I find it interesting that, several years ago when discussing AI, folks used to focus on how automation would first come for blue collar work. Things like driving being automated and truck drivers replaced. And then eventually on the horizon we would see things like this.
But we still have no self driving cars. Seems like the automation came by in the opposite direction. Perhaps driving will actually be the final thing automated.
While I agree with your statement overall, as someone who drives everyday in LA alongside Waymos without anyone inside of them I must disagree with this aspect of it.
We still have no general self driving cars. What we have is a bunch of geofenced tests. While we are closer than ever, we are still seemingly a long way away from truck drivers being obsolete was foretold over 10 years ago now.
A better formulation would be: we don't know if we have self-driving cars.
Because if it became fashionable for pedestrians to wear large green dresses with red balls, and everybody would wear that, we have no idea what the AI would do.
This is a general point rather than an argument supporting the GP comment, but it can be shockingly hard to know what the correct response to particular situations is, or predict how a system will respond. For example, one system I know had difficulty differentiating people laying on the road from unmapped speedbumps (particularly at night), but was very good at recognizing them standing up. Someone spent a lot of time testing with dummies until perception improved. Another had no idea how to handle people in fursuits and halloween costumes, because it was incorrectly applying animal behavioral models to pedestrians.
That's why the latest self-driving systems include reasoning models that draw from the massive amount of knowledge embedded in LLMs. Language evolved to communicate things that are not here and now after all.
Something I'm noticing a lot is that automation finds easy footholds in unregulated industries because there's no basis to push back on it. And there's a lot more tolerance for poor quality work in unregulated industries too.
An LLM automating web developers with 99% success rate saves the company millions of dollars and allows them to launch products and features faster.
A truck or a car that has a 99% success rate would be the top threat to society and quickly rise to the leading cause of death under 30 if widely deployed.
An LLM automating web developers basically puts the company out of business. Why would customers pay a developer when they can just build it themselves?
There are a lot of things that people are capable of doing themselves that they pay other people to do for them.
I agree that LLMs have and will continue to increase the number of activities that people can DIY. But I believe many people will still choose to pay for others to do things on their behalf.
“99% success” is too vague to be especially meaningful in such different use cases.
99% successful web development where the 1% missing is security that and the company leaks its entire repository of user data every few months is not good.
99% successful car driving where 99% of 5-year periods have zero accidents per car is an order of magnitude safer than the average driver. (And gut feeling, I suspect approaching the tail-end of the safest drivers on the road.)
>99% successful web development where the 1% missing is security that and the company leaks its entire repository of user data every few months is not good.
It feels like it's happening every Thursday and nobody cares. They just give people credit watch coupons or what's it called, and move on with their business. Worst case, a few pennies in fines.
Well 99.99% is you have a major data breach every decade or so, but there are lots of companies, so on aggregate across all of them it's every Thursday.
But if CVS leaks your purchases on a weekly basis, it's probably going to affect people's willingness to shop there. (Or at least to use non-cash.)
Self driving in a winter storm? Self driving in torrential rain? Self driving when mud covers a sensor it needs? Self driving if it hits someone? Pretty sure Tesla would say you were driving it then.
Driving is heavily regulated, and there are a million interest groups lining up to sue to stop self-driving. Imagine all the people who make their money from car crashes, for example.
Broadly speaking, this is proof that there will be no jobs coming after AI.
All past inventions have put some of the most miserable humans out of a job. Farming by hand is hard, so is copying books and don't even get me started on the dullness of making clothes by hand.
Driving all day isn't all that far off, though still much better than the above and definitely better than riding horses all day.
Telegraph operators? Hell no. Dude sitting in an elevator all day? No thanks.
It's always been from the other end. This time around, there's gonna be some 10-20 years of robot manufacturing, then a universal lifetime of robots fighting misaligned robots until enough paperclips have been manufactured.
Or you know, we block out the sun and live in caves or whatever you fancy.
I see your point but if self driving cars existed, worked like advertised and were economically viable they would and should be in every country in the world.
Developers are guilty of making reasonable projections based on contemporary industry trends, and they're getting slapped by unexpected revolution in computer science.
Yeah, as always in AI hard things are easy and easy things are hard. How many times have I heard condescending comments from programmers that radiologists are just about to be replaced by AI. And well, AI turned out to get kind of okayish at programming much faster than being acceptable at being radiologist.
The inconvenient truth is the “desk jobs” involving a human operating a computer are far easier to automate than piloting a physical vehicle through a complex environment… A desk job is doesn’t require lidar and cameras…
It was too easy to assume that an AI would never approach human intelligence and hence would only be used for “boring” jobs like driving.
I’ve long disliked the hype around AI but now see that human intelligence is perhaps slightly overrated, at least in terms of the loftier positions seemingly requiring such.
Claude Opus puts to shame any highly caffeinated “10x” software developer. Not saying it’s perfect, but damn good. It can knock out difficult code analysis/reverse engineering more than 20-40x faster than I can. There was even a time long before AI when people thought the same of me…
Hell, at this point, I’d rather deal with Claude than a low-cost offshore team that puts up crappy implementations with endless go-backs and rewrites. I say this from actual experience, not xenophobia. There likely are some high performing offshore teams, I just wasn’t lucky to have the chance of working with one. Have had a couple of local colleagues that were worse… Claude’s probably much cheaper too.
Unfortunately, the “knowledge worker” jobs historically tended to be higher paid.
On one hand, AI is a major asset to experienced professionals with access.
On the other hand, it robs junior engineers hard-fought experience - much like watching a movie instead of reading the book.
And longer term, I think the entire economic structure of society is going to change. Doctors, lawyers, software developers, engineers — none of these are safe from AI reducing the need for headcount.
human intelligence was always overrated, at least in terms of it's actual value to society. phds are often unemployed or "overqualified" for the roles they apply for, roles that pay the most are often roles which are less deskbound and more client-facing, roles that ai cannot replace.
alot of investment related roles, those require reading human emotions and pleasing people, a pretty face and pleasant demeanor with emotional intelligence helps more than raw iq.
Human level embodiment took 10,000x longer to evolve than human level abstract and symbolic information processing, and is also probably 10,000x harder to replicate in an alternative form.
One way to think about this is that the parts of our brain that are responsible for spatial reasoning and the basic automatic stuff we do day to day to just navigate the worlds are evolutionarily the most sophisticated and oldest. Evolution has so taken over this job that most of the time it's automatic and we don't even think about it, which makes it feel easy even though it's wildly complicated when you actually sit down and try to analyze and intellectualize basic stuff like navigating a 3D space in which almost anything can happen.
Intellectual pursuits are, by comparison, relatively new. Part of the reason they might seem so impressive to us is simply because we can't rely as much on eons of biological evolution, so even though what we are doing most of the time is really simple it seems "smart and sophisticated" compared to the far more complex mental operations that we are lucky are just automatic.
Think about it, you have an entire complicated as hell biological interior system that your brain manages continuously without so much as a conscious thought.
This is also why this hubris around the idea that LLMs rival human intelligence is quite laughable. If anything one of the major reasons they are so "great" is because they don't have to deal with a host of issues themselves (their computation is not responsible for keeping their bodies alive and hearts beating 24/7, data center ops takes care of that, unlike our computational system which, in addition to solving all these intellectual problems, also needs to regulate the body continuously)
I think fully automated driving is not so far off. Once the technofascists have their utopia, they will just allow full self-driving vehicles in every road in the country and the people will deal with it regardless of actual capabilities.
> They “used math for what they needed, which was to evaluate their models,”
OAI first paper on the prize was a pretty shitty thing to do. But the rest of this looks like they are just mining the corpus and taking a brute force approach to do it.
Im going to say this again, the "pure math" academic types need to get a grip. Plenty of your peers got PHD's and happily made a fortune in finance, while screwing the world (See 2008 housing bubble) and none of you said a peep about that.
> Im going to say this again, the "pure math" academic types need to get a grip. Plenty of your peers got PHD's and happily made a fortune in finance, while screwing the world (See 2008 housing bubble) and none of you said a peep about that.
Has OpenAI shared any new interesting _problems_? I.e. do more than strip-mine mathematics, as said in the article, but said: "Here is a question I cannot answer, it's interesting for this and that reason, I believe it's true for this and that reason, it's beyond me though, have at it." Has that happened yet?
Considering that this is where the mathematician excels over machine wouldn’t that only offer a new view from such a mathematician that “OpenAI has rapaciously stolen from mathematics that last dignity: of choosing worthy problems”? Perhaps with sufficient effort the community of mathematicians may yet forestall such an effort.
I'm in this boat too and I find it a bit strange that this hasn't been more emphasized by mathematicians themselves.
So far, a lot of the hullabaloo has been around solving legacy problems, but more interesting and more important for actually progressing mathematics is the generation of new insight and new problems. Someone had to pose all those problems OAI is solving in the first place, in some cases, ages ago.
The collective letter response to navier strokes, sort of touched on this, but still didn't really emphasize the epistemic value of problem generation explicitly. I find it sort of strange that this has been so underemphasized.
I think their expectation is that “problems” as we’ve understood them just won’t be nearly as important to mathematical research going forwards. The field will look more like e.g. economics, where progress sounds more like “developed our understanding of the microeconomic foundations of macro” than “breakthrough on the Coase transaction cost problem in five dimensions”.
I've never heard a particular reason why the Millennium Problems were of special interest. They had resisted solution despite significant effort, but I don't know if Riemann or Navier and Stokes had ever singled those out as especially high priorities. They started as ordinary problems and became known for their durability.
So I'm not sure if it's something we should expect from a machine, just because we don't expect it from mathematics either. Maybe the machines will interrupt that process because we've changed how we go about looking, but perhaps not because old problems are solved.
In order to celebrate mathematics in the new millennium, The Clay Mathematics Institute of Cambridge, Massachusetts (CMI) established seven Prize Problems. The Prizes were conceived to record some of the most difficult problems with which mathematicians were grappling at the turn of the second millennium; to elevate in the consciousness of the general public the fact that in mathematics, the frontier is still open and abounds in important unsolved problems; to emphasize the importance of working towards a solution of the deepest, most difficult problems; and to recognize achievement in mathematics of historical magnitude.
It then goes into details in each of the chosen problems.
> Much as one cannot unhear a movie spoiler or a crossword clue, one cannot explore a problem as profitably and richly once one is aware of an existing solution
Certainly, it’s like being made aware of cheat codes to video games: One doesn’t enjoy the game as much after but surely knowing something is true means you can find many other things that come from the pathway to it now that you know it’s true. Lots of science and tech is like this where someone shows something is doable and then people find better ways to do it but until that person showed it was doable there was no progress.
