I think some people insist on "consciousness" as being a necessary component of AI, and that "understanding" is a function of consciousness. And consciousness means "being conscious the way biological systems like ourselves are conscious". AND, the final nail in this coffin: "that's impossible". Hard to argue with.

QED, ergo, in conclusion regarding AIs ever "understanding" anything: Nope.

But what about....? Nope.

But maybe they'll...? I said no.

What if they invent a...? Doesn't matter, what part of "impossible" are you not getting here?

Just to be clear, I am not one of these people. But I think this is what we sometimes see. In order for AI to be "real", it has to have characteristics that are basically impossible to test for (i.e., consciousness and\or self-awareness). Thus, for these people AI can't ever be real.