Hello. Curt here.
So, almost every interview with Roman Yampolskiy opens with the apocalypse. I wanted to do something different today…
Roman popularized the term “AI safety,” and his career has shifted to proving the specific ways a superintelligence cannot be controlled. His latest book is titled, with no irony, AI: Unexplainable, Unpredictable, Uncontrollable.
The simulation he thinks we are probably living inside. Whether the models he studies are already, in some thin sense, conscious. What free will could possibly mean if a little quantum randomness is the only thing saving it from determinism. Somewhere in there I told him I agree with his spirit but disagree with his text, and we spent the rest of the conversation finding out which of us was the fool. NOTE: Probably me.
We cover:
The strongest thing he believes, stated flatly: “you cannot indefinitely control something smarter than you.”
Whether the models he studies are already conscious. He would not let himself off the hook: “ei…
Listen to this episode with a 7-day free trial
Subscribe to Curt Jaimungal to listen to this post and get 7 days of free access to the full post archives.