> I am horrified by the many person-years” of research, particularly by young mathematicians, “being casually disrupted or destroyed,” he wrote.
That’s what technology does. And he might be upset but we’re the ones paying for it and frankly, I’m happier to have more questions answered than for fewer answered but in an artisanal fashion.
Most notable is the fact that the released facts are like “treated waste”. But if the question was worth answering and the value is only in a good answer and not simply a correct answer, then the value has simply not been realized.
But that feels strange then that the value is in the good answer, not just a right answer and yet one finds oneself unable to go get the good answer because a correct answer exists.
The fact that any answers are made public at all is useful to us all. That’s free knowledge. Would it have been better if OpenAI had discovered this very achievable fact and not said anything? Surely the papers, post discovery would have spoken of the “hoarding” of such “essential human knowledge”.
> That’s what technology does. And he might be upset but we’re the ones paying for it and frankly, I’m happier to have more questions answered than for fewer answered but in an artisanal fashion.
Would you prefer if graph theory didn't exist but we could answer if you can pass all bridges in Königsberg, without going through same bridge twice.
Or if we could answer any question without knowing why? What about answering Hilbert without theory of Godel's Incompletness theorem? Or knowing Nth digit of Pi without calculus.
Because that's essentially what you're asking for. If you gave humans a Djinn that
could answer perfectly any question, it would kill most of our endeavors. Because it's not a problem that was unsolved for years "It's just the thing Djinn already answered so why bother?" Even if one is extremely curious, others would not see value in that curiosity.
Imagine you are Euler and magical Djinn's exist (that possess all mathematical knowledge written up to that point exist and function identical to LLMs). You ask it to tell you answer to the bridge riddle and it says that's not possible and even gives some plausibly sounding examples even. Then it throws a 11 pages of gibberish proofs at you, which are all true but very convoluted.
And what if you start digging into it? Won't other people say - "Hey Euler go back to ploughing the fields? We can have Djinn's answer all the math questions but this field isn't going to plough itself."
Today's, the Djinni-Euler's equivalent of someone looking at LLM's Navier Stokes equation proof.
Perhaps they would discover new Liquid Set theory that would turn Liquid dynamics from a field fraught with problems into a beautiful mathematically simple meadow.
> And what if you start digging into it? Won't other people say - "Hey Euler go back to ploughing the fields? We can have Djinn's answer all the math questions but this field isn't going to plough itself."
Seems like an overly pessimistic take to me.
If you have a magical Djinn happy to answer questions, why wouldn't you badger it a bit for the details? I can't imagine Euler of all people just wanted the answer.
It’s not pessimistic at all. Look at the advances in math thanks to hundreds years of pursuit to prove Fermat’s Last Theorem before Andrew Wiles finally did it in the 90s. You will not get that with LLM. That’s what mathematicians are mourning over.
This genie only knows the details that other people already formulated, because it doesn't actually think for itself, it just regurgitates human thought. That's what everyone seems to be forgetting.
Remove the human theorizing and the LLM won't be able to fill in that gap. All it can do is remix. It cannot formulate and, equally importantly socialize new mathematical theory. Without social acceptance and promulgation theory is meaningless. I might have a private "sort of graph theory but not" studying a slew of new mathematical objects called quizzgams, but if I cannot actually articulate this theory to anyone and get an active participatory body of people around it, it's pointless. Any "theory" an LLM can create de novo is such a theory. Because, again, it doesn't actually produce novel thought, just novel byte sequences. A thought is more than just its encoding, it is processual and social, which all the nerds and vc bros seem to not understand.
I’ve been on the other side of that, getting badgered for details by people who don’t have the conceptual frameworks to understand what I’m doing, and my experience is that it doesn’t really work. Sometimes, when they have a concrete goal in mind, they can communicate it to me and I can produce a demo showing that the software I’m working on achieves their goal. More often, they ask questions that don’t fully make sense, and can’t understand either my answers or my guidance towards the questions they should be asking.
If I were a Claude instance, rather than a software developer fielding these questions as a small component of my job, could I take a step back and instruct them from first principles until they do understand? Maybe. But maybe there’s just too big of a gap to be crossed in one step.
> I’m happier to have more questions answered than for fewer answered but in an artisanal fashion.
But the value in seeking mathematical answers was never in the answers themselves. It was in the things one discovers along the way, the questions asked, the tools developed. The artisanal fashion is the part that benefits non-mathematicians.
These AI proofs only serve as an I-told-you-so. They're not creating surrounding theory which can later turn out to have useful applications elsewhere.
> I’m happier to have more questions answered than for fewer answered but in an artisanal fashion.
Given that this is pure mathematics we're talking about, for the most part: Why?
Perhaps you are a mathematician and so these questions are relevant to you, but for most of us, these questions and their answers will not make any tangible difference in our lives or society.
Maths has made a very real and tangible difference to society. The easy, obvious, answer is online banking, which wouldn't work without crypto, which is pure maths. There are many more.
It used to be that Pure Maths got to affect the real world via Physics (and that odd thing where last decade's interesting mathematical exploration turned out to be the solution to this decade's Physics discovery). But t'internet consumes Pure Maths pretty much raw.
> The fact that any answers are made public at all is useful to us all. That’s free knowledge.
When mathematicians discover a new result, they automatically become experts during the process, and then go around giving talks to explain their results, methods, and intuitions to other mathematicians. After years or sometimes decades of refinement and a healthy dose of serendipity, some of the mental frameworks percolate to other fields.
No one from OpenAI is willing to give a talk on any of their 300+ results. They've decoupled formal proofs from actual knowledge.
Right so this goes to the other alternative; there's still useful work for mathematicians after all in dissecting, distilling, and generalizing these results. But according to some we now inhabit an excluded middle where the AI produces useless results but also makes these problems useless for mathematicians to study further.
Is it useful to us all? I read thru the results list and despite years of college and graduate math classes only understood the statement of like five or ten. Prof Tao and his colleagues are able to spot the thing in their work that make faster imaging possible; are you?
Terrence Tao was very supportive of AI and formalizing math with lean when it looked like it was giving ‘generalists’ like him over mathematicians too deep into one particular niche field.
But now that it’s letting everyone, including mathematicians who are perhaps too specialized, explore ideas and patterns from related math fields, he’s suddenly ‘on a war footing’.
I don't think it is self-serving. Tao has stated plainly that he thought LLMs would go the route of assistants, like co-pilots. But what he feels now is that they are actually being engineered to be replacements. His claim is that the way OpenAI dropped these math results isn't "assisting" anyone, so we should all read between the lines.
Consider any man at his job, fighting the good fight. Your boss says "I've hired this man to assist you" and you are happy to finally have some help with your difficult task. Then one morning you come into work and see that the coming day’s work was done overnight by this other fellow. And the realization sinks in, he isn't there to help you, he is there to replace you.
In this view his reaction seems totally expected. I expect many professionals, e.g. doctors, lawyers, accountants, etc. will make this realization sooner or later. The companies are saying these tools will assist professionals, but in reality they are aiming to replace them.
you may not be aware of it but society largely runs on the results of maths, clueless indeed. and regulated or not, if there are financial savings to be made anyone can be shoved under the bus.
You might be mistaking my comment as a value judgement rather than an attempt to charitably explain the latest media communications from Tao. I detect in his position that some trust he granted the AI companies has been violated.
You are, of course, right that doctors, lawyers and accountants have more control over the environment they work in compared to mathematicians. But it doesn't change the underlying argument Tao is making. Your observation that those professions will react differently is likely to be true and only time will tell how effective their response will be.
But AI "assistants" is a marketing approach being made to all professionals at this time. Tao is implying this is a Trojan-horse strategy. If you let the assistants in without reasonable protections, you may lose control over the environment you are trying to protect.
> The companies are saying these tools will assist professionals, but in reality they are aiming to replace them.
If anyone doubts this, consider that Anthropic considers its TAM to exceed $30 trillion which is nearly the entire GDP of the US and 25% of the world GDP.
Only makes sense if they believe they will massively replace labor. Which, to their credit, they used to admit openly and publicly and only stopped because some crisis PR people told them "hey maybe you shouldn't say the quiet part out loud while the hoi polloi still have the power to stop this from happening".
Do their own hiring practices reflect that ambition? They sure seem to be recruiting a lot of humans for typical roles in business, legal, and engineering. If they want to flex the power of their models to automate human jobs, why not lead by example?
> They sure seem to be recruiting a lot of humans for typical roles in business, legal, and engineering.
They have practically infinite investor money to throw at salaries (lost in the noise of how much they are spending on compute) so why not use it for now to reach their goal?
My thesis isn't that they have already achieved the ability to replace massive amounts of the labor pool, just that that has to be their goal given their stated TAM.
> If they want to flex the power of their models to automate human jobs, why not lead by example?
They don't want to flex this. They realized (after a false start) that it is better for them to avoid talking about this in public.
That doesn't mean their goals changed, they just realized that flexing before the goal is achieved could endanger the goal being achieved through political backlash.
>They have practically infinite investor money to throw at salaries (lost in the noise of how much they are spending on compute) so why not use it for now to reach their goal as fast as possible?
Wouldn't throwing that money into GPU be better because AI is smarter and faster ? So they would reach their goal even faster ?
I’m not making a claim one way or the other about the intent of these companies, but if your product made labor 3x more productive, you could still easily justify a TAM of the entirety of US GDP, without resulting in anyone losing their jobs. Obviously, here in reality, AI does act as a near 1:1 replacement for people in some roles, so we should expect some job losses. But I don’t think there’s any reason we should expect the entire US labor force to be automated away any time soon, even taking these assessments at face value
Tenured at age 24, I doubt he has any experience in industry to understand what he was getting into. Which makes me even more incredulous as to why people worship him.
The reason we fell first is because we embraced the copilots. We used them heavily, this gave them more and more training data on top of the code they slurped up from our using their assistants. Soon the assistant became the agent who killed the human.
They also had a rich starting point from open source code (which we somehow gave up without realising it). All of this won't happen as efficiently for these other fields, but it will happen, much in the way you've described. We (swes) turned out to be a rather dumb group of people, which is hilarious since we think we're clever.
The only moat we have is the human to human element, we need to talk to other humans in the business. Customers, etc. But once they're replaced by machines or automation, then machine to machine communication will be the preferred mode of communication and work and the legacy humans will be increasingly cut out of the loop. Human in the loop is a death sentence we just don't realise it.
Isnt OpenAI and specifically Altman openly talking about that being the goal?
