Wow, I never thought I would (indirectly) be mentioned on the Motte (I'm one of the authors of the "Asymptotically attaining the Moore bound" paper you linked)! Indeed GPT was quite helpful both for the initial ideation and for reviewing/sanity checking the construction, and for some of my other recent papers GPT was responsible for the bulk of (or even 100% of) the intellectual work. I would say that for that paper in particular the GPT use was a little less impactful, but it definitely helped speed up the research/writing process.
Interestingly enough, 3-4 years ago I was adamant that AI would never be able to compete with human mathematicians (read: me), and even bet that AI would never be able to solve the average USAMO #2/#5, much less be capable of resolving open research questions. Clearly I was wrong, and I am now fully "AI-pilled." I think it's a very exciting time to be doing math! Maybe that will change if (when?) GPT 7 or GPT 8 no longer needs any human input or guidance at all.
While research math is very different from other white-collar labor, my experiences doing math with GPT actually led me to recently make impactful career decisions - I'm not sure how much longer I'll be economically competitive in my chosen field, so recently made some decisions aimed at maximizing income in the next 3-5 years at the cost of career stability.
p
Wow, I never thought I would (indirectly) be mentioned on the Motte (I'm one of the authors of the "Asymptotically attaining the Moore bound" paper you linked)! Indeed GPT was quite helpful both for the initial ideation and for reviewing/sanity checking the construction, and for some of my other recent papers GPT was responsible for the bulk of (or even 100% of) the intellectual work. I would say that for that paper in particular the GPT use was a little less impactful, but it definitely helped speed up the research/writing process.
Interestingly enough, 3-4 years ago I was adamant that AI would never be able to compete with human mathematicians (read: me), and even bet that AI would never be able to solve the average USAMO #2/#5, much less be capable of resolving open research questions. Clearly I was wrong, and I am now fully "AI-pilled." I think it's a very exciting time to be doing math! Maybe that will change if (when?) GPT 7 or GPT 8 no longer needs any human input or guidance at all.
While research math is very different from other white-collar labor, my experiences doing math with GPT actually led me to recently make impactful career decisions - I'm not sure how much longer I'll be economically competitive in my chosen field, so recently made some decisions aimed at maximizing income in the next 3-5 years at the cost of career stability.
More options
Context Copy link