Post

Clément Canonne
@ccanonne.github.io

In light of the current funding situation (worldwide), a modest proposal: instead of pouring billions of dollars into GenAI claiming "it *could* accelerate science and research," consider putting 1% of that amount in what *will* accelerate science and research. Namely, funding science and research.

8 reposts70 likes

int_ijk
@tbrunner.bsky.social

and whatever da phuk means "accelerate sci & res"??? any metrics available here or just more colorful smoke? a longer critic is the usaf's study of nazi av gas and pilot training budgets. the nazis were always short on av gas, so they cut pilot training flights. the end.

Joshua Otti
@joshuaotti.bsky.social

Yeah, science is cheap. Especially math/theory. It's crazy undervalued because it's hard to understand. In terms of return on investment, the median may be modest, but that tail is LONG. Like, basically everything cool we have long.

Joshua Otti
@joshuaotti.bsky.social

As a software engineer who uses LLMs every day and a recent CS theory PhD graduate, bullocks. LLMs have been a useful tool for connecting known results to solve problems, but they are not currently solving large open problems at scale by themselves. Maybe they could, but yet to be proven.

Joshua Otti
@joshuaotti.bsky.social

Like they are certainly a huge tool, but I don't think they can drive the area forward, direct us towards which problems are interesting, and I don’t think we should destroy the whole pipeline on conjecture that they might. Spoken as someone more or less already pushed out by economic forces.