> The companies are saying these tools will assist professionals, but in reality they are aiming to replace them.
They are openly saying that the goal is to replace humans and make humans obsolete. Like, literally openly. I am not saying I agree with that goal or find it realistic, but they do think so.
A few-minute search on Google shows that Altman has said a lot of contradictory things on the subject. A few quotes, as examples:
“AI won't replace humans. But humans who use AI will replace those who don't.”
vs
"Some areas in the job market will be 'just like totally, totally gone' as they're replaced by AI agents.”
and
“AI will probably replace most of the jobs people do today,”
The most recent position I could find says he has revised his earlier position on a total replacement due to the lagging adoption of AI in many industries. But I don't think there is a single position we can point to that they are "openly saying". It is fair to grant even the CEOs of these companies a change in position as they gain more information about capabilities and adoption, rather than insist that they hold, have held and will continue to hold fixed opinions.
The point is that Tao's understanding has changed. He may have been naive before, he may have heard in their words something he wanted to believe. Perhaps he was misled by people who he trusted that worked in the company. But now that he has more evidence, in the form of the maths drops and the way they have been handled, his suspension of disbelief no longer holds. The companies could have done this in a way that included mathematicians, so they chose not to. And that choice reveals their aims unambiguously in a way that a media-trained CEO or careful PR department can avoid.
That suggests that Tao is not being "self-serving" as others claim. He is saying that his optimism in the best-case scenario turned out to be incorrect.
I've never perceived him as 100% supportive. He has rather been open-minded about a new technology and exploring what it can do. Now it turns out it can destroy his field, so he's naturally against it (at least against the destruction, not necessarily against the technology)
I think ‘destroy his field’ is ‘destroy his method of being famous in the field’.
Terrence Tao is famous because of solving problems better that most mathematicians because he has an uncanny ability to look across different math fields. AI gets everyone this ability for 20 to 200$ per month, eventually free.
Most mathematicians are happy to see interesting problems in their field being solved, and are looking to build on these to find more interesting problems.
One might say, in the end, what's wrong with being self-serving? Are we _all_ meant to be self-sacrificing? Are we all meant to be the eggs that get cracked to build this wonderful future ommlette?
I’m enjoying Terry’s “first popular math book” plug strategy - getting his name into discourse prior to release in two weeks without needing to plug the book. Maybe it’s just that AI maths has arrived at the same time. Well, I’ll buy it.
> Of course you get a lot of right-wing people saying "oh these Poles; they come over here taking our jobs". I think if a guy from Poland arrives and he doesn’t speak the language, he doesn’t have any money and he doesn’t know anyone, and he takes your job on the first day – you’re shit, fuck off."
For me it just shows that modern math is not that important and mostly ego driven. If it was important, mathematicians would cherish the progress. I can't imagine bio researchers get mad that AI cured cancer.
It was devastating to learn in my late 20s that 99.9999% of what the average person does is based on personal ego and not a communal vision of the future.
There are plenty people in every discipline that prioritize their ego over measurable progress towards a stated goal - even less so a community focused goal.
Pure Mathematics is closer to philosophy or art than it is to empirical science.
In Pure maths, people are not generally solving problems because of some pressing practical need, it's purely out of intellectual interest, and if we're lucky there's some immediate practical application. More often, the knowledge sits in pure form for a while and years down the road some new invention, or practical consideration, or other field finds a useful application.
Just like literature and visual art, it serves an important epistemic and social function even in spite of its immediate lack of raw practicality.
You are very wrong about bio researchers! I am a mathematician who has worked in bioinformatics and, just like we're seeing with a few curiously well-placed-to-obstruct-progress mathematicians, are in it for power, influence, and credit and anything that erodes any of that is bad, bad, BAD!!!
Is Terence Tao just another AI booster nowadays, by the sheer amount of PR generating posts? Reading HN one might think that there is only a single mathematician in the world left.
I remember just last year he said the LLM were like incompetent assistants, but not completely useless. The tech advances are mind boggling just looking back at that.
I do believe after the initial shock mathematicians will be able to use their math knowledge to solve practical problems, instead of lets solve this conjecture, prove this age old math riddle, how the pattern is possible. They may have some uses in fringe areas, but it doesn't seem all that useful. These guys are obvious quite brilliant, so when they combine it with other fields which they can branch to due to LLM's, they can still stay relevant. Ask your software engineer how that went.
IMO the most useful thing a mathematician can do is not solve “practical problems” but to come up with conjectures and riddles that look around several dark corners of math at once. The proving of a conjecture is a bit of a sport done to impress people with your cleverness; but the clever person was the one who made the conjecture to begin with.
Given the trajectory of LLMs to date I see them helping construct proofs for conjectures but not devising new conjectures of any material substance any time soon. The challenge is, as Tao has said elsewhere, part of being able to formulate a conjecture is by exploring the landscape around it by doing proof research. Seeking answers finds useful questions. The fear is the tools will take away the exploration of the landscape. But that’s silly IMO. It takes a special type of lazy to let a tool take away your understanding that’s important for what you’re goal is, and a special type of passion to learn with a powerful tool. The passionate will adapt and progress will accelerate, not slow down.
For instance Tao has posted some of his interrogations of proofs with AI assisted reasoning and it didn’t seem to diminish his ability to establish insight rapidly, but to shut down avenues rapidly and to pivot and expand his understanding rapidly. The insight was all his own, as was the path of inquiry. The understanding of the output was his as well. He carries it into tomorrow and uses it to formulate new more interesting conjectures the next day. The AI ceased to “exist” as soon as the inference turn was over.
The hand wringing is pointless and lame. John Henry moves aren’t necessary or useful, and when I see a tunnel boring project now I see tons of people well employed there, just the holes are bigger and better, and they’re engineers instead of strong backed hammer slingers.
Mathematicians might just be experiencing the first phases of the cycle of acceptance - resistance and anger, next is depression then eventually embracing it. Software developers as a whole are a bit further along the cycle.
"just". I think you should "justify" your use of "just".
Because it comes across as "ooh, you're just going through the five stages of grief, dummy! Look I've taught you psychology 101. Now that you know about grief, you might as well skip to the next cycle already. Support me on Patreon. Like, comment, subscribe"
Pointing out that someone is "just" grieving is not helpful.
I'm seeing this a lot in the past months of AI-for-math discussions. Had to get this off my chest.
Yeah, after I posted that I wondered whether someone would just focus on that word. You didn't just do that though, you went scorched earth - I meant no offense and sympathize with everyone affected by changes, I'm a developer and learning to embrace the change after a strong bout of existential crisis.
Math isn't useless. One of its defining features is its "unreasonable effectiveness" in formulating scientific ideas. It's hard to say exactly why it's so useful, but it is, even if mathematicians aren't trying to make it useful.
It's not at all clear whether the kind of math created by LLMs will retain the usefulness of the old math. LLMs are good at rigor but bad at intuition. It's not even really clear if the rigor was ever all that important. Theoretical physics is full of people using math in ways that don't meet the bar of rigor that mathematicians set for themselves, but they have enough physical intuition for what they're doing, so it doesn't matter. It's an open question whether post-LLM math will continue to supply scientists with the kinds of ideas that they need from it.
I find it weird to blame OpenAI for using math as a benchmark. If they hadn't done so, then in three months, say, after releasing Bel, someone else would have used math as a benchmark for it (like mathematicians already do since Fable 5 and 5.6 Sol). The result would have been the same (hundreds of open problems solved quickly).
It isn’t a benchmark though. Not yet. They rely on mathematicians to check the proofs including all lemmas. And on other mathematicians checking the reviews. Without being on the OpenAI payroll of course. Now they have flooded the zone, overwhelming the review capacity of the whole field by at least one order of magnitude. It is the move of 15-year-olds holding tools they shouldn’t have access to.
Did they force anyone to focus on reviewing those results instead of other ones? People choose to focus on them for a reason.
I think it's always better to publish preliminary results early (as long as you clearly say so, and they did), and let mathematicians decide which of those are worth focusing on.
A few people spend all their focus and tokens on one problem, and after some time I expect one of them to get at least closer to a solution. They can either work independently, or even better they team up and create a shared scratchpad for their agents. I don't see limits to scaling this, compared to OpenAI running however many subagents per problem.
And that multiply those few people by number of different problems.
No, that's not what I say. My argument is that once the Torment Nexus is built, it is stupid to argue that they used it to torment people, when someone else would have tormented people with it anyway. If one wants to argue something, it's that the Nexus should not have been built.
And I honestly have no idea what Tao wanted to happen. How could you ensure that the AI "empowers" scientists and doesn't replace them. Intelligence is intelligence. If it's powerful enough, it will replace other intelligence.
I very much wish they'd just let mathematicians have at it empowered by the model so we could all treat the people who claim they have a lean verified proof with an unintelligible explanation as the crackpots they are.
One aspect that currently frustrates me is that even paying for the highest plans and spending multiple quotas across all flagship models, I still haven’t been able to reproduce any of the amazing feats Anthropic and OpenAI gush about.
I suspect their internal harness, models, and thinking budget far outstrip anything us peasants have access to (never mind the compute).
So saying that they achieve all these headline results with 3 hours of GPT Pro is disingenuous. If that were so, hundreds of math phd students would’ve already recognized it.
I’m coming to the realization that my monthly subscription is only helping subsidize their real internal AGI, and my interactions are a small amount of human feedback and creativity which they can strip mine should they choose to.
They have been nerfing the models for us plebs for a while now. Anthropic does the same, and if the Chinese models aren't doing this already, they will eventually.
It's hard to prove, but I'm personally extremely convinced of this. The point of AI (from the perspective of these companies) is to solidify and centralize their control of the economy and the world.
If you ask Qwen3.8 to give you a list of things it would refuse to help you do, it will list a number of things related to building/creating/training/altering LLMs. This didn't make much sense to me until I read that the #1 thing these frontier companies are optimizing their models for is training/creating/editing newer AI models. That makes a lot of sense to me now, and that's also something which they will want to keep 100% under wraps.
Therefore the #1 focus of their research and efforts and expenditure will be something that they don't want any of us to get access to. I think many other behaviors like model nerfing follow from that. It helps explain these frontier research products which we ourselves cannot get access to.
If your takeaway from all of this is that some other group of people has the special-est skill that can never be replicated by a computer, I fear you’re missing the point.
How does that make any sense? Why do you assume that someone more intelligent would not be able to wield any AI tooling better than an average or below average human? Just curious if you're thinking short term or long term on this.
