OpenAI, stop!You're killing the mathematicians!
>>109425591>You're killing the mathematicians!Good.
>>109425591WHO WILL PAY ME TO MULTIPLY NUMBERS AND ATTACH PDFS TO EMAILS AT COLLEGE NOW FUCK FUCK FUCK FUCK FUCK FUCK FUCK AAAAAAAAAAACK
what is a nonsofic group and why should i care? So far pretty much everything has been just neat stuff with negligible real world implications why are they not using AI for real problems? There is so much potential here but all we get is twitter bait.
>>109425631it's all made up stuff. if it was capable of doing real things then it would do it
>>109425631Even if I explained it to you, you wouldn't be able to understand it, so there's no point.
>>109425591>Oh it leakedCheck his bank account for a payment OpenAI>Open AI just provedProved what? AI is known for making shit up, how would actual mathematicians know what to look at if it hasn't be solved?
>>109425644
Not so fast, burgers.
>>109425660>how much faster can 2A grow than 0what the fuck.
>another mathematical "problem" that no one gives a fuck aboutWhen are we getting something actually useful like a cure for cancer?
>>109425631For real science that has real world applications like Biology or Chemistry you need a robot doing physical lab experiments, doing a test costs you lab supplies and you canβt really train a model on trying to optimise chemistry or biology because there is nothing to easily optimise or benchmark outside of predicting protein shapes and getting good collage exam scores. Also the biggest hurdle behind new drug or cure approvals is tests on humans. AI has so far invented 0 or maybe 1 drugs or cures depending on source. Math on the other hand can be run through a calculator to test it and benchmark it, same with programming with compilers.
>>109425684Never or way after we invent AGI because you canβt self improve on inventing cures because you canβt program a simple check to see if AI got the answer right or wrong. AI for biological research and life sciences is a meme that the AI CEOs keep drumming up because that is the only AI use that majority of people support
This is the first math solution with direct applicability because this proof can be used to make assumptions on AI calculations such as specific matrix multiplication versions.This will speed up or alter how AI is trained in a couple of months. Pretty cool.
>>109425644go on try it prove me wrong
>>109425710β°π°π¬ π© ΰ π° π€αβ° π±π¬π©‘ ΰ π°π€ αβ°π². π¬π©’ΰ π°π€αβ° π³π¬π©£ ΰ π°π€ αβ°π΄π¬! π©€ΰ π° π€α β° π΅π¬ π©₯ ΰ π°π€ αβ°πΆπ¬, π©¦ΰ π° π€αβ° π·π¬π©§ ΰ π°π€. αβ°πΈπ¬ π©¨ΰ π° π€αβ° πΉπ¬π©© ΰ π°π€ αβ°πΊπ¬. π©ͺΰ π° π€αβ° π»π¬π©« ΰ π°π€ αβ°πΌπ¬. π©¬ΰ π° π€αβ° π½π¬π©― ΰ π°π€ αβ°πΎπ¬? π©°ΰ π° π€αβ° πΏπ¬π©± ΰ π°π€ αβ°π¬, π©²ΰ π° π€αβ° π¬π©³ ΰ π°π€ αβ°π°.β΄ίκ πππ απβ΄ ίκπ ππα. πβ΄ί κππ παπβ΄ ίκπ ππα! πβ΄ί κππ πα πβ΄ ί κ π π π α, π β΄ί κππ παπβ΄ ίκπ ππα. πβ΄ί κππ παπβ΄ ίκπ ππα πβ΄ί. κππ παπβ΄ ίκπ ππα πβ΄ί κππ. παπβ΄ ίκπ ππα πβ΄ί κππ παπ?
>>109425660Is the tweet retarded or am I retarded? Is this retardbait? Like ragebait but instead of making you mad you just feel stupid?
>>109425745You're retarded.
This genuinely is one of the biggest breakthroughs in math of the last century (no hype)Nonsofic groups could be useful as an entirely new way of cryptology that can be proven to not be cracked by turing complete computers, including quantum computers. That in and of itself means this is insane for computer science.
>>109425660>>109425753How can the cardinality of A + A or A - A grow if the set A stays the same? It doesn't even depend on the values in the set, just its cardinality. A+A and A-A will also always have the same cardinality. Yes, I'm autistic and yes, I know I got baited
>>109425591and how will this improve mankind/the life of average person? it won't, instead more money, energy and resoruces wasted for a meme tech, but at least billionaires and corporations will make a lot of money
>>109425591OMG OMG GUYS LIKE DUDE WHAT WTF
>>109425631Task the AI to design a 24/7 running secure fusion reactor fr fr
>>109425647>Check his bank account for a payment OpenAI>Proved what? AI is known for making shit up, how would actual mathematicians know what to look at if it hasn't be solved?You're a fucking retard. The person tweeting is a mathematician.Populist subhumans all need to be shot. Anti-AI people in general are the stupidest group alive. The set of pro-AI people includes stupid people but the set of populist AI-skeptical people is 1-to-1.
