Posted by aaraujo002 6 hours ago
It's akin somewhat to "donating 100 dollars to the node JS foundation, not the packages that appear when you run npm fund".
A non-zero amount of it is probably going directly to the AI companies through tokens...
Much the same way that VC money has been funnelled to Google and Meta through advertising budgets.
If it doesn’t pay for non-permanent staff salary that would be sad for the maths ppl indeed.
OTOH I have a phd in th physics, 15+ y of academic senior researcher experience, now working 60+ h /w on an unpaid internship and using lunch money to pay for tokens.
And I’m by far not most unlucky
Edit: just to put things in perspective, my phd field (string theory) has 15-30M€ annual funding in Europe (per google)
Mathematicians are in the same boat as Silicon Valley programmers in the end: AI is doing their job better than them and cheaper. Other professions will join shortly.
The question then becomes, what do humans do? Resigning to die quietly is hardly the most favorable option?
Just because somebody else can do something better than you ever could doesn't make doing it superfluous. There is more to mathematics and software engineering than finding proofs and algorithms.
Cheaper than them? That's dramatically unproven and unlikely just on power efficiency alone, even if we accept the suspect assumption the only efficiency anyone should care about is currency marked per line of code or theorem proved.
We're not just going to have to become better existential philosophers, we're going to have to become better examiners of assumptions and claims.
Pretty sure it was Tao but I could be wrong. In any case it now seems more defensible than it did when the idea was first broached.
Why? I thought we already established* that most professional mathematicians are not necessary (in the near future, then maybe "any"), so why would they doom these students with pity grants to basically create unnecessary jobs for them? More practically, wouldn't that money be best served (i.e., raw ROI) by purchasing more tokens on the frontier models?
(*) The existence of this effort itself is a good heuristic.
But that doesn't sound like a good argument to ride off into the sunset, right? You probably enjoy living and it does make a difference in the lives of some close friends and relatives?
We don't organize societies around ruthless efficiency optimization, we organize them to be at least somewhat hospitable to humans and human enterprise.
Even if you want to work in research, AI offers a lot of opportunities to do even more research, and less grunt work
I hope the future you built for others is imposed on you too.
I haven't escape note that I can do tons of things with AI because I spent the last 10 years building a bunch of side projects and now I can complete those side projects because I know all the magic words.
Gonna need to unionize because the gerontology is just getting stronger.
It's also not an OpenAI, Anthropic or any specific name caused problem. This type of progress has always been happening, it's more or less coincidental (like Stigler's law). The problem though is we need to rethink how we operate as a society and too few (IMO) are trying to make this conversation happen.
It doesn't help you that AI understands it, when you don't. You are the one getting screwed over by others who understand statistics better than you do.
When you're rich enough, you can pay others (or AI) to protect you. But then you're dependent upon them.
So, under a transactional model of social relationships, you are in an arms race over raw intelligence and competency/data. With only one winner, who takes all in the end.
Logically, most people have already lost in that game, so why keep playing that nonsense?
Let's say GPT-23 Astra grinds away for a full month one day and out pops a warp drive. Our best physicists ask it to explain. The resulting stream of text is essentially incomprehensible. But, they build a scale model at its direction to test and sure enough it does the Kessel run in 12 parsecs.
Are you saying we shouldn't hop aboard the USS Enterprise and check out Alpha Centauri because we don't understand the math?
I think what's really happening here is that there's an ivory-tower academic class that feels entitled to make their living at public expense thinking about and writing about their topic of interest, and along comes a new tool that's changing that dynamic up. Meanwhile, in the real world, the value remains with people who can take whatever the tool of the day is and do useful things with it. Which, I think, is why despite all the talk the only professional job losses blamed on AI appear to largely be companies that had a failed strategy of some sort or another and used AI as a boogeyman to do the layoff that they needed to do anyway. To people living in the private sector this isn't seen as a dangerous invader, LLM's are productivity multipliers.
In the real world, the frontier labs are still hiring lots of humans for business and software roles (https://openai.com/careers/search/) despite being better positioned than anyone to automate them away with their world-beating models and save a bunch of salaries.
For automated AI shilling ...
* Lots of people use math for calculations without having proven any theorems.
* You can understand a mathematical statement without understanding or even having a proof. There are proofs built on the axiom of choice and conjectured results.
