline
>>17052482not for the recognition that's for sure. that's goneyou are free to do it just to train your mind. no one is going to pay you for this though
>>17052496They still be paying them in academia since it's their job to keep their field running. But most won't find it outside the academia.
>>17052510it's a coin toss. math departments were being demolished even before the rise of ai. now any argument the math profs had against that is going to look very weakpasting a comment from reddit with a quote from this year's fields medalist:For context, here is the complete quote from mathematician Yu Deng to see the full picture: "Lately, the math community I've been in contact with has been permeated with a restless, anxious atmosphere... Regardless of what the future holds, this anxiety is really unnecessary. Since the timescale of AI iteration is visibly much shorter than that of human society, things will become clear once this current wavefront passes in a year or two. The mathematics community will naturally adapt over time, and a new ecosystem will gradually take shape—all of which will take far longer than just a year or two. For those who are anxious, lost, or even contemplating leaving the field, waiting a year or two before making a decision won't be too late (of course, to those trying to cash in on the hype and harvest quick money during these couple of years... good luck ). I am also genuinely curious about where the ceiling for AI actually lies. For any given PDE, you can theoretically pose an infinite number of questions. The existence of singularities is merely the most fundamental one—what about their measure in the space of all solutions? What about their (co)dimensions? Topological properties? Complete classification? If you tell me AGI can solve all of this: fine, witnessing a divine miracle in my lifetime means I haven't lost out either. At that point, all that's left for me to do is just go home and focus on writing my Yuri novels ."
>>17052247I'll just post all my memes
>>17052244>buckbroken manifestoLMAO. Le heckin I love le science crowd now want to slow down progress because they want their name in an equation lol
>>17052250i think physicists are excited as fuck, the field has been pretty dull for a whole 15 years and this could open an explosion of theoretical research
>>17052270>These level of delusionif terry tao and co had to manage a company it would go bankrupt in a year. Also the frontier models are already training themselves, so human competition is pointless.It is incredible how inflated your ego is when you don't understand exponential growth LMAO
>>17052284smart people are smarter with LLMs, you are probably just retarded
Honestly I just gave up and just watch tv in my underwear and even I am not as buckbroken and pathetic as the faggots who wrote this shit.I gave up honorably.I am a samurai
>>17052455this scenario is possible if progress hits a wall, but...>there is no sign of slow down>models become cheaper with time, Astra in 2 years will make you videogames from a prompt for 30$So these companies will be fine either way, their addressable market is too big and the economics of the business are very fucking good
>>17052482stop doing problems and focus on pushing theory
>>17052556lol. If you asummes the chinese won't be throwing enough compute. The big labs want to keep the models close and only very few enterprise users pay up majority of their bill. If open source models become as competitive as the closed source ones why would those enterprises would pay for you when they can run their own secure private models in rented cloud or compute infrastructure? It's price war that's keeping this companies from reaching profitability. They need to charge more to make sense. Making them ever cheaper harms their own business.
>>17052459>read math paper>extend or make result on your own using said math paper>i now have to give full credit to the previous author and get none of my own
>>17052564>few enterprise users pay up majority of their billthey don't want to pay frontier models, but the previous gen is cheap as fuck. As I said the business is more solid than people think. Using the equivalent of GPT5 is cheaper than frontier open weght models, when these labs stop pushing forward they will be cash cows
>>17052564And close by I mean they want govt to get involved and regulate the market for them. If there's no competition the regular users would barred from using latest models in the future.
Yep, it's time to start demolishing math department buildings and start building data centers in their place instead
>>17052568The definitely do lol. Keeping the same structure forever isn't sustainable for them.
>>17052577you don't understand how quickly frontier models become affordable, in 2 years you will be able to create videogames or an anime show from your laptop for 20$ a month and open source models usually specialize on less sophisticated tasks. These companies are designing their own data centers without the tech restrictions that chinese companies have, the optimatization gains that OPENAI is getting are absurd
>>17052584As I said again majority of the users aren't going to make up their necessary revenue. This price war will continue to point they're free. That's good for you but not for them.>without the tech restrictions that chinese companies haveInsane cope. You're having all sort of anti-ai and data center movement in us while chinese govt can just demolish all those restrictions and do unethical stuffs.
