One day this March in the Summer Palace of Beijing, a tour guide was struggling to hold his group’s attention. Instead of taking in the magnificent pavilions of the imperial gardens, the tourists – a gaggle of Western and Chinese technology experts – were more interested in talking to each other. “I felt sorry for the guy,” one of them recalled. “People are walking around this UNESCO World Heritage site, and it’s, like, beautiful, but they’re arguing about AI.”

The squabbles had an edge. For over a year, ChatGPT, an artificial-intelligence (AI) tool, had showcased the extraordinary leaps the technology was making. As big firms raced to develop their own faster, smarter products, debates intensified about how these models might evolve. Many experts fear that without robust guardrails, AI could be used to develop new diseases or cyber-weapons. Some believe it could advance to the point where humans can no longer control it, with potentially apocalyptic consequences. The fact that the two superpowers driving the development of AI – America and China – are locked in an escalating cycle of confrontation only adds to the sense of alarm.

The trip to the Summer Palace had been designed to foster camaraderie among an unlikely AI working group. It had been convened by Stuart Russell, a British computer scientist at the University of California, Berkeley, and one of the sector’s most prominent “doomers”.