The other thing that annoys me is that there are plenty of folks in the AI safety space who do not think very highly (at least now) of Yudkowsky. One of the comments I saw recently on a Substack article talked about how Yudkowsky has "beclowned" himself by not updating some of his earlier theories given how LLMs function.

So it's pretty annoying that this article is trying to paint everyone who cares about AI safety is in a "sex cult". There are lots of folks who are genuinely concerned about risks. If anything I discounted a lot of the AI safety folks because I thought they were not seeing the underlying limitations in LLMs. That may still be the case, but agents keep breaking through what I thought were limits on their behavior.

If anything, I don't understand at all the folks who think ASI is on the horizon (< 10 years) but think everything will be hunky dory and easy to control once we actually do build superintelligence.