AI Copium reviews interview excerpts in which Geoffrey Hinton discusses his expectations for superintelligence and his concern that safety methodsAI safety is the field and practice of reducing harmful failures, misuse, loss of control, and unintended consequences from AI systems. may not keep pace. His timelines are predictions, not established outcomes.
Geoffrey Hinton describes how model capabilities and shared digital training changed his assessment of AI. He argues that an agent could develop continued operation as an instrumental subgoal and discusses incentives for regulationGovernance sets responsibilities, rules and oversight for decisions about how AI systems are used..
Hinton also expresses his belief that current AI systems may be conscious; the narrator questions that view. Whether AI can have subjective experience remains uncertain and contested.
Watch on YouTube