This tool belongs to somebody (capitalist!) else and in a new world better tools will always win, not individuals anymore. And it feels like to go from being Rembrandt to CVS employee who takes passport pictures - result much better quality and cheaper but no skill and glory.
Interestingly, the dynamics strongly resembles the industrial revolution when from highly skilled manufactures society moved to factories with power machines. And it all came down who owns means of production. Now AI is a new means of production for all white collars jobs and if the industrial revolution taught is the way to go, nothing good will come out of it.
I'm somewhere in the middle on that. People used to think that computers would be owned only by large corporations, then home computers appeared and improved. It's getting tight now that logic scaling is mostly exhausted, at least with current operating principles. Even so, new specialized hardware and model improvements may help (for inference). Training seems to be the biggest problem, it seems hard to democratize.
Nothing good for skilled labor you mean, especially in the short to medium term - just like in the industrial revolution. If the new technology is any good, it will produce desired things (let's say medical breakthroughs?) faster.
I don't particularly like AI, but if AI replaces developers and other knowledge workers, not everyone is worse off, and I don't mean just capital owners.
No, but someone who has trained their whole life to run fast, surely will be able to coach someone else to run faster than someone who has been on the couch watching television?
I use LLMs to great advantage, solving problems in my domain. I no longer make for loops.
my point was that the new future where running faster doesn't matter any more might come, because everyone is driving (and potentially can't run much at all).
but the analogy would be a riflemen trained to aim, and a person pressing button for guided missiles. How good you are at the original task is unlikely to matter when the new way supersedes the old way.
Over someone who never fired any real weapon? Probably. You are not startled by violent sounds. You have intuition for ballistics, time to target, CEP, weather effects.
The social fabric that produces math got ripped up by marketing guys who didn't even bother to check their work.
He described it at 'strip mining', 'rude', and said that many grad students, who are given problems like these as a part of their training, had the rug pulled out from under them (because its customary to let people know you are working on something to prevent this kind of thing).
And your whole analysis is that Tao is worried about his 'identity', status, and 'being a differentiator'.
This is projection on your part. People have values that bear no resemblance to yours and it would behoove you to learn to recognize that.
Well they've been doing that to devs for a couple of years now and yet mathematicians seem to be much more vocal whingers (might be selection bias in the articles I'm reading though).
AI is coming for all of us knowledge workers, nobody is immune.
The rug being pulled is the expectation of a reasonably high-trust ethical environment. The question to be asked is: What will graduate study look like in an utterly zero trust ethical environment?
If there is still such a thing as graduate study. If there is still such a thing as study. Tell me what to say to someone about to enter college about why it isn't a waste of their time.
Most colleges, even before LLMs, were already a waste of time for most students. I mean they were somewhat effective for networking and economic signaling to employers but they had already lost the plot in terms of real value to society. And now with LLMs the entire education system will need to be rebuilt. Let's get to work.
This was already why people went to grad school. I think people weigh a complex bag of known and perceived motivations and risks. One of the risks is that your project gets "scooped." There were some weak ethical and practical safeguards around that. But completely rearranging the risk equation will change peoples minds about whether graduate study is worth it.
Disclosure: PhD, early 1990s. In my case, replicating my project would have required a sizable investment in equipment, much of it was purpose-built, among other things.
The ethics of academic research weren't idyllic, but they were the devil known. It was at least informally accepted that a student should have a fair chance of finishing their project. Cherry picking student projects and publishing them ahead of the student was considered unethical. I got my graduate degree under those conditions, as did my parents.
In math, proofs are "the coin of the realm," and could be devalued to zero. This leaves no answer as to how someone will earn a graduate degree. For instance, what kind of work will be accepted by a thesis committee in five years? The professoriat can only say "we will need to change some things." That's not good enough when the deadline for grad school applications is a few months away.
Strip mining interesting conjectures. AI can answer interesting(pedagogically valuable, carrying important implications, etc.) mathematical questions in formally correct gibberish far faster than mathematicians can generate them. Until AI can replace the conjectures it resolves on its own, it's consuming a finite resource.
It is usually not the monopolist who is protesting against strip mining, it is usually the monopolist doing the strip mining and the environmentalist protesting it. Which is to say, I don’t think mathematicians are claiming a monopoly on mathematics, but I think AI might.
It's time for the social fabric of math to be ripped up, it's an industry being disrupted.
No one has the right to prevent others from working on a math problem. If you're working on a problem the best way to not get scooped is transparency: publish whatever you have as soon as possible.
Tao's behavior is most consistent with any other worker resisting automation for selfish reasons.
Tao is very prominently an early AI adopter. He, and other mathematicians, have fruitfully integrated it into their work. The way it is integrated matters, and it is carelessness and selfishness on the part of OpenAI that he is objecting to.
> It's time for the social fabric [of math] to be ripped up
This is the attitude of a social arsonist. You can, and should, integrate changes without maximizing destruction for its (or Altman's) own sake.
> an industry being disrupted.
This is the SV euphemism for law-breaking and monopolist practices in a country that has forgotten how to prosecute those with the most power and privilege.
> This is the SV euphemism for law-breaking and monopolist practices
Quite the opposite. The mathematicians are the monopolists who are upset that new-comers are able to do their job better than them, and even worse, they are giving it away to the world for free!
The horror! People giving knowledge to the world for anyone to use freely! Won't someone think of the oligopoly that was coordinating to make sure each of the gatekeepers get their slice of the pie?
Monopoly: Exclusive control of a particular market that is marked by the power to control prices and exclude competition.
Mathematics is collaboratively and publicly produced, it is vetted for free by peers, and results are publicly available. There is no 'market' and no 'price'.
These results, like the invention of LLMs, are often used by private companies for the generation of private profit.
"Open" AI used a private, unreleased, for-profit model to generate those papers.
What, precisely, did the mathematics 'oligopoly' gatekeep you, or anyone, from?
> their slice of the pie
There's very little money in academia, and even then, its a struggle to get any funding at all. What are you even talking about?
Final point, 'knowledge' is socially produced as well. Until the papers are accepted, they are machine artifacts that might be knowledge, but that same system you disdain will be what incorporates and validates those insights.
> Mathematics is collaboratively and publicly produced, it is vetted for free by peers, and results are publicly available. There is no 'market' and no 'price'.
How wonderful then! You should be happy to know that the AI companies are releasing information for free for all to collaborate on then!
> What, precisely, did the mathematics 'oligopoly' gatekeep
Well what they are trying to do is convince the AI companies to stop releasing all of their findings. Because they are upset that other people are able to do their job better than they are. Its the classic gatekeeper move.
> There's very little money
Oh there is much more than money to silly games like these. Instead, clout and the admiration of their peers, in this tiny insular group for sure, is what most of these people are so desperate to acquire.
Fame (well... relative.... in their subgroup) and admiration of their peers is a much stronger motivation factor for humans than mere money.
> but that same system you disdain
I don't distain anyone who is willing to get with the times. My distain is held out only for those who want to prevent the rest of the world from using the latest tools to achieve real results. If a mathmatician simply doesn't care and isn't wineing about all this AI stuff, I have no issue with them.
The "not bothering to check" part is the most annoying.
There are already several of these "proofs" that have been retracted because they did not make sense.
I would have no problem accepting this whole ordeal if the humans behind each proof took the time to re-read, understand and synthesize the result in a readable paper.
There is ample evidence of how damaging fake information can be, and paper retraction never fully undoes the damage (see the Andrew Wakefield and his antivax "paper" for instance).
Until AI agents are embedded in glasses and everyone gets comfortable with sending their conversations to AI, the smartest person in the room is still going to be the smartest person in the room because they say smart shit. Or until the dumb guy types into their phone and says smart shit. Given the pushback to Meta's glasses, I think that'll be a differentiator for a bit longer still. Even before LLMs, you could check Wikipedia and different rooms treat the people that do that in different ways. And then there's the "Don't be a meat proxy" movement.
Not all rooms are real face to face. Smartest person also needs time to think and lookup information. Worst smartest person can think on one hypothesis at a time.
What? This is like a drive by shooting or something. Several PhD students almost certainly lost months of work, and others considering a PhD will have more doubts. Those problems aren't necessarily important because of the solutions, but because mathematicians can make unique contributions to mathematics in the process of solving them. LLM proofs haven't created any new mathematics techniques.
Everyone who built an identity on earning a living doing useful work for other people is going to have to come to grips with that no longer being part of their identity.
No. It's just people who built an identity on the power that their intelligence afforded them. We view that power and the intelligence itself as the same thing, but they're actually separate.
I don’t think that is correct. For consistently good results, you need to set the right goals and be able to judge the quality of your outcomes. Even if you don’t do the hard work in between some thinking is required for non trivial work.
You're out of date. The chats that found new Riemann conjectures were "look for a solution to the Riemann conjecture", followed by a bunch of "you can do it" and "I believe in you".
Yeah that's an interesting point of view. In my studies I got comfortable with the fact that there's always someone smarter than me.
At the same time it should be considered that with physical labor workforces lost their income source numerous times. Ironically they are currently being considered a safe haven.
I just finished reading "Outsmarted: The Hidden Costs of Being the Smartest Person in the Room".
Most of the HN crowd would resonate a lot with the points raised in this book.
The last(-ish) chapter does have a section on how AI has robbed the "cognitive elite" of our identity.
Same as code, the llm can generate vast amounts and it seems to work, but no one understands the documentation Claude Code, or other LLMs, produce. Essentially, garbage because, to make use of it you need to be able to understand it, usually at a cost (unquantified), and if only LLMs do then what is the point?
It turns out it doesn't matter, because outside of a few narrow bands like Claude Code, no one cares what documentation an LLM writes for your project, because no one is ever going to go looking for someone else's project or library to solve their problem anymore.
So either an LLM decides to pull in your project as a library for reasons, in which the LLM can understand the documentation, or else no one besides the creator is using it anyway. So it doesn't matter.
Why would anyone use someone else's AI-coded project if they can get their own version that better fits their own needs in less time?
Sure you can make Claude produce documentation & code that is readable. You need to steer it and show it how you want it to write. When it gets it wrong, find counter examples so it can learn the style you want. Otherwise it's going to write what it thinks is best. What is best for LLM comprehension and efficiency is not necessarily best for the human.
It's a little funny how quickly the conversation goes from "AI will never be able to {my-field}" before turning into "We cannot have AI ruining {my-field} with its obviously superior abilities"
If OpenAI had never made a push to get good at maths, we might still have a sizeable group of peopne stuck at that first line. I say that because I still hear in discussions the naive rebuttal "But we'll still need _someone_ to {one-step-up-in-hierarchy}" . Although you have to be curious now how long that sentiment can last.
