Agree, and most of the argument downstream of this collapses two separate questions:
1. How dangerous is this and how fast is it moving?
2. Given that, do you accelerate, race, slow down, or regulate?
Your answer to the first doesn't determine the second. You can hold a high p(doom) and still conclude racing is the least bad option available. Which is why people with near-identical risk estimates end up in camps that can't stand each other.
I can't predict the AI future, and I'm very unsure about what the right strategy or policies should be moving forward. What I do know is that people who scoff at the idea that we may be in grave danger have not read enough about AI.