In a notable exchange during a public event in Berlin, OpenAI CEO Sam Altman and British physicist David Deutsch agreed on a new benchmark for determining when artificial intelligence achieves human-level general intelligence, or AGI. The proposed test: if an AI system can solve the problem of quantum gravity and provide a coherent, insightful explanation of how and why it arrived at that solution, it may have crossed the threshold into true intelligence. The conversation unfolded during a fireside chat hosted by Axel Springer publisher Mathias Döpfner at the company’s Berlin headquarters. Döpfner, intrigued by Altman’s admiration for Deutsch’s book The Beginning of Infinity, invited the Oxford-based physicist to join the discussion via video link—an unexpected but enthusiastic moment for Altman, who visibly delighted in the connection. Deutsch, widely regarded as a founding figure in quantum computing and a leading thinker on the philosophy of science, has long questioned whether AI systems built through scale and data training could ever achieve genuine understanding. He acknowledged that models like ChatGPT have demonstrated remarkable conversational ability, even if they lack true comprehension. “ChatGPT proved me wrong,” he admitted. “It’s not an AGI, but it can converse.” Still, Deutsch insisted that real intelligence goes beyond pattern recognition and mimicry. He defined it as the capacity to create new knowledge—identifying problems, inventing solutions, testing them, and refining them through iteration. He cited Einstein’s development of relativity as a prime example: not just assembling existing ideas, but creating something fundamentally new, driven by deep curiosity and insight. Altman, whose approach to AGI relies on scaling up models and iterative development, asked Deutsch whether a future AI that could not only solve quantum gravity but also articulate the reasoning behind its approach—its motivations, challenges, and thought process—would meet the standard of true intelligence. Deutsch responded without hesitation: “I think it would, yes.” Altman immediately agreed: “I agree to that as the test.” The moment marked a rare alignment between two influential figures with different perspectives on AI. Altman, the builder focused on rapid progress, found common ground with Deutsch, the philosopher skeptical of brute-force methods. Their shared benchmark shifts the goalpost from performance to understanding—requiring AI not just to produce answers, but to demonstrate the kind of creative, self-directed inquiry that defines human intelligence.