I feel exactly the same way about the kind of LLM-based AI we have today.

But the AI x-risk argument is about what happens when we have "real" AI that can actually think like humans but a million times faster.

There are no real experts on this and yet it does seem worth worrying about in advance.