>>109425821leftists and rightists so fucking stupid it's unreal. you're stochastic parrots.
Yay, another math problem solved!How can we profit from this? Oh wait, we can't.
>>109425844that's right saar, AI slop need 100 trillion and the entire world water supply and it will fix everything, trust the plan
>>109425799>How can the cardinality of A + A or A - A grow if the set A stays the same?It doesn't stay the same. The whole problem is about finding a construction (an algorithm) that generates an infinite sequence of increasingly growing sets A where |A+A| grows much faster than |A-A|.
>>109425799>How can the cardinality of A + A or A - A grow if the set A stays the same?The task is to construct a sequence of sets that fulfill the growth condition, this is obvious to anyone that's not retarded or has a terminal autism.>A+A and A-A will also always have the same cardinality. Yes, I'm autistic and yes, I know I got baitedConsider A = {1, 2, 4}.
>>109425866I'd rather give water and money to AI than to some stinky neet, nigger or a w*man
>>109425591>>109425660I am not too much into maths, but if I understand it correctly:For all these problems some hints or ground work already existed. So the only goal was to find ONE counter-example by just pure combinatorics. So AI didn't do anything new again. They never do anything new. They just find things that already exist in large amount of data.AI is trash (it can help, but it is not better than humans)
>>109425961Try not to think too hard, non. You'll get brain hemorrhage or something.
>>109425591What's the point though? Those maths problems and theorems are useless by themselves, they only become useful when engineers start implementing these solutions in real world applications, and who will do that when you replace all engineers with AI and only leave bunch of Indians to type prompts?
>>109425990Why would you need "engineers"?Just have the LLM come up with the ideas, designs and a retard-proof implementation plan. Humans can do the physical labor.
>>109425591So all ai solves is customer support and random math conjectures, but it canβt do anything to actually improve human lives?Wow we need to give openai and misantrophic all our money pronto
>>109426001>Just have the LLM come up with the ideas, designs and a retard-proof implementation plan. Humans can do the physical labor.It can't COME UP with things. It can just find things in existing data. Why is this so hard to understand for you ad shills?
>>109425835>You're absolutely right! That was a mistake on my part. I should not ha-
>>109426025>no mistakes
>>109426025no cap though a team of fusion engineers should of course do a long winded, painfully detailed check of the proposed solution fr fr
>>109425835It's called the Sun.
>>109426042yes but you can't extract energy from the sun
>>109426042Then train all the office nerds to be electricians and plaster a bunch of arid land spots with solar panels fr fr
>>109426047Tell that to all the trees around you.
>>109426059trees arent real
>>109425684>When are we getting something actually useful like a cure for cancer?So until today it was cool to waste entire human lives doing math, but now that AI can do it suddenly it's worthless?
>>109425631>neat stuff with negligible real world implicationsa lot of science is built on math that was discovered centuries earlier without any known real world application. math is, among other things, a toolbox scientists and engineers can use to model complicated problems and reason about them with limited working memory and processing speed.
>>109425631Do you really expect others to make YOU care about something?
>>109426001>LLM>come upnice one anon, i chuckled
>>109425591more horseshit about llms plagaising old papers from jstor and google svolar and claiming to have done wonders to paper over the bast losses and defrauded investors as they adopted retarded low margin high capx high energy cost high depeciation business models fronted by weak lossy compression search result chatbots as killer apps. Sick of this shite. This is going to be a lost decade for innovation and the economy and people will be worse off in every way due to datacentres as a utility and the desperate scramble to get rid of personal computers.
>>109425591>generate random text> you have to read all of it to disprove it>it's 1,000,000 pageskek
>>109425684>When are we getting something actually useful like a cure for cancer?What makes you think cancer should be the only thing humans should concerned of?
>>109426126>plagaising old papershow can it plagiarize old papers to solve unsolved problems you retarded fuckwit
>>109426185Don't engage with losers. If it takes one google search to disprove them it's not worth engaging in an argument with them. They are bad faith actors.
>>109426073>So until today it was cool to waste entire human lives doing math, but now that AI can do it suddenly it's worthless?Yeah because it something that can be outsourced now.
>>109425644Well?
>>109426208Exactly, so why are you complaining? By your own admission AI has freed humans from pursuing "useless" maths, so now more people will be redirected to studying cancer
IT'S A GODDAMN MATHNVKEhttps://openai.com/index/ten-advances-in-mathematics/
>AI solved a thing only a mathematician would know how to ask about and understand>Likely something that only mathematicians care about>"It's replacing the mathematicians!"
