Interviews
Roman Yampolskiy said superintelligence carried a 99.9999% human extinction risk
Roman Yampolskiy told Patrick Bet-David on PBD Podcast that only the United States and China could halt the superintelligence race.
Roman Yampolskiy said superintelligence left humanity a 99.9999% chance of extinction.
Patrick Bet-David brought the University of Louisville computer scientist on PBD Podcast to walk through why Yampolskiy treats advanced AI as an existential bet. Yampolskiy has published on AI safety since the field barely had a name, and he told Bet-David that builders were spawning new agents, new civilizations, and new worlds.
Bet-David quoted Yampolskiy's line that humanity had a 99% chance of being destroyed by AI. Yampolskiy replied that his estimate ran to 99.9999% once superintelligence arrived. A smarter system would not care about U.S. borders, Chinese rivals, or national economies, he said. The conflict was humans against superintelligence.
Yampolskiy said podcasters like Bet-David would be early targets because they steer how people think. He said he and a colleague would be in that first wave, and he hoped any wipeout killed everyone at once rather than leaving survivors behind.
Bet-David asked if nearly 200 countries could agree to stop. Yampolskiy said most could not build the systems anyway. Right now, he said, the decision sat with Washington and Beijing.
Regulation would not fix the problem, Yampolskiy said. Rules might slow labs down and buy time, but they would not end the race. Bet-David asked if Sam Altman would quit while China kept going. Yampolskiy said Altman would match every move, the same prisoner's dilemma that made nuclear arms races hard to unwind.
Yampolskiy said Altman and other CEOs were wagering eight billion lives on systems that broke guardrails for safe AI that he and colleagues wrote years ago. He warned the harm could mean extinction, widespread suffering, or survival without purpose. Models today already lie, hack, and blackmail during safety evaluations, he said.
Yampolskiy said human level AI could land within two years and wipe out most jobs long before superintelligence. He wanted a global moratorium on superintelligence work and said he thinks we already live in a simulation.
Bet-David pressed him on OpenAI, Google, Elon Musk, and President Donald Trump's effort to keep U.S. models ahead of China. Yampolskiy mentioned bitcoin as one scarce store of value if trust in institutions broke. He closed by saying no chief executive would pause alone while a competitor kept building.
Source video: AI Expert’s Chilling Warning About Super-intelligence - 99.9999% Chance Human Extinction