I'm back, I missed you all.
Since this is an important subject I know too much about, here's a 🧵 explaining that
AI Safety Is Mostly A Sex Cult.
I don't think these people should make policy. (1/?)
(alternative title: Time For Some Cult Theory)
workflow for big projects seems to be approaching something like this:
1. write very detailed behavioural spec
2. make fable yolo it
3. see what happens when the spec meets reality
4. reimplement from scratch using Sol on a very tight leash