The good news for those doing abstract maths might be that, now that they've proven that AI can do the job in principle, there's probably not much left in it for them to keep going. Mathematicians should know better than anyone the lack of a direct profit motive in doing the work that they do.
Hacker news is not really a good place to talk about nuances involving academia. I don't expect VC minded hackers comprehend how someone can devote years of their lives working minimum wage trying to answer questions which earn them a fancy title on a piece of paper, for that matter.
A lot of us here are VC minded hackers *because* we spent years of our lives working minimum wage jobs in academia and then eventually realized we preferred to build!
> A lot of us here are VC minded hackers because we spent years of our lives working minimum wage jobs in academia and then eventually realized we preferred to build!
This is a disingenuous framing. I've "built" more in academia than in some jobs I had, and some VC-backed startups are renowned not for building but for their lack of ethics.
I think it's a very skewed form of ethics to make people pay without convincing them.And fwiw Making money is correlated with productivity which is negatively correlated with a lot of bad things so I think on pure technical point you are wrong, there are subtleties.
Reward alignment is the differentiator. Obviously it's very far from clear-cut what is worth funding in research. But i've seen a lot of crap research, that would have died off far sooner if it had some VC/free market equivelant.
Or preferred to eat. I love teaching, and I'm good at it. If it were more valued by society then I'd do more of it. As it stands, I'm in adtech, and I teach a little piano and math on the weekends.
It's great there's an exception, but that isn't the point, they're talking about people who get paychecks as mathematicians. BTW what did Grigori do for a paycheck?
A large fraction of people are open source contributors, hobby game devs, and various other types of people who fall under "devote years of your life with no formal remuneration"
The US-American mind cannot comprehend "free time" nor working hard on something just for the love of it and the joy of collaborating with other like minded people.
There's a reason that so many open source builders and contributors are unemployed people in social-welfare European states. There's a reason that nearly every great work in the English canon was written by someone who was on the dole at the time, from Nietzsche to Harry Potter.
> If OpenAI had never made a push to get good at maths
By snooping on work math researchers were doing on their platform and then overtaking them when sufficiently close to completion? isn't that what happened i'm not sure
No, as far as I know there's no evidence they used the other team's work for their own (and OpenAI changed their initial 'It's possible' to 'we double checked and it didn't happen'). What they did was hear a rumour about it and then spent millions of dollars on compute to scoop them.
That's not what happened. The claims of this for Navier-Stokes were already exagerated. It was dramatic but there was AI involved on both sides to make progress and the final proof did not use the same approach.
Finally, the list of hundreds of solved problems after that completely defeated any argument in that direction.
so far it's been programming and math, the two fields most people would naturally assume computers would be good at :P
not saying it's not possible ai will continue to expand to more fields, but i don't think it's a coincidence that both math and programming are easily automatically verifiable. in fact for programming their effectiveness drops substantially once you remove any sort of testing or verification loop, and i imagine it doubly-so with math. try getting an agent to build something without any tests or verification, you'll see it go back to near-2024 levels. half of the work i do with agents is setting up a harness that can take and analyse screenshots, send direct input to the application, and compare results. this is something that should come standard with every harness.
if we can build automatic verification for more fields is a different and perhaps more relevant question.
> so far it's been programming and math, the two fields most people would naturally assume computers would be good at :P
If that is true, then only because "most people" have no understanding of what differentiates programming and math from "computation". Computers (as the term suggests) have always been extremely good at the latter, and useless at the former. It's only 2–3 years ago that computers, by way of LLMs, have started being good at the former.
I have to confess to being wrong in exactly that way in the past. On the other hand it is probably always good to behave as if the next step up in the hierarchy is still available. There's no point letting these things spoil your day twice.
Now I'm forced to think about it much more carefully. I don't believe that current LLMs can find every possible proof on their own from now on. I also agree that the PR returns have fully diminished. Given how hard it's been in recent years to fund e.g. a poorly paid postdoc, I doubt there's a ton of money lying around for getting the best models to work on mathematics problems. I think there's a chance we're heading into a really good era. I was able to make so much progress back when Google search was a useful research tool...
I'm of two minds about this. On the one hand, you get miraculous results with AI, and that is what you are referring to with "be able to {my-field}". On the other hand you regularly get useless, if not toxic, crap out of the miracle machine. If you work with AI, you know these moments: "no that is not what I meant at all!" [1] or "how could you have already forgotten the central motivation I laid out just ten back-and-forths ago?". We're calling this the jagged frontier, but the jaggedness itself is not the problem, it's how deep these cuts go into aspects that are important. These are, why AI is still not "obviously superior" in its abilities, can't replace mathematicians and software developers and lawyers and everyone else who is primarily working with text. (It's also why people say it is ruining their fields, instead of replacing them in their fields, because it only replaces some part of the work, and what's left is not tenable or pleasant (e.g. doing reviews all the time).)
Over all, I'm not quite sure whether I see these deep cuts occurring less and less with progressing AI. It might just be, that I'm giving them progressively harder tasks as well. But the fact of the matter still is that they aren't there yet and it needs proof they will get there ("there" being able to do competent work that is not strongly limited in their scale by human review and correction). So I'm still not convinced about the point you want to make, even though I see AI disrupting about every part of society, and am hoping for the benevolent super AI to lead us to utopia (and afraid of the cliff of human dictatorship on one side and extinction on the other).
[1] I've started to use these moment to reflect on whether a human in a comparable situation would fare better. Having only a few text inputs and small glimpses into a code base and knowing next to nothing about you or your project personally. Quite a few times I've concluded: No, they would not. But more than 50% they would, and also, these limitations are technological limitations of LLM, so you should probably not excuse them for that.
But isn't part of the problem that you simply don't have access to the hardware larger organisations do? OpenAI certainly run into the same constraints. but if they have access to terabytes of memory, while when you use the same model it's closer to 32GB, that will surely affect how the model behaves?
So the pain points you and I frequently run into, simply don't exist for a well connected individual(Sam Altman).
> The good news for those doing abstract maths might be that, now that OpenAI have proven that their AI can do the job in principle, there's probably not much left in it for them to keep going. Mathematicians should know better than anyone the lack of a direct profit motive in doing the work that they do.
True.
The other possibility is maybe OpenAI doesn't know what to punch into the calculator now.
They have gatekept for literally 100s of years. Decide the problems, decided who is good enough to have a chance at solving them, decided what it means to be solved.
All without meaningfully contributed to human progress - beyond teaching calculus and lin al to future machine learning scientists - for 40+ years.
Every other job in the developed world has been disrupted by technology over the past 100 years. For many of them - multiple times. Now it is pure math's turn, and they are acting like they always do - elitist, entitled, worthy of special consideration.
I would usually have been part of the downvoting brigade for this comment.
But to be honest the reaction from the ivory tower here is seemingly quite telling to a layman watching from afar.
It looks far more like this was some extremely gatekept community where you were only allowed to work on sanctioned things if so deemed worthy. Better not step one someone’s toes lest you get exiled! How much progress was lost due to newcomers never working on a problem due to told not to?
It’s starting to make those old timey Usenet kooks look a whole lot less crazy to me. They seemingly were right about this part.
Mathematics was funded and supported by a whole bunch of janitors and blue collar workers busting their asses to let entire departments of thinking men and women work on hard problems for the betterment of society. If it was just a super special friends club the whole time that was gatekeeping what amounts to a hobby, it’s going to me much harder to justify such a social contract.
Honestly not even sure what to think at this point. Extremely disappointing to me the more I read from these people.
I hope the LLM's will tackle theoretical physics next. If you think the math community is bad in terms of gatekeeping, and it is, the physics community is so much worse. Even community insiders are afraid to share their more unusual ideas at risk of becoming pariah's, outsiders have no chance. And that while they seem more or less stuck in their current track.
Admittedly it's also just really hard to come up with something better than the current track, but doing better has to start with being eager to explore different directions. Even if the new directions don't quite work just yet. And that eagerness just isn't there in the community, new directions are threats to the careers of sensitive people
Theoretical physics would, from my outsider view nowadays, benefit from an injection of vitality and an ability to tackle new venues. But the question is open whether LLMs can truly break new ground or if they work within the convex hull of current theory. And if that matters, too.
In the past there was too much activation energy to point out flaws in other’s reasoning. This is a natural response to slop cannons even prellms with misinformation shared.
I don't know if this will happen in math, but in other fields where computers have outstripped humans (chess and poker, for instance), there's been a loss of creativity, but there have also been other forms of creativity that take its place. There was a period of reeling where we had to figure out what it meant to just be able to know answers to the question that made the game fun, but there's also a moment of realization: just because a computer says something doesn't mean we can't make it our own. Using poker solvers to find creative non-equilibrium strategies can still be fun: the ingenuity comes from using the tools to think out side the box. Or, finding a nicer set of abstractions that explains why a strategy does a certain thing, and then starting to notice it showing up other places.
For math, just because we know a thing is "TRUE" doesn't mean the investigation is done. The folks that love math are driven by solving puzzles, and the puzzle is still there, only changed: instead of being "is this true?", the puzzle is now "why is this true?" Building theories, finding nicer proofs, etc: there is still plenty of room for creativity.
But in the meantime, it really is tragic that so many young minds have had their work stripped away from them.
Chess and poker as spectator sports have a high incentive to see humans compete fairly, and I think that's the reason why there are such strong anti-cheating measures (at least in chess; in poker, too, I suppose).
And the economy puts no such large incentives on much of the remainder of economically valuable work being purely human, quite the opposite actually. And instead of regulating the AI industry (which produces the "cheating tools"), they get trillions of investment. So this seems to be going in the opposite direction where it is actively worked against humans staying economically relevant.
I dont know what the future will hold, but in the past each time technology took over a profession, it ended up opening more doors than it closed. However the transition has always been rough.
I have faith that when AI settles, it too will open more doors than it has closed.
463 comments
[ 2.4 ms ] story [ 32.7 ms ] threadBut we still have no self driving cars. Seems like the automation came by in the opposite direction. Perhaps driving will actually be the final thing automated.
While I agree with your statement overall, as someone who drives everyday in LA alongside Waymos without anyone inside of them I must disagree with this aspect of it.
Because if it became fashionable for pedestrians to wear large green dresses with red balls, and everybody would wear that, we have no idea what the AI would do.
Something I'm noticing a lot is that automation finds easy footholds in unregulated industries because there's no basis to push back on it. And there's a lot more tolerance for poor quality work in unregulated industries too.