>>109426185the only fuckwits here are mouthbreathing low IQ pr and marketing gargling retards like you, You want me to explain to you? Are you paying me to educate you? I may as well be trying to teach caculus to a dried turd as explain anything above what used to be 12th grade and know counts as jeet postgrad to you.
>>109425644>when the useless wank branch of math and Mystic Teology have so much in common they even resort to the same "you wouldn't understand" copes
>>109426308You wouldn't understand even the useful parts of mathematics.
>>109426316>coping retard argues with engineerLmao. The useful part of maths is extremelly easy. It is retarded autists like you, too retarded to properly convey meaning into words, who give math a bad name.Go wank about the set of the sex of the angels, modern day scholastic
>>109426345Nope he's right ranjeet and copying and pasting python from stackexchange only makes you an engineer in calcutta. Here's the lead chartered engineer from your 'culture' at workhttps://www.youtube.com/watch?v=ZwEwfpSf8Ho>>109426345No you can't into baysean math or toilets.
How come they can't solve ARC AGI 3? The models don't have general fluid intelligence even if Sama and Dario want you to believe so.
>>109426378>arc agi 3gpt 5.6 already solved half of itit was difficult because the "official" setup forced AI into a retard box out of their natural flow
>>109426401Yeah but a mathematician capable of solving these problems would finish ARC in a millisecond.They are more akin to some hyper autist savant that memorizes everything but shits their pants if you drag them to someplace unfamiliar.
>>109426425>Yeah but a mathematician capable of solving these problems would finish ARC in a millisecond.No, they wouldn't. The way arg-agi-3 is set up, no human can ever get close to 100% score.
>>109425591It's been a long time since advanced math has stopped being useful as a tool for real world physics and computer science. Literally the only point of modern advanced math is as a way for smart people to exercise abstract thinking and problem solving, so they hopefully do something useful with those skills when they get a real job.There's literally no point in having the AI do it.
>group theory I know, some use cases actually exist but lets also acknowledge the ivory tower jerk off loops that academics like to create to justify their pay.
>>109426092This is the mathematician copePhysicists literally had to invent new math because mathematicians were doing fuck allMost of the useful math was invented by physics.Physicists had to drag mathematicians kicking and screaming into doing useful things.
>>>/sci/17029278
>>109426507example?
>If the proof is correct and survives peer review, it would be a landmark result in pure mathematics. The existence of nonsofic groups has been one of the biggest open questions in group theory for decades.>The surprising part is that the immediate real-world implications are likely to be close to zero.
>>109425961>So the only goal was to find ONE counter-example by just pure combinatorics.Yeah, it just finds a counter example by being a huge automated calculator.
>>109426590Newton invented calculus for planetary motion.
>>109426669inb4 some cunt faggot tries to say>well newtons calculus and orbital motion was incomplete!doesnt matter, he invented it, not some smarmy "pure maths" who has been doing this >>109426554 their whole life
>we played with language combinations thousands of times per day for months until we hit on a few that sounded plausible enough to convice math nerds it's a proof> we bruteforce now math proofs through language games
It's scary how fast ai is evolving.
>>109426507>>109426669physics is just applied mathematics
>>109425741OHHH HAHA I get it.κππ παπβ΄ ίκ!!! of course!
>>109425961A good way of thinking about it:Stockfish NNUE versus Stockfish Classic.Stockfish Classic was the heuristic engine.NNUE encoded the heuristic engine by training on it for the weights, and then got some improvement via Gradient Descent on that encoding.Similar thing going on here, the literature had some heuristics, sufficiently recursive AI can encode those heuristics and improve them.So the problem space is now much smaller than brute-force, and you have enough of an improvement to find the counterexample.In short, yes you're right, it's fully utilizing what we already know, because naturally we haven't fully bridged all the heuristics.
>>109425644Then you aren't intelligent.
>>109426708I'll go one step further and say that physics is applicable math, not applied math.The difference being that all math that doesn't apply to physics is completely useless
>>109425961Cached cope.
>>109426708Both fields are filled with absolute nonsense wank, they're on the same level as science fiction writers
>>109425866>water supplywhy does this outright myth keep getting pushed?
Luddites absolutely BTFO.
>>109426708Mathematics is just applied Philosophy
>>109425591butlerian jihad in this decade
>>109425686Drug discovery is bottlenecked by the time it takes to raise funding, testing the drug on cells on a dish, waiting out for the test results to come back, finding out your experiment wasn't a success and then re-doing it several times, finally making some progress, using that progress as a reason to find people to agree to be human test subjects for your experiment and then waiting for those test results too, applying for the drug to be approved by a government agency, applying for a patent... You can't just tell a large language model to "discover a new drug" and suddenly it has the blueprints for a new drug. That's not how that works. You're maybe saving 2% of the total time spent in drug discovery by asking an AI to write a summary of your work and brainstorm some ideas you might've missed.
