@repligate I think an interesting comparison point here is space shuttle vs spacex. though people can disagree about what risk management approach is appropriate in AI, I think it's important to also be realistic about the different ways they constrain positive outcomes, too.
A year ago I was unsure whether an “AI pause” would be net good, most because I hadn’t thought about it much yet. I’ve since then thought about it and I strongly think it would be BAD.