[qwertylicious]

Yeah, LLMs share a lot of the same challenges as self-driving cars: when they work great, we get complacent, and then when they fail, they fail in ways that humans are really bad at anticipating.

And the discourse around all this is also the same: the detractors point out these flaws, and the proponents chime in with "Yeah well humans suck too" and around and around we go...