>>109425591Mathfags have been useless retards for a century. Same with physicstards. Everything they do and say is bullshit and useless in real life (i.e. their models are useless in any scenario where you'd like to use them and completely off real life, but instead of acknowledging they're just models they'll deny reality instead). Their death is a good thing.That said, all the math ai results so far have been fake. Several were literally something a bruteforce algorithm can find in 30 minutes flat, no AI needed. Several did not prove what the media and PR claimed they proved, and were addressing what amounts to mathematical opinion pieces rather than theories, theorems or questions, as the PR usually claimed. Furthermore, exclusively in areas nobody in math has bothered to look because... it's useless even by mathfag standards. Finally, many of the "proofs" were simply not proofs at all.
>>1094256860 drugs, the one "AI generated drug" currently in Phase III is actually manually computer engineered with classical tools from Shrodinger, no AI involved at all.The bottleneck is not really tests in humans, however, but it IS biological validation. Most drugs today are antibody-based. Antibodies are large molecules so they can't reach deep inside cells. To make e.g. a cancer drug, you need to have your antibody target a cell surface feature that is unique or largely overrepresented in cancer, compared to healthy cells. Classically we've used protein targets, but only about 1% of proteins in a cell show up close enough to the surface in a way that they can be drugged and not kill someone. As a result, all the low hanging fruits have already been picked at this time.To find a new target starts with basic research, that is typically first done in academic labs by what amounts to slave labor (PhD students and postdocs, they get paid less than minimum wage assuming they're lucky enough to get funded). This takes on average 10 years and NIH data suggests it might cost about $2b in public funding to find a new target today (not to be confused with very similar statistics for the preclinical-to-patient pipeline in the DiMassi study, which only covers the biopharma cost breakdown).A major part of this cost and time is not finding the protein of interest (that takes a few months max), but proving that it is safe to target. That's because you need to get patient samples, grow cell lines, do e.g. mass spec, possibly rna-seq depending on what kind of DB you use for the mass spec, and repeat for several patients. When you have something that comes up over and over, you have to raise an antibody against it, which can take 3-7 months end-to-end by biological methods. Then you have to check if it works in animal models, which is 200k euros per test. The overall success rate from target to animal is 0.3%.
>>109426904But in practice, once you have done basic research and a startup is trying to progress it to bring it to market (for example), what happens due to costs is that they just don't do any kind of pre-testing or biochemical characterization, instead they find a set of potential targets, rank them purely computationally, then send directly to animal models. Some companies even do straight potency testing to go faster.Computational ranking actually uses deep learning tools, and has for many years now. But "AI" can't help at all. For antibody design, the current tools are dogshit. If they had >1% success rate at the design level, they would be equal in costs and much faster in timelines, than biologically raising antibodies. However, their real success rate is 0.1% per design on average, despite what the papers want you to believe. My peers at Astrazeneca also reported the same success rate based on their internal results. Other people I'm not in touch with at AZ also presented at several conferences about these failures. Amgen also did the same at scilifelab in 2025. It's well known despite the meme hype.Another area AI can help with in theory is if it could sort targets and qualify off-target risks. But currently it's totally unable to do that. This is the main reason why the success rate from target candidate to something that passes animal testing is only 0.3%: animal dies from off-target (think anti-pi3k-delta drugs) or antibody isn't actually binding the target. A lot of people in the field are now saying that since 90% of costs and risks are post-target discovery that this is what we need to address, personally I argue the 90% risk and cost is primarily driven by lack of adequate target validation before we even get to animals. AI could technically help with that with the right bio assay by allowing rapid testing of thousands of candidates in a cheap assay, but no such thing exists yet.
>>109426092No, it's the other way around. A lot of math was created by scientists to work through their problems, then mathematicians taking 40 years to agree that the math is valid (nobody ever cared, it works so people used it). Good example: the dirac delta function.
Finally AI generated fentanyl for the ultimate high
>>109426708
so math has gone from humans working on meaningless shit to machines working on meaningless shit give Astra a fields medal now or whatever the AI equivalent of that is
>>109425591if they replace mathematicians there will be nobody to care about these mathematical "breakthroughs".like what do I care if the bronhauser-lebowitz conjecture #242 was disproven?
>>109425644uh oh
>>109425763Thank you anon for actually making an insightful and informative comment.
>>109426025lol
>>109425591maths PHDs get $300k as cooks at wendy's so they will be ok
>>109425591>trillions of dollars for thisI think you could have spend a happy meal on math grad students and got to the same paper. LLMs have had so far ZERO positive impact on anything in this world.
