Setting aside the marvelous murk in that use of "you," which parenthetically I would be happy to chat about ad nauseum,

I would say this is a fine time to haul out:

Ximm's Law: every critique of AI assumes to some degree that contemporary implementations will not, or cannot, be improved upon.

Lemma: any statement about AI which uses the word "never" to preclude some feature from future realization is false.

Lemma: contemporary implementations have already improved; they're just unevenly distributed.

I can never these days stop thinking about the XKCD the punchline of which is the alarmingly brief window between "can do at all" and "can do with superhuman capacity."

I'm fully aware of the numerous dimensions upon which the advancement from one state to the other, in any specific domain, is unpredictable, Hard, or less likely to be quick... but this the rare case where absent black swan externalities ending the game, line goes up.

"every critique of AI assumes to some degree that contemporary implementations will not, or cannot, be improved upon."

They're token predictors. This is inherently a limited technology, which is optimized for making people feel good about interacting with it.

There may be future AI technologies which are not just token predictors, and will have different capabilities. Or maybe there won't be. But when we talk about AI these days, we're talking about a technology with a skill ceiling.