The way you describe it, it seems hard to argue that AI won't be a boon to research in math and science at large. Does it matters though that this is not the path of least resistance to using the technology? (The path of least resistance in terms of how frontier AI is designed, where it's increasingly difficult and unpleasant to talk to, I would argue, is to sacrifice understanding and basically turn into a reverse centaur, a la Corey Doctorow.)
Hmm, interesting thought though that the incentives of the companies building the tech are not necessarily the same as those trying to use it. I haven't noticed any obvious misalignment so far (I don't actually find it unpleasant to talk to), but I suppose it could happen in principle. But you'd think the financial incentive alone might help.
Yeah, on second thought, I guess it really is just one model (Opus 5.0) which I feel is especially difficult to talk to, but it seems to be better already in Opus 5.5.
The way you describe it, it seems hard to argue that AI won't be a boon to research in math and science at large. Does it matters though that this is not the path of least resistance to using the technology? (The path of least resistance in terms of how frontier AI is designed, where it's increasingly difficult and unpleasant to talk to, I would argue, is to sacrifice understanding and basically turn into a reverse centaur, a la Corey Doctorow.)
I guess you address this in your last two paragraphs.
> Of course, all of this assumes that the researcher is trying to write better papers and increase their own understanding.
And I guess the quality of academic output has always depended on this condition.
Hmm, interesting thought though that the incentives of the companies building the tech are not necessarily the same as those trying to use it. I haven't noticed any obvious misalignment so far (I don't actually find it unpleasant to talk to), but I suppose it could happen in principle. But you'd think the financial incentive alone might help.
Yeah, on second thought, I guess it really is just one model (Opus 5.0) which I feel is especially difficult to talk to, but it seems to be better already in Opus 5.5.