>>17052590>aren't going to make up their necessary revenue.People spend like 300$ a month already on Shitflix, HBO Max, Hulu, videogames, amazon prime and the like. All of that can be bundled up to your taste with an AI>But it is too expensiveLook what GPT4 costed and how much it costs to run now. You don't get the economics nor the technology
>>17052607You are using circular reasoning. What I said is that as the cost per compute gets cheaper it reaches the point where running previous models costs almost nothing. A subscription strategy is unsuitable for pure lab companies and they won't survive more flexible rivals who will eat up their revenue. They will never become profitable only coping idiots like you believe that. It's simple free market survival. They are extremely narrow sighted which is why they are so keen on pushing regulations.
>>17052491woah it's exactly the same shit as with artists! almost like everybody bitching about ai only cares about his job and nothing else
>>17052513>go home and focus on writing my Yuri novels Based Fields Medallist
>>17052637AI already automated him in that sector
>>17052482I think you should read this quanta article published a year agohttps://www.quantamagazine.org/mathematical-beauty-truth-and-proof-in-the-age-of-ai-20250430/
>>17052554you'll never be a japanese
>>17052244Comes across as pretty whiny. Like they're made that no funding went to them, and instead it was all contained within a corporation.It was still mathematicians directing the LLM, they were just employees of OpenAI instead of in academia. That must be infuriating if you're in academia.
>>17052244I actually read this entire thing and it's literally just>stop taking our jobslike they don't raise a single actual point regarding why it's bad for AI to take over the field, it's just whining>b-but if AI does everything then what are we going to sit around and circlejerk over???This isn't like a creative field where people can create infinite works. There is a finite amount of math to "solve" and these dummies already had 5,000 years to figure it out. Sorry, you were too slow, your profession is now obsolete.
>>17052628>the cost per computethe cost per compute is the same, you just need less compute because the algorithms are optimized, and the open source models are good at specific things (coding, image generation, etc) but they lag behing in general cognition and the big AI models are 1 pit stop with a massive user base and incredibly optimized algorithms. You are really not getting anything lol.
>>17052649Retard I meant LLM compute let me rephrase cost per token. As for the fact that nothing is stopping the chinks from creating an equivalent model even without access to more advanced chips they can simply brute force training with a cheaper energy budget. And contrary to you they are definitely better at optimisation other wise they couldn't have caused the sudden price slash among other companies. That's really all that matters especially now that their government has gotten involved.
>>17052276Clout isn't really it. If we're talking about OpenAI, this is really Altman who had some success with LinkedIn, using his reputation with technophilic hype to get investors to give him money. Investors certainly are not rational people. They have all sorts of superstitions and often fall for hype. If they were persuaded somehow that they're going to make a return, they will go for it whether or not it will really happen.For Nvidia, it's about selling GPUs and LLM companies burn through GPUs really fast. Bottom line goes up so the investors throw more money at Nvidia. But they burn them up so fast that GPU and RAM prices skyrocket. Investors only care about numbers going up. They support all sorts of degenerate things as long as the line goes up. Jensen like the good goy CEO wants his big money so he pushes this to make the line go up.Meanwhile OpenAI has only proven to be good at turning billions of dollars into millions of dollars. If they keep it up they lose the investors and go out of business. Jensen's line goes down because he loses a significant portion of the market. All this economy around data centers takes a huge hit. OpenAI seems to be doing the Pure Math stuff because it seems like the model could possibly generate a paper with a results and if they get one they can defer the collapse a bit. "OH LOOK AT OUR BIG RESULT. DON'T DROP US PLEASE. GIB MORE MONEY." To try to get the investors to hold on a little longer. They want to keep the line going up until Altman and Jensen retire. Then it can finally collapse because it'll be someone else's problem.
>>17052376holy shit I remember that. god fucking damn she was an artist
>>17052486Lmao. How can you be tricked into signing documents from openai to advertise for them? Pray tell, what are the specifics? Oh right they don't exist because Tao was a greedy faggot who took AI money, then cried when it had strings attached.
>>17052359Legit, I lost a lot of respect for Tao. He now pretends he never was an AI shill.
>>17052642>quanta
>>17052681doesn't he already get paid like 500k a year?
>>17052282It's like a little kid wanting a lollipop for capital gains over competitive wankery. They never won.
>>17052244this guy is over 50 btw. he looks 16
>>17053257No he doesn’t
