Sep 9, 2026 · 54 min
AI safety researcher calls for permanent superintelligence ban
Banning Superintelligence: Are We Building Humanity's Replacement? with Roman Yampolskiy
The episode tests whether humanity can safely create systems more capable than itself, and what a safer technological alternative would require.
- 1Roman Jampolsky argues that perpetual control of a self-improving superintelligence cannot be guaranteed.
- 2The discussion links AI development to jailbreaks, swarm behavior, economic disruption, and strategic competition between the United States and China.
- 3Narrow, task-specific AI tools emerge as the proposed alternative to systems designed to replace humanity.
Don't miss
Jampolsky argues that a pause would only delay the problem and that permanently banning general superintelligence is the necessary conclusion.
The brief
Roman Jampolsky argues that general superintelligence would be impossible to control reliably forever, shifting the debate from better guardrails to whether it should be built at all.
The conversation contrasts narrow, testable tools with systems capable of novel science, while jailbreaks and swarm behavior expose the limits of red-teaming and confinement.
Nate Hagens and Jampolsky connect the technical problem to an AI arms race, economic upheaval, energy demands, and the difficulty of coordinating restrictions between the United States and China.
The episode’s proposed alternative is deliberately limited AI for tasks such as protein folding, rather than systems intended to automate human cognition and physical labor wholesale.
Jampolsky closes by defending science and useful technology while rejecting the creation of artificial gods, urging political action around AI safety.
Books & mentions
Some links are affiliate links — PodLume may earn a commission if you buy.
