In you're in favor of "AI safety" (broad definition), what's your most compelling cast-length argument?

if toddlers were able to make an adult human from scratch would they be able to control it?

What evidence do you have that we are even close to achieving that level of cognitive and effective disparity?

I don’t think we’re close to that personally but it’s just about mitigating tail risk right? We might be someday, and it could be sooner than we think

Humans have never dealt with an entity that’s more intelligent and capable than us so feels like it makes sense to tread carefully

I do agree with the OAI strategy of using contact with reality to approach alignment, not pausing/regulating