I'm also not actively doing much research these days (though hoping I can continue in some fashion..) but what I'm hearing from my colleagues is that LLMs can sometimes find bugs in code that might normally take days of debugging to find. I think this is one of the clearest upsides to the tech.

Show Post

Yeah, I'm not a "booster", but for my former job it's hard to see how could not be extremely helpful. E.g doing pointless computer work, quickly making prototypes to check if research avenues have potential, and most important of all: finding errors.

Show Post

Yes — and science was already struggling with the quantity. Now the problem has been magnified at least tenfold, if not much more. Don't get me wrong, LLMs can be incredibly powerful in the right hands. The long term effect on science as a whole is not yet obvious to me though

Show Post

That's certainly a problem, though there was a ton of (effectively) "slop", fabricated science, papers with trivial mistakes, etc., well before LLMs.

Show Post

In principle, having LLMs help out with the time-consuming drudgery is great. But in science, part of the drudgery is *really important*, and you might not know beforehand which is which. Basically, LLMs may make some things a lot faster. But faster isn't necessarily better in science

Show Post