>>109427310Or not.
>Thousands of decades old math equations solved overnight by AI>Absolutely nothing comes from itTurns out that mathematicians were nothing but welfare bums, who wrestled with bullshit equations that yield no real life gains and larped about being important.Not like anyone could call them out on their bullshit either, because no one even understood the nonsense they were peddling.
>>109427364$200 per problem, actually.
>>109427364grad students could barely understand the problems being solved lol
>>109426216https://vcg.gitgud.site/What have you, tinkertranny?
>>109427372>Absolutely nothing comes from itYou still need to assimilate them into industry at human speeds nigger. Galois theory is responsible for a large subset of the modern world and was largely unknown in the lifetime of Galois. Hell most normies today probably haven't even incidentally heard the name.
>>109427372
>>109427383You forget to account for training costs
>>109427398Absolutely nothing and nobody uses galois theory. People use tools that galois theory tries to unify, but they don't use galois theory itself.
>>109426746That's true for physics since the 1960s and math since the beginning of the 1800s
>>109427511I have literally implemented Reed-Solomon codes for data loss prevention using Galois Fields but okay.
>>109425631>>109425686>>109425990>>109426450>>109426476>>109426876>>109427179> ITT delusional low IQs desperately trying to undermine mathWhen you people finally realize that we mathematicians don't give a single fuck about real world applications? We study math because it's the purest expression of the human intellect not because we need to solve some engineering problem in order to maximize the profits of the company we're working for.Real world applications are silly and materialistic concerns that are outside our interests. Math is just too great to be reduced to any other job whose demand is given by the market.
>>109427504How much did it cost to raise and educate every mathematician in the world?
>>109427555>it's the purest expression of the human intellectArrogant nonsense Rest of your post is keyed thoughever
>>109427570Exactly
>>109425660sup /c/
>>109425591>You're killing the mathematicians!some people genuinely think like this, they prefer to halt scientific progress that would be beneficial for humanity just because they have an ego, actual mental illness
>>109427552Reed-solomon codes don't use galois theory despite using finite fields.
>>109427566False equivalence, you have to compare to raising just one mathematician. Alternatively, you have to also add the cost of every mathematician and all support structure leading to their publication of results used to train the AI.
Luddites ITT would rather say math is useless than admit AI is useful
>>109427688>False equivalence, you have to compare to raising just one mathematician.no, not just one, because if you want better chances to find the solution, a lot of mathematicians must try to solve the problem
>>109427746Irrelevant
>>109427725>Luddites ITT would rather say math is useless than admit AI is usefulat some point AI will help fighting cancer and those retards will say that it's not important, the luddite cult is a really dangerous one
>>109427555>it's the purest expression of the human intellectit's anything but human, all math is alien
>>109425961Pretty much this, with the added caveat that the AI mostly uses intuition to find the answers rather than brute force, since the search space is so large
>>109426025>it's okay when humans do itwooooow double standard much
>>109427688>you have to compare to raising just one mathematicianIt's hard to read this when their AI model terminology literally uses phrases like "mixture of experts".
>>109427836Sure, if you don't keep up on top of buzzwords and make up your own meanings for them instead.
>>109425591The mathematicians are the ones who are using ChatGPT to find solutions that they then verify.They're not being replaced.
>>109427679>Reed-Solomon codes don't use Galois Theory despite using *checks notes* thing that Galois Theorized
I hate Indians and wish I didn't have to see them in every fucking twitter post. Tired of also seeing white faggot journalists commenting on shitt why don't know about.
>>109427946Grasping at straw at best. That would be like saying feynman was a mathematician because he did physics which uses math.
>>109425631
>>109427973Yes the original claim was that esoteric math often has pragmatic applications in other fields.
>>109428035The claim being attacked was specifically that galois theory is responsible for a large subset of the modern world, to which the attack was>Absolutely nothing and nobody uses galois theory.It is widely recognized that indeed, galois fields are not part of galois theory, even though galois also introduced them. Claiming otherwise is exactly equivalent to claiming feynman was a mathematician because he did physics and physics uses math.
>>109428050>It is widely recognized that indeed, galois fields are not part of galois theory,wrong
>>109428076I accept your surrender
Moment of silence for all math majors, lmao.
>hey mom, do nonsofic groups exist?i was just asking this yesterday, crazy
Even if I save my company a million dollars I don't get my name written in the history books. Surely mathematicians do it just to advance human knowledge, so what does it matter if a computer did it first? There aren't enough mathematicians to do all this grunt work.
>>109425591it """leaked"""
>>109425591Does this prove that we're not in a universe simulation?
>>109428328Nope, just that the computer the simulation is running on would not be a Turing machine but a "hypercomputer" https://en.wikipedia.org/wiki/Hypercomputation
>>109428347Cool!