A truck or a car that has a 99% success rate would be the top threat to society and quickly rise to the leading cause of death under 30 if widely deployed.
You have to look at the benefit of something when balancing against a cost.
Wait, aren't human-driven cars already the leading cause of death under 30?
I agree that LLMs have and will continue to increase the number of activities that people can DIY. But I believe many people will still choose to pay for others to do things on their behalf.
This is beside the point of the analogy though.
99% successful web development where the 1% missing is security that and the company leaks its entire repository of user data every few months is not good.
99% successful car driving where 99% of 5-year periods have zero accidents per car is an order of magnitude safer than the average driver. (And gut feeling, I suspect approaching the tail-end of the safest drivers on the road.)
It feels like it's happening every Thursday and nobody cares. They just give people credit watch coupons or what's it called, and move on with their business. Worst case, a few pennies in fines.
But if CVS leaks your purchases on a weekly basis, it's probably going to affect people's willingness to shop there. (Or at least to use non-cash.)
In 2025 it was occasionally driving weird. Today it's a much better driver than me.
All past inventions have put some of the most miserable humans out of a job. Farming by hand is hard, so is copying books and don't even get me started on the dullness of making clothes by hand.
Driving all day isn't all that far off, though still much better than the above and definitely better than riding horses all day.
Telegraph operators? Hell no. Dude sitting in an elevator all day? No thanks.
It's always been from the other end. This time around, there's gonna be some 10-20 years of robot manufacturing, then a universal lifetime of robots fighting misaligned robots until enough paperclips have been manufactured.
Or you know, we block out the sun and live in caves or whatever you fancy.
Do you live under a rock?
The same isn't true for elephants.
[1] https://pages.skybrian.com/2026/waymo-robotaxi-fleet/
Developers are guilty of hubris and they are getting slapped back to reality.
It was too easy to assume that an AI would never approach human intelligence and hence would only be used for “boring” jobs like driving.
I’ve long disliked the hype around AI but now see that human intelligence is perhaps slightly overrated, at least in terms of the loftier positions seemingly requiring such.
Claude Opus puts to shame any highly caffeinated “10x” software developer. Not saying it’s perfect, but damn good. It can knock out difficult code analysis/reverse engineering more than 20-40x faster than I can. There was even a time long before AI when people thought the same of me…
Hell, at this point, I’d rather deal with Claude than a low-cost offshore team that puts up crappy implementations with endless go-backs and rewrites. I say this from actual experience, not xenophobia. There likely are some high performing offshore teams, I just wasn’t lucky to have the chance of working with one. Have had a couple of local colleagues that were worse… Claude’s probably much cheaper too.
Unfortunately, the “knowledge worker” jobs historically tended to be higher paid.
On one hand, AI is a major asset to experienced professionals with access.
On the other hand, it robs junior engineers hard-fought experience - much like watching a movie instead of reading the book.
And longer term, I think the entire economic structure of society is going to change. Doctors, lawyers, software developers, engineers — none of these are safe from AI reducing the need for headcount.
Which role is that, and why you think so ?
On other hand, in a completely fair market, i think AI can make profit. A lot of traders just stay at home and monitor news.
https://en.wikipedia.org/wiki/Moravec%27s_paradox
Intellectual pursuits are, by comparison, relatively new. Part of the reason they might seem so impressive to us is simply because we can't rely as much on eons of biological evolution, so even though what we are doing most of the time is really simple it seems "smart and sophisticated" compared to the far more complex mental operations that we are lucky are just automatic.
Think about it, you have an entire complicated as hell biological interior system that your brain manages continuously without so much as a conscious thought.
This is also why this hubris around the idea that LLMs rival human intelligence is quite laughable. If anything one of the major reasons they are so "great" is because they don't have to deal with a host of issues themselves (their computation is not responsible for keeping their bodies alive and hearts beating 24/7, data center ops takes care of that, unlike our computational system which, in addition to solving all these intellectual problems, also needs to regulate the body continuously)
There is more than one way to skin a cat.
OAI first paper on the prize was a pretty shitty thing to do. But the rest of this looks like they are just mining the corpus and taking a brute force approach to do it.
Im going to say this again, the "pure math" academic types need to get a grip. Plenty of your peers got PHD's and happily made a fortune in finance, while screwing the world (See 2008 housing bubble) and none of you said a peep about that.
They call it the most populist post of all time.
Considering that this is where the mathematician excels over machine wouldn’t that only offer a new view from such a mathematician that “OpenAI has rapaciously stolen from mathematics that last dignity: of choosing worthy problems”? Perhaps with sufficient effort the community of mathematicians may yet forestall such an effort.
This will be interesting to watch: Where will AI creating and solving problems in a loop lead to.
This leads to the comment I've seen many mathematicians make that humans should refocus on understanding not just proofs and mechanics.
So far, a lot of the hullabaloo has been around solving legacy problems, but more interesting and more important for actually progressing mathematics is the generation of new insight and new problems. Someone had to pose all those problems OAI is solving in the first place, in some cases, ages ago.
The collective letter response to navier strokes, sort of touched on this, but still didn't really emphasize the epistemic value of problem generation explicitly. I find it sort of strange that this has been so underemphasized.
So I'm not sure if it's something we should expect from a machine, just because we don't expect it from mathematics either. Maybe the machines will interrupt that process because we've changed how we go about looking, but perhaps not because old problems are solved.
From the very top of https://www.claymath.org/millennium-problems
In order to celebrate mathematics in the new millennium, The Clay Mathematics Institute of Cambridge, Massachusetts (CMI) established seven Prize Problems. The Prizes were conceived to record some of the most difficult problems with which mathematicians were grappling at the turn of the second millennium; to elevate in the consciousness of the general public the fact that in mathematics, the frontier is still open and abounds in important unsolved problems; to emphasize the importance of working towards a solution of the deepest, most difficult problems; and to recognize achievement in mathematics of historical magnitude.
It then goes into details in each of the chosen problems.
Certainly, it’s like being made aware of cheat codes to video games: One doesn’t enjoy the game as much after but surely knowing something is true means you can find many other things that come from the pathway to it now that you know it’s true. Lots of science and tech is like this where someone shows something is doable and then people find better ways to do it but until that person showed it was doable there was no progress.
> I am horrified by the many person-years” of research, particularly by young mathematicians, “being casually disrupted or destroyed,” he wrote.
That’s what technology does. And he might be upset but we’re the ones paying for it and frankly, I’m happier to have more questions answered than for fewer answered but in an artisanal fashion.
Most notable is the fact that the released facts are like “treated waste”. But if the question was worth answering and the value is only in a good answer and not simply a correct answer, then the value has simply not been realized.
But that feels strange then that the value is in the good answer, not just a right answer and yet one finds oneself unable to go get the good answer because a correct answer exists.
The fact that any answers are made public at all is useful to us all. That’s free knowledge. Would it have been better if OpenAI had discovered this very achievable fact and not said anything? Surely the papers, post discovery would have spoken of the “hoarding” of such “essential human knowledge”.
Overall, I find the thing unconvincing wholly.
Would you prefer if graph theory didn't exist but we could answer if you can pass all bridges in Königsberg, without going through same bridge twice.
Or if we could answer any question without knowing why? What about answering Hilbert without theory of Godel's Incompletness theorem? Or knowing Nth digit of Pi without calculus.
Because that's essentially what you're asking for. If you gave humans a Djinn that could answer perfectly any question, it would kill most of our endeavors. Because it's not a problem that was unsolved for years "It's just the thing Djinn already answered so why bother?" Even if one is extremely curious, others would not see value in that curiosity.
Now math isn't my wheelhouse, but every LLM I've seen provides far more than just the final answer and is quite willing to answer questions.
Imagine you are Euler and magical Djinn's exist (that possess all mathematical knowledge written up to that point exist and function identical to LLMs). You ask it to tell you answer to the bridge riddle and it says that's not possible and even gives some plausibly sounding examples even. Then it throws a 11 pages of gibberish proofs at you, which are all true but very convoluted.
And what if you start digging into it? Won't other people say - "Hey Euler go back to ploughing the fields? We can have Djinn's answer all the math questions but this field isn't going to plough itself."
Today's, the Djinni-Euler's equivalent of someone looking at LLM's Navier Stokes equation proof.
Perhaps they would discover new Liquid Set theory that would turn Liquid dynamics from a field fraught with problems into a beautiful mathematically simple meadow.
Seems like an overly pessimistic take to me.
If you have a magical Djinn happy to answer questions, why wouldn't you badger it a bit for the details? I can't imagine Euler of all people just wanted the answer.
Remove the human theorizing and the LLM won't be able to fill in that gap. All it can do is remix. It cannot formulate and, equally importantly socialize new mathematical theory. Without social acceptance and promulgation theory is meaningless. I might have a private "sort of graph theory but not" studying a slew of new mathematical objects called quizzgams, but if I cannot actually articulate this theory to anyone and get an active participatory body of people around it, it's pointless. Any "theory" an LLM can create de novo is such a theory. Because, again, it doesn't actually produce novel thought, just novel byte sequences. A thought is more than just its encoding, it is processual and social, which all the nerds and vc bros seem to not understand.
If I were a Claude instance, rather than a software developer fielding these questions as a small component of my job, could I take a step back and instruct them from first principles until they do understand? Maybe. But maybe there’s just too big of a gap to be crossed in one step.
https://www.youtube.com/watch?v=HQuH5nI-pHQ
But the value in seeking mathematical answers was never in the answers themselves. It was in the things one discovers along the way, the questions asked, the tools developed. The artisanal fashion is the part that benefits non-mathematicians.
These AI proofs only serve as an I-told-you-so. They're not creating surrounding theory which can later turn out to have useful applications elsewhere.
Given that this is pure mathematics we're talking about, for the most part: Why?
Perhaps you are a mathematician and so these questions are relevant to you, but for most of us, these questions and their answers will not make any tangible difference in our lives or society.
It used to be that Pure Maths got to affect the real world via Physics (and that odd thing where last decade's interesting mathematical exploration turned out to be the solution to this decade's Physics discovery). But t'internet consumes Pure Maths pretty much raw.
When mathematicians discover a new result, they automatically become experts during the process, and then go around giving talks to explain their results, methods, and intuitions to other mathematicians. After years or sometimes decades of refinement and a healthy dose of serendipity, some of the mental frameworks percolate to other fields.
No one from OpenAI is willing to give a talk on any of their 300+ results. They've decoupled formal proofs from actual knowledge.
Terrence Tao was very supportive of AI and formalizing math with lean when it looked like it was giving ‘generalists’ like him over mathematicians too deep into one particular niche field.
