The ‘But China!’ Dilemma Driving the A.I. Race
About this episode
Sheehan, a Carnegie China-AI specialist, unpacks the 'But China!' dilemma paralyzing US AI regulation: every frontier-safety proposal crashes into the claim that slowing down surrenders the race to Beijing. He argues the race metaphor misleads — China doesn't share Silicon Valley's FOOM obsession, is severely compute-constrained, and focuses on diffusing AI applications rather than consolidating for takeoff; ironically, China regulated AI more heavily than the US while catching up via distillation of American models. Beijing views Washington as both trying to strangle it (chip export controls) and recklessly deregulatory, breeding deep mistrust. Ahead of Trump-Xi and Bessent-led talks, Sheehan proposes modest wins: a recurring dialogue, technical safety exchanges between US CAISI and China's Working Group 9, and crisis communication channels (even literal faxes). He closes genuinely alarmed: the evidence increasingly favors the longtime warners, and timelines are shortening.