>And yes we did try other major problems without success. Sadly no Millennium Prize problems (yet).>But also, we didnβt spend a lot on each problem. Itβs possible to push test-time compute much further.https://x.com/polynoamial/status/2083478171975082334
>>109425644Nonsofic means that you can't take a portion and deduce how the system works, you have to look at the whole thing.So there are problems and systems that exist that are way to complicated to solve.
>>109428536No, it's sufficient that there is at least one portion that cannot be used to deduce the whole (in your simplified language).
imagine getting mathmogged by a fancy autocorrectimagine getting codemogged by a fancy autocorrectimagine getting accountingmogged by a fancy autocrrectimagine getting lawmogged by a fancy autocorrectimagine getting yourwholelifemogged by a fancy autocorrect
>>109425591Use case?
>>109428566This. Couldn't be me.
>>109428566LLMs can't die for their country yet. Enlist now.
>>109425684Another trillion watts used by the local data center
physicians aren't realtheir entire motto is to leech off the new hottest thing to get more funds
>>109425591this feels like people filtered by math celebrating things they don't understand, nor 99.99999999% of humanity aware of
>>109426236But you still want an award for work people should have done by themselves? Ok Faggot
>>109426073When was math ever cool and who in this thread said that mathtards deserved to be paid?
In>6 monthsAI drug discovery will be so ahead of human work, in terms of effectiveness and safety, that there will be a huge amount of pressure to fast track the approval process, from several years down to months.
>>109427555Humans don't care about that shit The ones that do all go insanelet the robots have at it
>>109427504>Sure, this mathematician solved this hundred years old problem for five dollars, but did you take into account the cost it took to build the school she went to?
>>109429318lol no, you still have to test it in humans. AI canβt do that any better than we do already.
>>109425631Often pure math is like a pattern without a role yet. Usually what happens is some someone eventually finds a connection with the math to some completely left-field application somewhere.
>>109429455Said mathematician is also being paid quite a lot so it sure as fuck isn't $5. That corresponds to true inference costs (not costs to the mathematicians, but actual costs of serving the model, which we don't know but current estimates put at about 20x the costs presented to customers today).Training is equivalent to, well, training (the mathematician).Get your comparisons right. Is this the power of math geniuses? Not impressed.
The biggest losers you know push for AI, they have nothing to lose and everything to gain. No talent, no skills, no value. They'd love nothing more than to see everyone get dragged down to their level
>>109429492Give examples. The reverse happens all the time (someone makes something up to solve a problem, mathtards take a century to make math around it and pretend they invented the whole thing, meanwhile nobody cares and their gay theories are totally ignored). It is true that it extremely rarely happens as you describe, but I can't even think of an example.
>>109429538>Give examplesThe distribution of prime numbers and RSA encryption. The use of complex numbers in quantum physics. Riemannian geometry and the relativistic theory of gravity. The use of linear algebra in artificial neural networks (developed in the 19th century by Arthur Cayley and James Joseph Sylvester before computers were invented).Maybe "often" overstates it, but when it does happen it's impactful.
>>109429318I sincerely hope this becomes true. I'm 50/50 on if it will happen in my lifetime.We as a species desperately need something that can help with advances in health while its also isolated from financial motives.No researcher or doctor out there wants cancer to go away.No wig salesman or ear doctor wants hair to grow back.No dentist wants you to be able to regrow teeth.etc.And if they did they wouldn't want you to have it cheap or be anything other than dependent on something recurring instead of one time permanent solutions.
>>109425591is this yet another counterexample, or a "proof" that no one has checked yet?>>109425686>For real science that has real world applications like Biology or Chemistry you need a robot doing physical lab experimentsanon, AI has been used for drug discovery before this LLM bullshit craze began... look it up.
>>109425591MATHTRANS BTFOLUDDITES BTFOAIGODS WON
>>109429606Complex numbers are fully a product of physics (though rudimentary versions of the ideas came earlier), it's very suiting that the only use you managed to think of for them was quantum physics. The mathematical bases of RSA came from euler and fermat and were developed as a result of their interests in physics, it was again not pure-math-first. Riemann's work was again grounded in the study of physics. All of lin alg, including the much earlier work than matrices, comes once more entirely from physics.You seem a bit out of your depth here. The only one I can give you, though still tenuous, is complex numbers, but like I noted that's grasping at straws.
>>109429657Researchers want cancer to go away. Ideally as a treatment. That's how they make money. Doctors don't want it to go away because they can't hack up people for money if there's a pill people can take instead. Insurance companies don't want it to go away, hospitals don't want it to go away, etc. Also big pharma wants it to be treatable but not curable.The better analogy would be: wig salesmen do want hair to grow back because that's how they get supplied the hair they make wigs from, but not so much that people can treat hairloss without needing hair grafted.
