Post

This is a very silly framing of the relevant issue. It's like saying "if a text-to-speech converter is analyzing speech, interpreting it, and converting it to written language, but is not 'listening,' the word 'listening' has no meaning." No. "A meaning different than mine" is not "no meaning."

Whether an LLM is "understanding" is the very definition of a definitional issue. And what is common sense to one person (Will Stancil, say, or not-Will-Stancil), may not be to another (not-Will-Stancil, for instance, or in the alternative, Will Stancil.)

AIs are doing things that resemble human cognition in some ways, and don't resemble it in others. They are—and this is a thing Stancil gets right—far more capable at doing some things that only humans could do until now than some knee-jerk AI critics admit. But.

As is so often the case, Stancil has concluded that if someone else is saying something that's incorrect, it follows that (1) that person is egregiously wrong, (2) that person is operating in bad faith, out of animus, or out of peer pressure, and (3) Stancil's opposing claim is obviously correct.