Lots of loose talk about how likely AI is to kill us all. One better way to think about the risks of AI is to learn from near misses, incidents like the Hugging Face attack and many others, to estimate actual risk, not just existential risk.
johnkroman.substack.com/p/using-near...
johnkroman.substack.com
Using near misses to estimate p(doom) in AI
Lessons from crime policy research