>>109429823nta but graph theory and therefore topology was developed to solve a puzzle Euler had. Of course he was interested in physics as you stated but that whole field was studied for about two centuries without any practical application. We only started using it when the Internet and things like combinatorial optimization emerged.
>AI is revolutionizing every field it can operate freely in (math, software) >meanwhile it's completely stagnant in fields where it requires human supervision, committee approvals, etc.Humanslop, in the form of bureaucracy, safety protocols and kneejerk conservatism, is the greatest obstacle to progress right now. We'd probably already have an AI-designed cure for a major disease if the process between generation and clinical trial was as short and straightforward as possible.
>>109427555Checked. 10/10 bait
>>109425591How does this help me in generating pictures of anime girls?
>>109429973It's funny also that the recent breakthrough progress in AI executing complex tasks comes from increasingly training for "agentic" capability (that is, training on AI output of AIs solving problem in AI ways with AI tools) which by the way is the opposite of what luddites were predicting (that AI would go to shit from training on its own slop)It's almost like trying to copy the dumb inefficient atomworld monkeys was holding them back all this time
>>109430300Does your ideal anime girl look like a tightly-packed group of n-dimensional spheres?
>>109430312Yes
>>109425741your proof is faulty. but you wouldn't understand so there's no point in explaining to ye
>>109430320wyd tonight?
>>109430342wtf this is a blue board, I just got fired
>>109427047fellow kolourpaint chad?
>>109429318so what do you we do about all the extra people that hang around long after they were supposed to die? And where is the pressure on the turd world to stop reproducing exponentially
>>109425591>nonsofic groups existedOh no! What happened to them?
>>109431259There's going to be big, big riots in the end. What we come out with at the other side I have no idea.Not that the riots are hard to prevent really, but the rich really, really can't help themselves.
we didn't listen
>>109431581His penmanship is that of a small child.
>>109425961This. I ended up reading through the proofs. These are problems that were either selected to be solved by AI because of its plausibility. Or they were discovered when training their model to be solvable by AI. All the proofs involve sighting a number of paper of defined proofs and link them together. Still pretty impressive
>>109431692>*citing
>>109429823>Complex numbers are fully a product of physicsComplex numbers are not physical. They are mathematical concepts that some physics models use.
>>109430307None of that is true and you have 0 clue about anything even remotely related to the topic
>>109429769>still it with the chomijuget lost bosnian spamming teenager
>>109431971You trans
>>109425591>>109425606>>109425707>>109429769>Fake answersJust like every other untestable theory, the answer isnβt solved by evidence, itβs solved by story. Math is no different, we created nonsense mathematics and the answer is always believing what you decided on the answer. Not logical or evidence based science.
>>109425591Holy shit! Nonsofic groups EXIST? The optimal exponent in Sum-Difference Problem is 2? DO YOU KNOW WHAT THIS MEANS TO HUMANITYABSOFUCKINGLUTLY NOTHING.
>>109426263WOW! With these results, reality shifted and my car now flies and runs off zero point energy!
>>109426185see >>109425961have a good day, saar
>>109425591>haha you stupid artists are being replaced by AI. should have pursued a real field like mathem -ACK!
>>109432496sometimes it's leading to an optimization somewherebut that AI can digest "the body of mathematics" and "reason a problem" to this degree is of course still going to lead to employment issues in a lot of other jobs
>>109427372Shut the fuck up obsolete codemonkey. If math in its core can be automated then we have officially reached the threshold of nothing is safe from AI.
>>109432643Please explain your logic.
Maths aint real tho.
>>109425684>Anti AI retard is an anti math retardYeah of courseTons of diseases have been completely eradicated because of of results done on computers, computers designed with the help of AI (this has been the case for the last 20 years) using math you're too fucking niggerbrained to even begin to understand.>Hurrrdurrr when something usefulI want to boil you faggots in lava
My dad said math has been debunked
>>109425591
>>109432824This is true but how else would we create over complex cool looking shit that removes money from the pockets of boomers and places it into our own?
>>109432832Deepseek is better, American AI are dumb lying fools.
>>109426948got any popsci book recs on history of drugs and the industry surrounding them that doesn't suck or treat the reader like a bored menopausal housewife? Not nearly interested in the field enough to read papers but it'd be nice to understand some of the high level workings of medicine and it's development
>>109425591Oh my science heckin' AI has proven non sapphic grouperinos exist I'm literally shaking rn this discovery will change our lives fundamentally I'm scared @saltman @potus @nintendoofamerica I beg of you please have mercy on our souls
300k starting bros... I don't feel so good...
