Login
You're viewing the functional.cafe public feed.
  • Nov 15, 2026, 5:18 AM

    I'm seeing a lot of single-author papers like "I used [LLM] to help me understand [LLM-produced result], and then wrote up my own understanding."

    Rather than multi-author papers like "We got together in a reading group to try to understand [LLM-produced result] and here's what we came up with."

    Yet another example of how, even when people pursue human understanding of LLM-produced math results, *using chatbots to do it* weakens community ties instead of strengthening them. A real missed opportunity.

    #math #AI #academia

    💬 3🔄 19⭐ 20

Replies

  • Nov 15, 2026, 6:04 AM

    @joshuagrochow but first asking a machine, just like first opening a book or asking a colleague, immediately deletes the possibility that the person might have reached a genuinely personal understanding... How don't people realize this?

    💬 2🔄 1⭐ 3
  • Nov 16, 2026, 6:25 AM

    @antoinechambertloir I often wonder this.

    And there is some real disconnect b/c people even tell their undergraduate students they learn better by doing (or at least first seriously attempting) the problems themselves. Why do they think the same doesn't apply to their own research and proofs?

    💬 0🔄 0⭐ 0
  • Nov 16, 2026, 6:48 AM

    @antoinechambertloir Oh, that's an easy one: the reason people dont realize that is because it's completely false! You can totally gain someone else's understanding from a book and afterwards still gain your own personal understanding by continuing to think about it. Maybe getting the answer first makes it less likely that you will still make the effort to gain your own personal understanding, but it by no means "immediately deletes the possibility". In fact, I'd be shocked if you told me you never found your own personal insight into something you already understood from else point of view! @joshuagrochow

    💬 1🔄 0⭐ 6
  • 💬 0🔄 0⭐ 0
  • Nov 15, 2026, 9:39 AM

    @joshuagrochow I teach a pre-doctotal intro statistics course. Some students come from very quant backgrounds, some not, and I'm supposed to homogenize them. Few years ago, it would naturally happen that each non-quant student would pair with a stronger one to work on problems and R, and that part of the class had a very lively environment. Now they all work in silence, aided by their LLM. (And it's not that they dislike each other! They hang out together in the weekends and stuff.)

    💬 0🔄 0⭐ 0
  • Nov 15, 2026, 3:08 PM

    @joshuagrochow That’s probably partly a selection effect - it will take a longer time for groups to meet (with busy schedules), understand the results communally, and then write things up.

    💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0
  • Nov 16, 2026, 3:50 PM

    @matthras @Ianagol maybe some. But I’ve also seen plenty from very well-known / relatively senior people. Isolated or not, it would not be hard for such people to organize an online reading group…

    💬 0🔄 0⭐ 0