I liked this post - it seems generally good to talk about the context behind people's writings and changes in perspective for a more rational AI risk discussion. As an aside, I find that of the many people who derogatarize "doomers" or dismiss AI risk as fiction don't even try to engage with their arguments. I previously had the impression you were one of these people, which I've now updated against (although to be fair, I hadn't read your writing comprehensively, so this was a vague impression).
Swarm behavior, scary RL stuff in general with recent models? Maybe the thing around the misleading letter to congress?
although in a more reasonable world, i would imagine that this war would not be necessary and that their immediate reaction right now should just be to do everything to coordinate with oai for some kind of slowdown
In what way is J-lens blantantly dual-use? And I mean this in that I thought about it for a bit, and am curious how you interpret this. The way this is phrased seems to imply more than just problems with safetywashing
Overall, im pretty symphathetic to the general argument btw