Will AI Destroy Humanity? 11 Experts Offer 11 Answers

Summary

The San Francisco Standard asks 11 people from AI safety research, venture capital, politics, industry, academia and protest movements how seriously to take fears that AI could threaten humanity. Their answers diverge over both the danger and the remedy. Malo Bourgon of the Machine Intelligence Research Institute says recent AI disasters and the prospect of recursively self-improving autonomous agents make existential risk credible, while Will Wilson of Antithesis says even aligned superintelligence could make people economically and militarily irrelevant and enable lasting authoritarian control. Others reject or narrow the doom narrative: UC Berkeley’s Shreeharsh Kelkar considers human extinction unlikely but treats accidents in complex systems as a continuing risk, and Dylan Baker of the Distributed AI Research Institute argues that speculative “kill switch” debates distract from current environmental, labor, child-safety, mental-health and copyright harms. Several contributors call for regulation. State Sen. Scott Wiener supports government standards, transparency, incident reporting and possible kill switches, citing California’s enacted reporting law after an earlier catastrophic-risk bill was vetoed. Josh Constine favors rapid development to maintain strategic strength, but wants cautious deployment, stronger pre-release monitoring and possible restrictions on models or training environments designed for catastrophic misuse. Aniket Shah says regulation is unlikely to slow development without a disaster on the scale of a nuclear accident, while Christopher Ackerman views recent public incidents as an inflection point for AI-safety politics. Other perspectives focus on human agency, employment anxiety, profit incentives and the risk that AI safety becomes partisan. Taken together, the interviews show no consensus on whether extinction is plausible, but broad agreement that oversight, deployment choices and social consequences require attention.