Yet this has zero utility (other than your own satisfaction). Once you understand something, you will likely not even do a better job at explaining it to someone else, so the utility is strictly limited to yourself.
So even ignoring ability to understand the problem is the cost/efficiency of understanding (and applying it).
I agree with your other statements, but I feel you're suggesting the non-sense is scoped to just math. It's not it's almost everything that we do (exceptions are social interactions including care taking, etc.)
So in this frame your "nonsense" is (almost) everything that we do. Personally I'd like to keep doing things.
John von Neumann
No one wants a data center in their back yard, but in the private sector I think adoption isn't as political as some make it on Hacker News.
There's no particular reason to believe that solving math theorems generalizes to engineering or robotics without a whole lot of additional work.
Never understood this sentiment. Theses top researchers could do almost anything else with their skills outside Academia and make far far more money. You don't get rich off grants... They could all presumably walk into Anthropic or OpenAI for near $1M a year in TC.
If you yourself are as cracked as you think you could also do the same...
Essentially gifts for the annointed.
Math proofs are right in the sweet spot for training LLMs with RLVR. It isn't some general "LLMs are getting insanely smart!". Even most software development itself is miles away from being as well suited to training as math proofs are.
The majority of math research is akin to professional chess. Completely useless and the folks doing it are using their giant brains to play a game that no one outside themselves benefits from.
Mathematicians also decide what to prove in their research - taste is not a skill we have RLVR'd away (yet, not claiming this is fundamental).
Of course, the field is being fundamentally transformed, it will probably end up looking more like experimental science in the future.
All I'm saying is the math advances of late really hasn't changed much outside of math as far as AI is concerned.
The foundation was started in 1994, Wikipedia says "one of the largest charitable organizations in the United States with assets of over $5 billion in 2022" - Jim Simons founded the Renaissance Technologies hedge fund.
The foundation was an early backer of machine learning for science, including funding the development of the Lean proof assistant, which has revolutionized how mathematicians work and made AI-assisted proofs possible. Additionally, the foundation has convened workshops, lectures and collaborations that have brought mathematicians together to explore and discuss AI’s role in the field.
Second, the whole point of arXiv is open access. It didn't use to be this way. Research being disseminated as widely as possible is a good thing.
[1] https://www.simonsfoundation.org/news/from-trust-to-verifica...
Quanta magazine is the propaganda arm.
The headline in this piece made it look like maybe a foundation was trying to help mathematician's push back against the unwanted assault on their profession by the AI labs. The section I quoted shows that's not the case.
Also not sure how you interpreted "foundation started by the Renaissance Technology hedge fund billionaire" as a mindless boost.
I'm going to defend my comment. I said:
> The obvious question here is "who is this foundation and are they working with the AI labs". I answered both.
You said:
> Respectfully, you didn't answer either.
My first comment started by quoting:
> Under a partnership, the Simons Foundation, Anthropic and OpenAI will commit $15 million in combined funding and resources
That answer the question: are they working with the AI labs?
And then I said:
> The foundation was started in 1994, Wikipedia says "one of the largest charitable organizations in the United States with assets of over $5 billion in 2022" - Jim Simons founded the Renaissance Technologies hedge fund.
I think that's a pretty solid starting answer to "who are they", and it tips people off that Wikipedia can answer in more detail.
It also leaves opportunity for people like yourself to speak to Jim Simons' reputation and fill in more gaps.
OAI and Ant got blowback for essentially strip mining mathematical discoveries into PR wins for their models, accidentally laying waste to the research field in process. Throwing ~$5M each into PR whitewashing (partially in token credits), seems self-serving.
Kudos to the Simons Foundation for seizing the moment to offer the labs a fig leaf of respectability. Might have been the fastest "Yes" on a grant proposal ever.
It’s like saying the power loom laid waste to the field of textiles.
Ironically, that has been that way for most people since forever.
Comparing with the situation in programming, where code evaluation (obviously analogous to "understanding") is that bottleneck now, should get people the idea of being in the same boat.
It is my understand that there isnt a lot of innovation here. Rather the (broad) field of mathematics got mined for solutions that already existed within the corpus.
It implies "We didn't know we already knew how to solve these problems". How is that a good look for mathematics as a field?
I think we’re still trying to figure this out. Once the proposed proofs are verified and deciphered I think we’ll have a better answer.
