← Home
The Diary of a CEO with Steven Bartlett · September 3, 2025 · 2h

Roman Yampolskiy: Only 5 Jobs Will Remain in 2030

Roman Yampolskiy — AI safety researcher. His thesis: we cannot control a superintelligent AI, and we're building one anyway. The simulation argument. Only 5 types of jobs will survive AI automation. The most pessimistic credible voice on AI risk.

Highlights

We cannot control a superintelligent AI — the alignment problem is unsolvable
Yampolskiy's thesis: alignment (making AI do what we want) is not a hard engineering problem — it's an impossible one. You cannot verify the goals of a system smarter than you.

Misc

University of Louisville computer scientist
Most pessimistic credible AI researcher