you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 24 May 2026
17 points (94.7% liked)
TechTakes
2586 readers
91 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 2 years ago
MODERATORS
Look, I admire professional math researchers as much as the next guy, but they're literally balancing on top of the economic pyramid of needs. This is a profession that cannot exist outside a culture with a surplus that can afford higher education, because there's no economic incentive to hire them to do that kind of work.
So GenAI can solve Erdos problems. Can it unplug a drain or repave a road? Fuck no. Can it handle refuse for a city? Also fuck no.
Maybe professional math researchers will go the way of professional typesetters. I think humanity will retain its relevance.
His drama knob is always set to 11.
See also the previous article and how he talks about the "[fucking] NYT." Or better, dont look at that article.
isn’t that the old “basic science is boring and unsexy” issue though? There are economic incentives, but not in a short term-big-bux sort of way, so capitalism can’t be trusted with it.
To conjure up a recent example, something like “The number of curves of genus two with elliptic differentials”, published back in 1997, probably had limited commercial value at the time, but 20 years later completely sunk a promising post-quantum cryptography algorithm (“An efficient key recovery attack on SIDH”) which might have had some non-trivial commercial implications if SIKE had got through the key exchange algorithm competition.
Anyway, the Erdős problems are good candidates for llm work because they have been specified in a careful and formal way, which requires a reasonably competent mathematician to do. That then opens up mathematics to the same deskilling problem that other sectors afflicted with llms have, and because capitalism is shortsighted and stupid we don’t know what the future economic impact of that will be, right?
yeah I was aware of the value of pure math (specifically the discrete stuff for crypto) when I wrote it
my point is the Scott-A is massively overvaluing the societal worth of pure mathematicians. Assuming (I know, big ask) that AI can succesfully automate that field, humanity is not worse off with regards to outcomes. Humanity remains relevant.
It's a weird moving of goalposts, not often commented upon, that GenAI is succesful mostly in replacing (in the Ersatz sense) stuff that's not really foundational to human society. We don't have robot cars, nor are we actually close to getting them, which would actually transform society in a massive way. Instead we have robot copywriters and bespoke porn creators.
but for people like Scott-A, whose entire selfworth is being "smarter" than anyone else (not withstanding that he would be eaten alive in the post-apocalypse), being replaced by a robot is not just about losing your livelyhood but your self-worth as a human as well
I think pure mathematics is as valuable as the humanities. Unlike many stembros, I think the disconnect is that we vastly undervalue humanities, not that pure mathematics is overvalued. Agreed Scott is probably overvaluing them.
I am on the same page.
I think the challenge is that the value of results of any kind of basic research are so wildly variable that normal rational economic thinking stops working. In Nassim Taleb terms you're actively seeking black swans in a world where everyone knows all swans are white. Sometimes you venture into the depths of the rainforest and come back with a revolutionary new medicine, but most of the time you're gonna have a few cool pictures of new bugs or something - not without value in the real sense, but hard to capitalize and transform into profit. Even if you end up discovering/creating an entirely new framework for understanding life itself that revolutionizes everything from agriculture to medicine to politics in the following century, that doesn't necessarily work in the specific context of economic rationality - who remembers the name of the guy(s?) who funded the Beagle? And sometimes, as you referenced, the cool bug picture doesn't have an obvious or immediate return but ends up being the important piece of data in a different context decades down the line.
This is a field of human endeavor where the economic best-case scenario is probably Bell Labs. And despite having an absurd number of patents and prizes they still couldn't survive within being largely a vanity project for the original Telco monopoly. The ludicrous returns that came from repeatedly revolutionizing electronics and computing couldnt justify their position on a quarterly balance sheet.
The Beagle voyage was a Royal Navy project, and it had a defined purpose: charting. Having a young naturalist onboard was rational because what if you found the next tea plant?
The captain of the ship remained an implacable opponent to Darwin's later theories.
I believe Bell labs was mostly a fig leaf for covering up Bell's legal monopoly (i.e. look, we're doing some good stuff with all the money you're legally obliged to pay us)
there is a degree of risk that is acceptable for business. 90% drug trials fail; if you are inventor, you make a startup, package your pre-trial drugs and associated IP there, then pitch it up and cash out. vcs have money for clinical trials. sometimes you have phase 1 results that you got on your own too. the further it goes in trials the more it is worth; result is the biggest gacha in the town
Not a pro mathematician so apologies if Im hanging my butt out here, but I'm not convinced these solutions (or something analogous) aren't in the training data. Wouldn't be the first time. Several seem like the kind of thing you might figure out for a lemma on the way to something else unless you'd previously examined tHe ErDoS PrObLeMs.
Note: there are over 1200 problems that could be arguably labeled "Erdos Problems." It appears that a chatbot enjoyer vibecoded a website in 2024 and was somehow declared the chief keeper of the things. The site can be found here: https://www.erdosproblems.com/faq
The relevant part:
You fuck cows in retrospect
More than anything else, the AI-rdős Problem cottage industry has damaged my trust in mathematicians. First of all, just how bad does a company have to be before you boycott their products? Just where is your line? Because the industry passed my line about seven thousand lines ago. Second, we know that in other fields, the output is shit, that people brainfuck themselves by counting the hits and forgetting the misses, that users de-skill themselves through slop dependence... What makes you so special? Piping the output through an automated theorem-prover, or any other hack to improve the reliability of the stochastic text extruder, can at best shift the probabilities.
Its the “expert in one field = expert in all fields” stuff that I hate. There’s still plenty of jobs and areas of expertise that require humans: medicine, psychology/therapy (AI has arguably made the need for human therapists way more important), physical labour like construction work, linguistics etc. Hell human authors are definitely keeping their jobs since so many people hate slop
Relatedly, I think another part of the problem is the implicit assumption that 'able to do one narrowly defined/narrowly constrained type of problem within a field' = 'expert in a field'.