For example [1]:
> In many fields and activities, years of training have traditionally served not only to produce a final answer or product, but also to develop understanding and the ability to formulate new questions and ideas. However, building on a vast body of previous human work, AI systems are becoming increasingly capable of producing the results of such work directly, and these goals cease to align.
Or [2]:
> Use of AI tools, when done responsibly, can support not just the initial generation of a solution, but all of the valuable followup activities that traditionally flow out of that solution. But at the current time, the opposite is often occurring: problems are being solved autonomously by AI prompters who have no interest in the broader field itself once their initial target is "solved", and do not understand the AI output well enough to answer questions on the result, give talks, or otherwise interact with the rest of the field.
> ...
> solutions to open problems are now being harvested at large scale in an unsustainable fashion, leaving entire fields of mathematics much less fertile than when such problems were solved in the traditional "Math 1.0" fashion.
This applies to all fields and pursuits where AI is effective. Yes I think about this with regards to various things I’m working on.
This is just the nature of things, you need to adapt and not throw a tantrum like a child. You need to not behave like Terrance Tao who is essentially like a child in an adult’s body in terms of his behavior and lack of wisdom, and keep in mind he was a child prodigy so this isn’t a knock at all or statement about his intelligence, only his development as a man.
You can mourn, as I have for the field of programming as I knew it which I have loved since I was a child, but you need to move on.
If AI can actually replicate mathematicians, sure. We don’t have evidence it can. We have evidence it can do useful stuff once promoted by human mathematicians. But if all the Terence Taos “move on,” we won’t have those in America anymore.
(The irony of ironies would be the mathematicians prematurely moving on causing American AI progress to stall in the future.)
Best I can tell the biggest complaints are really just that the future of their career is largely combing through lean and they don’t want to do that. Too bad. Adapt or get off the pot.
Mathematicians need to completely rethink the prestige paradigm. Counting proofs no longer works. That doesn’t mean there’s nowhere for them to go. Now it’s about gleaning human understanding from the lean. Maybe in the future LLMs can do that too and it’ll be all about formulating new conjectures.
I’m in the funding markets around this stuff. Nobody took this into account. It’s more for PR with government and potential hires (and maybe some corporate marketing).
Softbank's rally began in Japanese markets on 3 September and American markets on 4 September. OpenAI announced on 8 September. When OpenAI delayed its IPO plans on 13 September, Softbank sank back to its 7 September level [1].
[1] https://asia.nikkei.com/business/markets/softbank-group-plun...
Jeeze, that's an almost impressively charged way of saying "private company performing SOTA research on open problems that anyone has been welcomed to approach for over a century" lol
I think it does a disservice to the dialogue to strip away any nuance here and just paint these companies as evil for doing what everyone has been anticipating companies like them doing well before LLMs even existed. It's clearly just trying to fit these events into a specific narrative which does nothing to actually address the inevitability of everything that comes next. Feel free to pressure OpenAI and Anthropic to stop performing math research so that it can just be picked up by whoever catches up and doesn't care about bad PR. I'm sure Musk with Grok will be extra considerate of pure math researchers when he sees an opening to pick up those scraps lol
In math the reward is finding the decomposition that reduces the problem to simpler cases, and in many cases that reward is very far. Furthermore, the LLM may process the many cases fast, but the decomposition often comes from a human, and the model takes the credit for the proof.
This foundation could provide the LLMs with the decompositions, generated by human ingenuity, that they need. But once supplied, the model can run the cases and get the credit. Hence more credits for LLMs and fewer jobs and worse prospects for humans, even if it isn't an intended consequence.
On ChatGPT: Settings -> Data controls -> Turn off “Improve model for everyone”
On Claude: Settings -> Privacy -> Turn off “Help improve our AI models”
They can scrape your work after you publish it.
Now that things are settling a bit after the Navier-Stokes and theorem-dump shocks, something like what mathematics will look like post-AI is starting to emerge. Of course, things should be conditioned on the assumption that capabilities will continue to improve, but imagine something like this--
An institute with a tremendous amount of compute, either rented or owned, built on the Humboldtian model of the unity of teaching and research. The main draw for big names would be the access to compute, free of the fear of being scooped. New forms of attribution and collaboration would develop, and we'd gain some clarity for students and PhDs on what a future career looks like.
