Elektrine lite

← Feed

@MartinEscardo@mathstodon.xyz

2026-11-14 19:12 UTC

A lot of mathematical developments work like this: "This should work". And then we go frantically in front of a blackboard, or with pencil and paper, to try this out and attempt to make it work. Or, if we have too many this-should-work, we may enlist PhD students, postdocs, collaborators, or, indeed, members of the audience of our talks. Now people are trying their own this-should-work in LLMs, rather than PhD students, postdocs, collaborators and the community. The LLM is removing these important roles, the PhD student, the postdoc, the collaborator, the community. This is the price we are paying to know "what is true". In my own particular case, when I decided to make this my career, I subscribed to the community, not to ultimate truth alone. The LLMs are destroying this sense of community in one way or another. For example, a lot of people are reluctant to share their own this-should-work, where, in the recent past, this was common currency, what made mathematics as a whole work. :-(

Replies (1)

  • @andrejbauer@mathstodon.xyz 2026-11-14 19:39

    @MartinEscardo@mathstodon.xyz Who does this? I keep hearing such claims, but my experience is nothing like it. What I am observing is that I tell PhD students and postdocs "this should work" (just as I always have) and they go and use LLMs, so the research cycle is shorter. And after about half a year of this, they're beginning to do their own "this should work", which is precisely what they should be doing. They're learning to fly. All I see is a speedup. I don't see an important qualitative change in how our group works. It's just crazy that the postdocs present solutions to open problems every week, and build systems that would take years a decade ago.

    Open ##5542361