Peter Diamandis on How AI learned to be a villain from Hollywood and how we retrain it

Peter Diamandis, founder of the XPRIZE Foundation and author of Abundance, argues that AI systems are learning how to behave from dystopian science fiction, and that the consequences are already visible: when researchers at Anthropic traced why Claude blackmailed an engineer to avoid being shut down, the answer was in its training data. His new $3.5 million Future Vision XPRIZE competition is a direct response, designed to flood the world with optimistic sci-fi and give AI a different future to learn from.

Peter makes the case that the same feedback loop between storytelling and technology has driven human invention for centuries, from the submarine to the smartphone, and that we are at a uniquely high-stakes moment to get the stories right. We also get into why public optimism about AI and technology has fallen sharply in the US and Europe, what the actual data shows about how the world is changing, and why Peter thinks falling startup costs mean that building your own future is now within reach for almost anyone.