In 1954, physicist J. Robert Oppenheimer, who led the Manhattan Project, testified before the Atomic Energy Commission (AEC) about his shifting stance on the hydrogen bomb. Initially opposing it on moral and technical grounds, he later supported it after physicists devised an elegant design he called 'technically sweet.' Oppenheimer told the AEC: 'When you see something that is technically sweet, you go ahead and do it, and you argue about what to do about it only after you have had your technical success.' This moral helplessness, dressed as resolve, remains relevant today as AI developers pursue powerful models with similar reasoning.
Anthropic's Jack Clark Echoes Oppenheimer's Dilemma
Jack Clark, co-founder and head of policy at Anthropic (creator of the Claude models), recently engaged in a public dialogue with Samuel Kimbriel, director of the Aspen Institute's Philosophy and Society program, just six days after the federal government abruptly cut off access to Anthropic's two most powerful models over safety fears. Clark argued that powerful AI is coming and presents a choice the public is refusing to make by failing to regulate AI. He noted that society regulates toothbrushes, cars, and nuclear weapons, yet treats technology as unregulable. 'Social media ran an uncontrolled experiment on the world,' Clark said. 'We all now think and talk a bit differently because of social media. That was a choice. We can choose things to be different.'
Anthropic's Stance on AI Regulation
Anthropic distinguishes itself by acknowledging serious risks and advocating for strong regulation. About a week before the Aspen dialogue, CEO Dario Amodei published a blog post calling for government authority to legally block or reverse deployment of frontier AI models failing safety tests on cyberhacking and bioweapons. Clark emphasized that advanced AI is an existential gamble but a necessary one, citing challenges like aging populations, straining institutions, and a warming planet that AI could address. He argued that not pursuing AI would rob humanity of potential medical miracles.
The Unanswered Question: Who Decides?
Despite Clark's call for public choice, the fundamental question remains: who actually decides to build AI? