So #LLM can't be trusted.

[mattesilver@fe.disroot.org]

So #LLM can't be trusted. OK, same with people! We build trust over time, but there are plenty of contexts where it doesn't apply. That's why we have procedures and oversight.

Perhaps we'll build similar procedures for AI over time? Maybe even semi-automated?

Not a big fan of dumping everything on #AI, just an argument in the discussion.