Sam Altman, CEO of OpenAI, poses a question to physicist David Deutsch about what it would actually take to believe an AI is thinking: The setup is a discussion of Einstein and general relativity which Altman calls one of the most beautiful things humanity has ever figured out, maybe even number one. But his point isn't about the physics. It's about the story. As Altman puts it: "Einstein had a story. We knew what he was working on." We knew the problems Einstein wrestled with, the questions he chose to chase, and the path he took to get there. That narrative is part of how we recognise genuine understanding. So @sama builds a hypothetical to test the line between imitation and real reasoning: "If in a few years GPT-8 figured out quantum gravity and could tell you its story of how it did it and the problems it was thinking about and why it decided to work on that, but it still just looked like a language model output but it really did solve it… would that be enough to convince you?" In other words: not just the right answer, but the reasoning, the choices, the why this problem. The same things we'd want from any human physicist. Deutsch's response is short: "I think it would. Yeah." And Altman accepts it as the bar: "I agree to that as the test." The real test for AI might not be whether it can pass as human, but whether it can produce something genuinely new: solving a problem that's eluded us for a century and account for how and why it got there. Output alone isn't enough. The story is what makes it convincing.