Login
You're viewing the functional.cafe public feed.

Replies

  • 💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 1
  • 💬 0🔄 0⭐ 2
  • 💬 1🔄 0⭐ 2
  • 💬 0🔄 0⭐ 2
  • Oct 31, 2026, 7:58 PM

    @chrisamaphone the last bit about speaking about it to show people that they are not alone really resonates with me. Both from needing that affirmation and for also trying to be vocally opposed for those who don't dare

    💬 0🔄 0⭐ 1
  • Oct 31, 2026, 8:30 PM

    @chrisamaphone I was talking with one of my close friends who happens to work in cybersecurity a few weeks ago, talking about generative AI and code reviews.

    And I pointed out that code reviews are of infamously limited effectiveness, and mostly effective at catching major misunderstandings and shortcomings in cooperative, non-adversarial settings.

    I tried to make an analogy of a proof checker that's being used by a serious mathematician dedicated to their craft, versus a proof checker that's being used to validate proof-carrying code in maliciously adversarial scenarios.

    And of course, generative AI might not quite be a full-on maliciously adversarial actor, but it's certainly not a cooperative actor. In many ways an appropriate model is much closer to adversarial than cooperative. Thus the idea of having humans review generative AI code seems deeply suspect.

    He doesn't really have any background in theorem proving or formalizing mathematics, so I don't know how well my analogy stuck, but he seemed to get most of what I was trying to say.

    💬 0🔄 0⭐ 3
  • 💬 1🔄 0⭐ 0
  • 💬 0🔄 0⭐ 0