A paper by incoming Anthropic economist Chad Jones models a theoretical scenario where a 33% risk of human extinction is optimal

@RyanPGreenblatt @andreamiotti Why do you think this particular utility is unrealistic?

@MatthewJBar @andreamiotti I think it doesn't accurately model preferences in extreme scenarios, like a 75% chance of annihilation. My understanding is that more generally people tend to have sub logarithmic returns, but less confident in this.

@albrgr @ChadJonesEcon I'm shocked that Chad's tweet has attracted so much attention from normies. It's got to be a bad thing

@RyanPGreenblatt @andreamiotti For the parenthetical are you referring to Bostrom's Optimal Delay paper or are you talking about other papers?

@RyanPGreenblatt @andreamiotti No, the real tradeoff includes offsetting other existential risks with AI