[api]

A lot of these questions hinge on words like “think” and “understand” for which we do not have precise definitions. It’s really kind of calling bullshit on a lot of the terminology we use to talk about thinking, revealing that most of it is hand wavey anthropocentrism.

It does what it does. We do not know what that means. We do not really know how close what it is doing is to what we do.

It’s also possible that it’s doing similar things to what we do but in some very different way. “Neural network” has always been a term to make biologists cringe as they’re really only the loosest of analogies to how brains work.

[xp84]

So true! We basically define those words almost self-referentially. Understanding is what human brains do, so by definition nothing else but a human can 'understand' complex things (we say).

We make only slight allowances for animals, allowing that they can "understand" but only simple things.

As this tech gets better, those who cling to that definition are going to get more and more uncomfortable, and tie themselves into more and more logical knots trying desperately to explain why it "doesn't count."

[lxgr]

A similar thing has happened (and arguably is still happening) when it comes to animal cognition.

"The only species capable of recognizing itself in the mirror. Ok, recognizing itself in the mirror and object permanence. Ok, all of that and mourning its dead. Ok, hang on for a minute..."