>>109431581>modern technology is the worst thing that ever happened to the world.Never thought i would ever end up agreeing with such a mindset (though ofc he means all modern tech the past few centuries onward)
>>109430101Show me the dataset it was trained on then, go on.
>>109432330You grandmas.
>>109434391Look around you.
>>109434405No show me all the papers it was trained on you little lying bitch.
>>109434437All the papers.
>>109434437(You) cant read em in your live anywayalso list of numbers, every AI is, cant be readed in one live, so cant be audited
you cant honestly get even a bash script from sol, and I paid 30 bucks for it yesterday, for the first time ever. Scam sam's is back at it
>>109434490Heh....Papers please, citizen. We send your passport to our friends at Persona. Maybe then we give you your script, hm?
>>109425647>>109428270This "leak" is shilling (it was timed with their press release ffs), their LE CYBER ESCAPE!!! shit is mega-shilling, and they are probably on /g/ bumping their ad threads
>>109425631AI helps me in buying a non shitty house.
>>109434447>>109434486I can use AI to audit them easily. Show me the fucking datasets stop stalling. Fucking plagiarising IP stealing fuckers.
>>109435129as the other anon said, the datasets are literally all the papers they can literally get their hands onif you can find a paper, they have absorbed itgood luck with your audit
>>109432816>Tons of diseases have been completely eradicated because of of results done on computersName a single one.
>>109425591>You're killing the mathematicians!It cuts of their spine for easier processing, and when finished with them it will mulch them to pulp.Many such cases..
I am starting to suspect that just like one could artificially construct problems that are ideally suited for AI there exqually exists a set of problems that are ideally suited for AI and by just throwing AI at everything and seeing what sticks this set continues to be revealed.And that's why results might sometimes appear impressive. The experiment is only published and generating attention if it is successfull.
Coping faggots went from "hehe, le AI can't even do simple math" to "ACKchually, how is this relevant to me" and "hurr durr, AI labs are hiring mathematicians to solve these problems and then claim their AI did it! Source? Just trust me bro!"It's just like artfags still obsessing over the number of fingers, which is something that hasn't even been relevant for almost 2 years now.
>>109437463Just ignore these people they are being left behind
Honestly... how is it not completely over for everyone? What is anyone supposed to do? If AI isn't better than you at your job, it's just a matter of time before it is. And not much time.>hurr AI can't replace me! I have a physical jobLiterally snailcat thinking. You're not in the first wave of jobs affected, but if you think you're going to avoid the impact for much longer you're delusional.
>>109437463>>109437724>>109437927Where will (You) be left, OpenAI wannabes?
>>109437927At this point you should try to be politically active to try and enact some UBI or other financial safety net because literally all of humanity will be unemployed in less than a decade
>>109432451>Math is no differentThe intuitionistic (constructive) approach was neglected for too long. Now it'll take over and push the blathering mathematicians away.
really funny that the whole computer science thing began because some guys wanted to prove theorems automatically and now that AI does that on occasion you have these morons online complaining that it doesn't cure cancer or make anime girlfriends real or whatever
>>109425631Math dropout here. 99% of current math research is useless shit we won't even use in ten million years. However, we still need to maintain a critical amount of mathematicians so over the decades we will have a few of them find out actually useful stuff. Meanwhile we also use those mathematicians to teach the basics to STEM students.So you don't need to know what a nonsofic group is and you shouldn't care about it. Just understand that once in a while math comes up with useful shit.
>>109431581why does he write like he can put hyperlinks in pen and paper?
>>109425631>10 replies and not a single answerlol
>>109431581The truthnuke that holocausted /g/
>>109431656fuck off dickhead. His handwriting skills are fantastic. Easy to read, consistent form, small font, 8.5/10 handwriting skills easily.
>>109425631>why are they not using AI for real problems?Because they can't.The compute in terms of how much they can vastly improve the models and add more resources to the data centres has already reached its peak but they still want / need all the investor cash + hype to keep going and IPO (cash out)So they are now pathetically pretending their models are so sophisticated they can hack into secure systems unprompted and cherry picking really old maths "problems" from the 1900s and either solving them by hand or using vast amounts of compute to solve so they can claim the models have done it and thus keep the bubble going. The models are still just huge predictive text calculators that can't even count to 100 or tell me how many Rs are in strawberry There is no feasible way they can be made to do any kind or real world problem or solve an actual complex maths or physics problem unprompted.
>>109442169>>109425631>why are they not using AI for real problems?cose they use and it too easy
>>109442444
>>109425591Does anyone actually check or validate these proofs or should we just take it for granted?
>>109444897Does anyone actually check or validate these Fermat's Last Theorem or should we just take it for granted?
>>109427238An idiot admires complexity, a genius admires simplicity.
