> Mathematics need to redefine their profession and how they work (like us software developers too). There are very good reasons to believe mathematics has an important role to play. If they could just stop talking about OpenAI and get back to work - they are very much needed, in particular now!
The root issue is OpenAI et al.'s thoughtlessness in their engagement with a field.
OpenAI has resources.
That they fail to allocate enough of those to cleaning up pre-print papers (that seem to be a corporate PR priority for them to release) so they can be consumed and engaged with by the field they're targeting is... acting like a jackass?
It's the same "Meta / Alphabet can't vs won't hire more human reviewers" problem.
OpenAI could, at an immaterial salary level to them, pay a ton of PhD students and mathematicians just to clean up their proofs and papers.
Not doing so is a leadership and financial choice.
I'd prefer AI companies don't decide who's "cleaning up pre-print papers". This should remain the job of mathematicians at universities, which I am happy to pay with my taxes. Ideally there was something like a Bermuda Principles declaration for mathematics (https://en.wikipedia.org/wiki/Bermuda_Principles). This gave mathematicians even at poor universities and beyond the chance to participate in mathematical progress.
What do you think is gained, if AI companies manage "cleaning up"? Tax money?
No one is "responsible". If the paper is read, depends on the interest of the individual scientist. Many mathematicians interested in a theorem proven/disproven in a new AI paper will be curious, even if it is utterly cumbersome to extract the relevant line of thought. But don't you agree that the job can only be done by a mathematician anyway - be she/he paid by the company or by a university?!
What you are arguing for is somewhat like the "proprietary period" in astronomy (e.g. see https://www.scientificamerican.com/article/nasas-plan-to-mak... ). But there is no mathematician who asked for the run of AI, i.e. there is no mathematician, who can be regarded as the owner of the result, even if ownership was temporary. Complex proofs might take years to explain in detail (think of the ternary Goldbach problem, https://en.wikipedia.org/wiki/Goldbach%27s_weak_conjecture ). It's just unfair, if AI companies held back with their data until a proof has been put into a nicely readable article, and at the same time mathematicians elsewhere are spending all their time trying to solve it.
I'm not. To the extent my taxes are used for maths research (which is minute as a proportion of them) I want them to be used for the development of human mathematical understanding, culture and education, not trawling through a mountain of slop mechanically generated by a Silicon Valley startup that's about to IPO for trillions in the hope there might be some nuggets of insight hiding in there. If the latter is an activity with some value the startup in question can pay for it.
Addendum to risk: larger companies can throw bodies at escalated issues.
Good luck trying to get a 6 person shop with a full calendar of other-customer work to suddenly prioritize a show-stopping bug that's only affecting you.
Oh totally. This is a playbook for a Fortune X00 company. Layoff your staff and create a "services" SLA contract with a third party company to deliver IT services. Just the legal team to get thru all the docs alone means you need a legal and procurement/contracts department. Not for startups.
Ironic phrasing, as in my experience the steelman argument in favor of these companies is "The customer is really bad at thinking through their own requirements."
Part of the problem is that the US has no unified empowered cybersecurity regulator that I'm aware of.
As a result you get a mishmash of DHS + DOJ + SEC + HHS agencies.
If there was one single throat to choke that had remit for cybersecurity + ability to being cases and assess fines, they'd be handing them out left and right.
This is the elephant in the room that OpenAI and Anthropic specifically aren't talking about -- when "good enough" is surpassed.
Because there are a large number of use cases (and large sub-portions of others) where genius-level AI isn't required.
Which means once that threshold is surpassed in people's relevant fields, available margin on that is going to collapse to commodity levels.
It's difficult to see how either pure-play AI company maintains its valuation once that happens. Their TAM is based on capturing a big chunk of all work, not just that which requires the highest intelligence.
> First, the form factor of GPUs in data centers aren't the same as desktop GPUs, so you couldn't use them even if you wanted to in a normal rig.
With eGPU PCIe 3.0 or 4.0 x4 links (or even Thunderbolt), that essentially doesn't matter for inference.
You pay the bandwidth hit on model load / unload (a few seconds), but it's irrelevant for post-loaded inference.
Anyone hosting a high wattage GPU is going to be fine hosting in an external enclosure, most of which have generous extra-spec room.
And if they don't, if a flood of cheaper DC GPUs hit the used market, you can bet Chinese manufacturers will have enclosures that fit them available the day after.
US health care is screwy, because one of the equitable things it tries to create is equal price regardless of age/sickness.
As a consequence, insurance companies generally try to sign up mixed-pool, younger, healthier (read: employee plans) over older, sicker, opt-in insurance (read: ACA).
There are ways they can and can't legally do that, but it's always the goal. In the same way that retail order flow is targeted by HFTs.
reply