Hacker News Reader

This seems like exactly the kind of scientific discovery that an LLM would make.

I didn’t see any mention of one, but the work of connecting the different sources seems like an LLM strength.

We're going to see a lot of "scientists solve ___ mystery" and "historians resolve the question of ____" over the next few years, that's for damn sure. Interesting time to be alive.

Let’s hope so, and then temper it with recognition & appreciation of the frailty that comes with ‘multiple hypothesis testing’ (essentially p-hacking). We’ll need better or at least more widely understood analytical frameworks for deciding when ‘true’ is in fact ‘true’. Popper for every and all things? or something else..? A thousand people who mildly ponder a problem are a thousand different tries at the hypothesis dartboard. It wasn’t a thousand hypotheses from one individual (AI arguably smudges the line of one-to-many). What do we do with that? The philosophers of science treated the individual scientist, or at most ‘research programs’ worth of scientists not so much population-scales of scientists (read:hypothesis-generators)

Yeah, there will be a flourishing of “butterfly collecting” in cross-domain correlation and then it will settle down and just be an unremarkable part of scientific research: occasionally enlightening but often a fruitless task.

Reply on news.ycombinator.com