But now that it’s letting everyone, including mathematicians who are perhaps too specialized, explore ideas and patterns from related math fields, he’s suddenly ‘on a war footing’.
https://news.ycombinator.com/item?id=50128468
Eh... That “everyone” is a multi billion (trillion?) $ company paying top mathematicians specifically hired for generating math proofs.
Consider any man at his job, fighting the good fight. Your boss says "I've hired this man to assist you" and you are happy to finally have some help with your difficult task. Then one morning you come into work and see that the coming day’s work was done overnight by this other fellow. And the realization sinks in, he isn't there to help you, he is there to replace you.
In this view his reaction seems totally expected. I expect many professionals, e.g. doctors, lawyers, accountants, etc. will make this realization sooner or later. The companies are saying these tools will assist professionals, but in reality they are aiming to replace them.
lol. lmao even.
I guess there is a first time for everything.
Regulated fields - another clueless poster lol.
I know many of you really hate whats happening but don't try and project it on other professions where the environment is different.
Society largely never cared about maths.
You are, of course, right that doctors, lawyers and accountants have more control over the environment they work in compared to mathematicians. But it doesn't change the underlying argument Tao is making. Your observation that those professions will react differently is likely to be true and only time will tell how effective their response will be.
But AI "assistants" is a marketing approach being made to all professionals at this time. Tao is implying this is a Trojan-horse strategy. If you let the assistants in without reasonable protections, you may lose control over the environment you are trying to protect.
Out of sheer, idle curiosity: were you actually laughing when you typed that out?
If anyone doubts this, consider that Anthropic considers its TAM to exceed $30 trillion which is nearly the entire GDP of the US and 25% of the world GDP.
Only makes sense if they believe they will massively replace labor. Which, to their credit, they used to admit openly and publicly and only stopped because some crisis PR people told them "hey maybe you shouldn't say the quiet part out loud while the hoi polloi still have the power to stop this from happening".
https://www.anthropic.com/careers/jobs
They have practically infinite investor money to throw at salaries (lost in the noise of how much they are spending on compute) so why not use it for now to reach their goal?
My thesis isn't that they have already achieved the ability to replace massive amounts of the labor pool, just that that has to be their goal given their stated TAM.
> If they want to flex the power of their models to automate human jobs, why not lead by example?
They don't want to flex this. They realized (after a false start) that it is better for them to avoid talking about this in public.
That doesn't mean their goals changed, they just realized that flexing before the goal is achieved could endanger the goal being achieved through political backlash.
Wouldn't throwing that money into GPU be better because AI is smarter and faster ? So they would reach their goal even faster ?
As one of the smartest people in the world, he is so naive to believe that those capitalists will stop at that.
They also had a rich starting point from open source code (which we somehow gave up without realising it). All of this won't happen as efficiently for these other fields, but it will happen, much in the way you've described. We (swes) turned out to be a rather dumb group of people, which is hilarious since we think we're clever.
The only moat we have is the human to human element, we need to talk to other humans in the business. Customers, etc. But once they're replaced by machines or automation, then machine to machine communication will be the preferred mode of communication and work and the legacy humans will be increasingly cut out of the loop. Human in the loop is a death sentence we just don't realise it.
> The companies are saying these tools will assist professionals, but in reality they are aiming to replace them.
They are openly saying that the goal is to replace humans and make humans obsolete. Like, literally openly. I am not saying I agree with that goal or find it realistic, but they do think so.
“AI won't replace humans. But humans who use AI will replace those who don't.”
vs
"Some areas in the job market will be 'just like totally, totally gone' as they're replaced by AI agents.”
and
“AI will probably replace most of the jobs people do today,”
The most recent position I could find says he has revised his earlier position on a total replacement due to the lagging adoption of AI in many industries. But I don't think there is a single position we can point to that they are "openly saying". It is fair to grant even the CEOs of these companies a change in position as they gain more information about capabilities and adoption, rather than insist that they hold, have held and will continue to hold fixed opinions.
The point is that Tao's understanding has changed. He may have been naive before, he may have heard in their words something he wanted to believe. Perhaps he was misled by people who he trusted that worked in the company. But now that he has more evidence, in the form of the maths drops and the way they have been handled, his suspension of disbelief no longer holds. The companies could have done this in a way that included mathematicians, so they chose not to. And that choice reveals their aims unambiguously in a way that a media-trained CEO or careful PR department can avoid.
That suggests that Tao is not being "self-serving" as others claim. He is saying that his optimism in the best-case scenario turned out to be incorrect.
Terrence Tao is famous because of solving problems better that most mathematicians because he has an uncanny ability to look across different math fields. AI gets everyone this ability for 20 to 200$ per month, eventually free.
Most mathematicians are happy to see interesting problems in their field being solved, and are looking to build on these to find more interesting problems.
Link: https://youtu.be/bAnwMdfcMVw
https://www.quantabooks.org/books/six-math-essentials/
Funny when it’s someone else eh?
Just learn a new trade Terry.
(I am kidding if it wasn't obvious)
> Of course you get a lot of right-wing people saying "oh these Poles; they come over here taking our jobs". I think if a guy from Poland arrives and he doesn’t speak the language, he doesn’t have any money and he doesn’t know anyone, and he takes your job on the first day – you’re shit, fuck off."
https://www.youtube.com/watch?v=t2ICmOYP8gs
Somehow I think "taking our jobs" concerns are going to be a lot more palatable.
If the pole is willing to work for a third of the price the local(I'm assuming Englishman), of course it is detrimental to UK labour.
Pretty good for the capital owning class, though.
I think this sort of progress, compounded over the next N years will serve to cleanly split useful math from the ego driven math you speak of.
Certainly there are enormously impactful pure math results yet to be found, but we have no idea really which directions to pursue.
I wonder if this is going to look like the chemistry-alchemy split in the rear view mirror.
There are plenty people in every discipline that prioritize their ego over measurable progress towards a stated goal - even less so a community focused goal.
In Pure maths, people are not generally solving problems because of some pressing practical need, it's purely out of intellectual interest, and if we're lucky there's some immediate practical application. More often, the knowledge sits in pure form for a while and years down the road some new invention, or practical consideration, or other field finds a useful application.
Just like literature and visual art, it serves an important epistemic and social function even in spite of its immediate lack of raw practicality.
Case in point: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
The information in the article suggests that he regrets his prior hyping of AI.
Is the cost prohibitive , or a proprietary model?
Given the trajectory of LLMs to date I see them helping construct proofs for conjectures but not devising new conjectures of any material substance any time soon. The challenge is, as Tao has said elsewhere, part of being able to formulate a conjecture is by exploring the landscape around it by doing proof research. Seeking answers finds useful questions. The fear is the tools will take away the exploration of the landscape. But that’s silly IMO. It takes a special type of lazy to let a tool take away your understanding that’s important for what you’re goal is, and a special type of passion to learn with a powerful tool. The passionate will adapt and progress will accelerate, not slow down.
For instance Tao has posted some of his interrogations of proofs with AI assisted reasoning and it didn’t seem to diminish his ability to establish insight rapidly, but to shut down avenues rapidly and to pivot and expand his understanding rapidly. The insight was all his own, as was the path of inquiry. The understanding of the output was his as well. He carries it into tomorrow and uses it to formulate new more interesting conjectures the next day. The AI ceased to “exist” as soon as the inference turn was over.
The hand wringing is pointless and lame. John Henry moves aren’t necessary or useful, and when I see a tunnel boring project now I see tons of people well employed there, just the holes are bigger and better, and they’re engineers instead of strong backed hammer slingers.
Because it comes across as "ooh, you're just going through the five stages of grief, dummy! Look I've taught you psychology 101. Now that you know about grief, you might as well skip to the next cycle already. Support me on Patreon. Like, comment, subscribe"
Pointing out that someone is "just" grieving is not helpful.
I'm seeing this a lot in the past months of AI-for-math discussions. Had to get this off my chest.
It's not at all clear whether the kind of math created by LLMs will retain the usefulness of the old math. LLMs are good at rigor but bad at intuition. It's not even really clear if the rigor was ever all that important. Theoretical physics is full of people using math in ways that don't meet the bar of rigor that mathematicians set for themselves, but they have enough physical intuition for what they're doing, so it doesn't matter. It's an open question whether post-LLM math will continue to supply scientists with the kinds of ideas that they need from it.
I think it's always better to publish preliminary results early (as long as you clearly say so, and they did), and let mathematicians decide which of those are worth focusing on.
No? How much did OpenAI spend on this and how many trials failed? I can’t imagine anyone not working for them can pay the absurd token costs!
Can you elaborate what does that look like?
And that multiply those few people by number of different problems.
And I honestly have no idea what Tao wanted to happen. How could you ensure that the AI "empowers" scientists and doesn't replace them. Intelligence is intelligence. If it's powerful enough, it will replace other intelligence.
Unfortunately, though, it is what you said. And more or less what you
> If one wants to argue something, it's that the Nexus should not have been built.
Or that it is being built in a way that is causing forseeable problems, while making no effort to anticipate and solve or mitigate those problems.
I suspect their internal harness, models, and thinking budget far outstrip anything us peasants have access to (never mind the compute). So saying that they achieve all these headline results with 3 hours of GPT Pro is disingenuous. If that were so, hundreds of math phd students would’ve already recognized it.
I’m coming to the realization that my monthly subscription is only helping subsidize their real internal AGI, and my interactions are a small amount of human feedback and creativity which they can strip mine should they choose to.
It's hard to prove, but I'm personally extremely convinced of this. The point of AI (from the perspective of these companies) is to solidify and centralize their control of the economy and the world.
Therefore the #1 focus of their research and efforts and expenditure will be something that they don't want any of us to get access to. I think many other behaviors like model nerfing follow from that. It helps explain these frontier research products which we ourselves cannot get access to.
The only thing that will matter is capital, which will translate directly into all the talking and doing needed to make more money.
Interestingly, the dynamics strongly resembles the industrial revolution when from highly skilled manufactures society moved to factories with power machines. And it all came down who owns means of production. Now AI is a new means of production for all white collars jobs and if the industrial revolution taught is the way to go, nothing good will come out of it.
I don't particularly like AI, but if AI replaces developers and other knowledge workers, not everyone is worse off, and I don't mean just capital owners.
would someone who runs faster 100m also somehow drive better?
I use LLMs to great advantage, solving problems in my domain. I no longer make for loops.
my point was that the new future where running faster doesn't matter any more might come, because everyone is driving (and potentially can't run much at all).