The current melange of math university departments all following their own paths through the crisis without consolidating resources makes the transition more difficult. Think of the path physics took beginning with scattered cyclotrons (the basement of the Niels Bohr institute was even dug out to put one in), then the larger labs that followed. That transition was gradual because physics led it from within, whereas with mathematics every university department is suddenly at a massive compute disadvantage with respect to labs that aren't following the same model, and are driven by different incentives.
It could be this simple--maybe we don't need to completely throw away the old model, attribution practices, and so on, and maybe we can preserve the transfer of knowledge to the next generation of mathematicians, if we simply give them lots of compute.
An analogy to physics to highlight how the divergent interests of the labs could be managed in a state of "peaceful coexistence" might be: physicists got well-funded accelerators and labs that followed standard academic practices. These were separate from "war work" (Los Alamos) that followed government priorities, even with the crossover of personnel. You could have the same thing with a "math accelerator" working on the academic model, while the AI labs follow their own priorities, and some subset of people cross between.
It might be objected that this is merely a conservative attempt to preserve the existing academic conventions and practices currently under threat, but tenure itself already does that, and is arguably the very thing that exacerbates the problem for young mathematicians.
(Yes I know it's not the group directly receiving the founding)
He was also a major backer of autism research, because his family member had it. People spend money to where their passion is.
Mathematicians are not dummies?
In a sense, yes. There is no Simons Foundation for laid-off programmers. Taxi drivers and truckers, meanwhile, are really just trying to block change.
> wouldn't that make them the group that needs support the least?
I don't know. From a practical perspective, I am more confident the world will need mathematicians in 50 years than that it will need coders.
They may be a vanguard.
> find it hard to believe that AI doesn't represent a net potential benefit to human understanding of mathematics. Yes it can be misused
Have you read Tao’s criticism of AI in mathematics? The concern isn’t mis-use, it’s ecosystem collapse.
A good analogy would be offshoring. If AI causes higher mathematical thought among humans to stop happening for a generation, we could wind up in a state where the field stops progressing because AI is unable to generate new interesting questions or cross-link findings into practical applications.
More directly, the risk is that happens in America but not in a more-thoughtful country.
Just because AI is capable of churning out vomit inducing prose, doesn't mean we need to surrender to a future filled with it. Humans who actually care about what they're doing will use AI when and where it should be used. I trust the intelligence, curiosity and tenacity of mathematicians to learn to work with AI. I don't understand the sky is falling worst case scenario Tao and others seem to be taking as the most likely future.
…yes. That’s why the human mathematicians are having this debate. How to keep the human mathematicians and their founders motivated to keep doing math, potentially more productively with these new tools.
If AI saps motivation from some mathematicians, I question where their motivation came from in the first place. Maybe they're just too wedded to the traditional way of doing things and aren't fully appreciating the actual productivity gains AI makes possible.
Again, have you read Tao’s recent criticism of AI in mathematics? Nothing you’re saying is disputed.
> If AI saps motivation from some mathematicians, I question where their motivation came from in the first place
Question away. I care about results. If a substantial number of great mathematicians say this is a problem, and a substantial number of students agree, whether they’re unappreciative of their potentially pending unemployment is irrelevant to both that fact and the fact that they’re not going to be producing mathematics. (Again, here. The American mathematical ecosystem collapsing doesn’t mean the same will happen abroad.)
We don’t know this! We might just have no more American mathematicians because folks who aren’t mathematicians don’t understand that we’re turning mathematical breakthroughs from just a couple decades ago into real, working technology.
> It's not even clear if the human race can survive until then
It’s pretty obviously clear the human race can survive for hundreds more years…
"Climate change is going to kill everyone in 50 years so we'd better work fast if we want to burn all this coal by then"
Not in the same extinction-class event way. In large part because hiring developers is still profitable.
Also, not remotely as well organized or proactive as mathematicians anpparently are. So maybe less vanguard than the cannon fodder at the front.
You have to actively design society to accommodate for the dangers it poses.
Of course everybody should "learn to use AI", but the real questions here are wildly different: can everybody use AI? Do the big US companies succeed in capturing the system, effectively shutting out every competitor and how do you imagine to benefit from that? Etc.
knowledge workers are building their own gallows and it probably can't be stopped.
they’ve only had a few weeks to deprogram so I’ll let it shake out just a little longer