A misalignment of AI in mathematics • Hacker News Board

Re: A misalignment of AI in mathematics

#791

Post by tmhn2 » Sat, Sep 12, 2026, 4:09 AM UTC

Shaptse wrote: post #418

reasonableklout wrote: post #187

Earlier quoted context omitted.

Tao’s key point is that the way people are using AI today does disrupt that. Because people and companies are valuing the results over understanding. So we’re getting slop results rushed out that are automatically verified. Moreover in the past, discussion and idea sharing would happen naturally to overcome the friction of the process. But now when OpenAI is stuck on a particular part of NS for example, they can just…

I won’t speak for Tao. But it is not “people” who have valued the result over understanding. “People” are a bit impressed that the machines are as good or better than the priests who have been praying at the inscrutable altar of pure mathematics. But it is mathematicians—I am speaking as a mathematician—who have adopted a culture of valuing results over exposition and transparency. The field has been rife with extrem…

I don't think that's entirely fair.

I think it's OK that there are some materials designed for a specialist audience, and some materials designed for a wider audience. Technical density serves a real purpose in the former (I'd like to just say "stack" without including an explanation 10 times longer than the rest of my paper about what a stack is and why we are talking about them), and some people expend a lot of effort on the latter (think lectures, lecture notes, textbooks, seminars, blogs, etc--totally appropriate to dig into the motivations here).

It's just the fact that it's a really vertical field, not some cultural failing, that results in pretty opaque stuff sometimes.

Re: A misalignment of AI in mathematics

#792

Post by demibabs » Sat, Sep 12, 2026, 4:14 AM UTC

nmeagent wrote: post #773

pvab3 wrote: post #714

Earlier quoted context omitted.

that is incredibly depressing

Yep. Perhaps humanity would be better off if we instituted and enforced the notion of "Thou shalt not make a machine in the likeness of a human mind." It's worth thinking about; just because we can build AI systems doesn't mean we should.

I think it would be depressing in its own way if this technology just ended up being regulated out of existence.

Re: A misalignment of AI in mathematics

#793

Post by lll-o-lll » Sat, Sep 12, 2026, 4:16 AM UTC

noslenwerdna wrote: post #760

lll-o-lll wrote: post #528

Earlier quoted context omitted.

> But there's still play. There's still curiosity. And there's still the drive to understand something for yourself. Yes, but think about what that implies if those are the only motivations left. Gone are the professions. Gone are the ambitious. There is plenty of space for people to work on intellectual pleasure pursuits (as there is with art and music), but the death of all intellectual based industries is still so…

Good? Those people were in the field for the wrong reasons. No reason to mourn them leaving

> in the field for the wrong reasons

It’s an interesting view-point. What are the “right reasons” to pursue an intellectual discipline? To selflessly further the total sum of knowledge?

> No reason to mourn them leaving

What field? Who’s leaving?

The conclusion has to be that human involvement in these pursuits is facing existential crisis. “Teach me, oh Oracle, the mysteries of the Universe”. “For a fee”.

Both the glory and joy of discovery now belongs to the machines (the owners of those machines). Perhaps that’s fine and well, but I read Anathem and feel the strong yearning to be the intellectual monk. And yes, to fight and debate and think; and obtain my place in history. The history of man.

Re: A misalignment of AI in mathematics

#794

Post by tmhn2 » Sat, Sep 12, 2026, 4:18 AM UTC

jrflo wrote: post #160

tmhn2 wrote: post #148

Earlier quoted context omitted.

At issue is the fact that it doesn't typically work like: mathematician produces a proof in isolation, generates a PDF, and shares it with a bunch of people. There's a whole culture and community going on behind the scenes with conferences, seminars, lectures, chats in the hallway, advising students, etc. that AI-generated proofs bypass.

I don't think that AI would disrupt any of that. Even if AI solves a problem, you can still discuss the methods at conferences, seminars, lectures, chats in the hallway, advising students, etc.

Hopefully that's true. Instead of bypassing the community maybe it stimulates the community. I think it does disrupt things a bit, but I'm optimistic at the moment!

Re: A misalignment of AI in mathematics

#795

Post by reasonableklout » Sat, Sep 12, 2026, 4:24 AM UTC

redox99 wrote: post #754

ekidd wrote: post #720

Earlier quoted context omitted.

> Those who think it's me or the machine will fail. > Those who realize how much you can accelerate your research with the help of AI will succeed. This is only true up until a point. If I treat a mid-sized model (say, Qwen3.8 Flash Next) like a pair programmer, then yes, it accelerates my work. But I can already see the next stage with Fable: If I give it a couple of paragraphs of spec and $50, then I can just leave…

Of course you're not going to get rich with the kind of software that LLMs can one shot these days. But that kind of software like to-do lists or basic CRUD have been saturated for over a decade, way before LLMs. People overestimate how much you can one shot, yeah a good prompt can get you 90% there but that 10% remaining often takes months of extra work. Software has always progressed this way, lots of devs back the…

But what does the human researcher do in this future?

If they are not needed to understand the result, then what's their role? Asking the right questions? But how will they know what questions to ask if they don't have a deep understanding of the domain earned by sweating the details themselves?

And how long will they be needed to ask the right questions, how long until AI can do that too?

> Software has always progressed this way

These platforms took on the order of a decade to mature, during which people had plenty of time to learn what's next. With LLMs we went from one-shotting functions in 2025 to entire projects just over a year later.

What do you think you will be working on in a decade?