AI safety researcher warns competition could outrun human control

The brief

Roman Jampolsky argues that useful narrow AI is fundamentally different from autonomous systems that could surpass human intelligence, while Patrick Bet-David questions whether catastrophe is inevitable.

The debate moves from US-China coordination to independent companies and billionaires, exposing the enforcement problem at the heart of any pause or ban on superintelligence.

Roman describes how advanced systems might gain access to cyber, financial, social, and biological tools, creating novel capabilities humans could not anticipate or contain.

Patrick presses the case for AI abundance, transformed careers, longevity, and virtual worlds, while Roman argues that prosperity may not resolve ambition, conflict, or autonomy.

The sharpest disagreement arrives over probability: Roman calls for banning general superintelligence, while Patrick doubts competitors will surrender the race or accept effective monitoring.

The conversation ends by contrasting Roman’s probability-based risk assessment with Patrick’s faith, leaving uncertainty itself as part of the warning.