See the beauty of analogies is you can make one to support any point.
He described it at 'strip mining', 'rude', and said that many grad students, who are given problems like these as a part of their training, had the rug pulled out from under them (because its customary to let people know you are working on something to prevent this kind of thing).
And your whole analysis is that Tao is worried about his 'identity', status, and 'being a differentiator'.
This is projection on your part. People have values that bear no resemblance to yours and it would behoove you to learn to recognize that.
AI is coming for all of us knowledge workers, nobody is immune.
Disclosure: PhD, early 1990s. In my case, replicating my project would have required a sizable investment in equipment, much of it was purpose-built, among other things.
In math, proofs are "the coin of the realm," and could be devalued to zero. This leaves no answer as to how someone will earn a graduate degree. For instance, what kind of work will be accepted by a thesis committee in five years? The professoriat can only say "we will need to change some things." That's not good enough when the deadline for grad school applications is a few months away.
Strip mining what, platonic reality? Mathematicians don't have a monopoly on mathematical truth. It doesn't exist for their benefit.
No one has the right to prevent others from working on a math problem. If you're working on a problem the best way to not get scooped is transparency: publish whatever you have as soon as possible.
Tao's behavior is most consistent with any other worker resisting automation for selfish reasons.
> It's time for the social fabric [of math] to be ripped up
This is the attitude of a social arsonist. You can, and should, integrate changes without maximizing destruction for its (or Altman's) own sake.
> an industry being disrupted.
This is the SV euphemism for law-breaking and monopolist practices in a country that has forgotten how to prosecute those with the most power and privilege.
Quite the opposite. The mathematicians are the monopolists who are upset that new-comers are able to do their job better than them, and even worse, they are giving it away to the world for free!
The horror! People giving knowledge to the world for anyone to use freely! Won't someone think of the oligopoly that was coordinating to make sure each of the gatekeepers get their slice of the pie?
Mathematics is collaboratively and publicly produced, it is vetted for free by peers, and results are publicly available. There is no 'market' and no 'price'.
These results, like the invention of LLMs, are often used by private companies for the generation of private profit.
"Open" AI used a private, unreleased, for-profit model to generate those papers.
What, precisely, did the mathematics 'oligopoly' gatekeep you, or anyone, from?
> their slice of the pie
There's very little money in academia, and even then, its a struggle to get any funding at all. What are you even talking about?
Final point, 'knowledge' is socially produced as well. Until the papers are accepted, they are machine artifacts that might be knowledge, but that same system you disdain will be what incorporates and validates those insights.
How wonderful then! You should be happy to know that the AI companies are releasing information for free for all to collaborate on then!
> What, precisely, did the mathematics 'oligopoly' gatekeep
Well what they are trying to do is convince the AI companies to stop releasing all of their findings. Because they are upset that other people are able to do their job better than they are. Its the classic gatekeeper move.
> There's very little money
Oh there is much more than money to silly games like these. Instead, clout and the admiration of their peers, in this tiny insular group for sure, is what most of these people are so desperate to acquire.
Fame (well... relative.... in their subgroup) and admiration of their peers is a much stronger motivation factor for humans than mere money.
> but that same system you disdain
I don't distain anyone who is willing to get with the times. My distain is held out only for those who want to prevent the rest of the world from using the latest tools to achieve real results. If a mathmatician simply doesn't care and isn't wineing about all this AI stuff, I have no issue with them.
There are already several of these "proofs" that have been retracted because they did not make sense.
I would have no problem accepting this whole ordeal if the humans behind each proof took the time to re-read, understand and synthesize the result in a readable paper.
There is ample evidence of how damaging fake information can be, and paper retraction never fully undoes the damage (see the Andrew Wakefield and his antivax "paper" for instance).
[1] https://news.ycombinator.com/item?id=50036781
So there's that.
This is what Terence Tao's ChatGPT chats look like: https://chatgpt.com/share/6a5fdc7a-d6f8-83e8-bbea-8deb42cfed.... Can a layperson prompt the LLM like that?
"Garbage in, garbage out" is a polite way of saying the LLM's output is a reflection of the prompter's own intelligence.
At the same time it should be considered that with physical labor workforces lost their income source numerous times. Ironically they are currently being considered a safe haven.
We know it's a filter, and we know it's an idiot parroting Claude.
The last(-ish) chapter does have a section on how AI has robbed the "cognitive elite" of our identity.
It turns out it doesn't matter, because outside of a few narrow bands like Claude Code, no one cares what documentation an LLM writes for your project, because no one is ever going to go looking for someone else's project or library to solve their problem anymore.
So either an LLM decides to pull in your project as a library for reasons, in which the LLM can understand the documentation, or else no one besides the creator is using it anyway. So it doesn't matter.
Why would anyone use someone else's AI-coded project if they can get their own version that better fits their own needs in less time?
If OpenAI had never made a push to get good at maths, we might still have a sizeable group of peopne stuck at that first line. I say that because I still hear in discussions the naive rebuttal "But we'll still need _someone_ to {one-step-up-in-hierarchy}" . Although you have to be curious now how long that sentiment can last.
The good news for those doing abstract maths might be that, now that they've proven that AI can do the job in principle, there's probably not much left in it for them to keep going. Mathematicians should know better than anyone the lack of a direct profit motive in doing the work that they do.
This is a disingenuous framing. I've "built" more in academia than in some jobs I had, and some VC-backed startups are renowned not for building but for their lack of ethics.
"Building" is not the discriminating factor.
Earning money is not curelated with usefullness these days. It is somewhat correlated with lack of ethics.
Grigiri Perelman, who was awarded the Fields Medal, notoriously refused the prise because he stated no other recognition was needed.
There's a reason that so many open source builders and contributors are unemployed people in social-welfare European states. There's a reason that nearly every great work in the English canon was written by someone who was on the dole at the time, from Nietzsche to Harry Potter.
if you think that's the takeaway, you're not listening to any of the people affected by AI
By snooping on work math researchers were doing on their platform and then overtaking them when sufficiently close to completion? isn't that what happened i'm not sure
Finally, the list of hundreds of solved problems after that completely defeated any argument in that direction.
Using LLMs doesn't mean some easy mode where humans don't do difficult work
not saying it's not possible ai will continue to expand to more fields, but i don't think it's a coincidence that both math and programming are easily automatically verifiable. in fact for programming their effectiveness drops substantially once you remove any sort of testing or verification loop, and i imagine it doubly-so with math. try getting an agent to build something without any tests or verification, you'll see it go back to near-2024 levels. half of the work i do with agents is setting up a harness that can take and analyse screenshots, send direct input to the application, and compare results. this is something that should come standard with every harness.
if we can build automatic verification for more fields is a different and perhaps more relevant question.
If that is true, then only because "most people" have no understanding of what differentiates programming and math from "computation". Computers (as the term suggests) have always been extremely good at the latter, and useless at the former. It's only 2–3 years ago that computers, by way of LLMs, have started being good at the former.
Now I'm forced to think about it much more carefully. I don't believe that current LLMs can find every possible proof on their own from now on. I also agree that the PR returns have fully diminished. Given how hard it's been in recent years to fund e.g. a poorly paid postdoc, I doubt there's a ton of money lying around for getting the best models to work on mathematics problems. I think there's a chance we're heading into a really good era. I was able to make so much progress back when Google search was a useful research tool...
Over all, I'm not quite sure whether I see these deep cuts occurring less and less with progressing AI. It might just be, that I'm giving them progressively harder tasks as well. But the fact of the matter still is that they aren't there yet and it needs proof they will get there ("there" being able to do competent work that is not strongly limited in their scale by human review and correction). So I'm still not convinced about the point you want to make, even though I see AI disrupting about every part of society, and am hoping for the benevolent super AI to lead us to utopia (and afraid of the cliff of human dictatorship on one side and extinction on the other).
[1] I've started to use these moment to reflect on whether a human in a comparable situation would fare better. Having only a few text inputs and small glimpses into a code base and knowing next to nothing about you or your project personally. Quite a few times I've concluded: No, they would not. But more than 50% they would, and also, these limitations are technological limitations of LLM, so you should probably not excuse them for that.
So the pain points you and I frequently run into, simply don't exist for a well connected individual(Sam Altman).
True.
The other possibility is maybe OpenAI doesn't know what to punch into the calculator now.
They have gatekept for literally 100s of years. Decide the problems, decided who is good enough to have a chance at solving them, decided what it means to be solved.
All without meaningfully contributed to human progress - beyond teaching calculus and lin al to future machine learning scientists - for 40+ years.
Every other job in the developed world has been disrupted by technology over the past 100 years. For many of them - multiple times. Now it is pure math's turn, and they are acting like they always do - elitist, entitled, worthy of special consideration.
They are not.
You are right, we should get rid of their profession
But to be honest the reaction from the ivory tower here is seemingly quite telling to a layman watching from afar.
It looks far more like this was some extremely gatekept community where you were only allowed to work on sanctioned things if so deemed worthy. Better not step one someone’s toes lest you get exiled! How much progress was lost due to newcomers never working on a problem due to told not to?
It’s starting to make those old timey Usenet kooks look a whole lot less crazy to me. They seemingly were right about this part.
Mathematics was funded and supported by a whole bunch of janitors and blue collar workers busting their asses to let entire departments of thinking men and women work on hard problems for the betterment of society. If it was just a super special friends club the whole time that was gatekeeping what amounts to a hobby, it’s going to me much harder to justify such a social contract.
Honestly not even sure what to think at this point. Extremely disappointing to me the more I read from these people.
Admittedly it's also just really hard to come up with something better than the current track, but doing better has to start with being eager to explore different directions. Even if the new directions don't quite work just yet. And that eagerness just isn't there in the community, new directions are threats to the careers of sensitive people
What are you even talking about here? Mathematics was funded by janitors?
How do you like them apples?
That's a lot less than i expected. The value he gave them should be at least 2 magnitudes higher.
For math, just because we know a thing is "TRUE" doesn't mean the investigation is done. The folks that love math are driven by solving puzzles, and the puzzle is still there, only changed: instead of being "is this true?", the puzzle is now "why is this true?" Building theories, finding nicer proofs, etc: there is still plenty of room for creativity.
But in the meantime, it really is tragic that so many young minds have had their work stripped away from them.
And the economy puts no such large incentives on much of the remainder of economically valuable work being purely human, quite the opposite actually. And instead of regulating the AI industry (which produces the "cheating tools"), they get trillions of investment. So this seems to be going in the opposite direction where it is actively worked against humans staying economically relevant.
I have faith that when AI settles, it too will open more doors than it has closed.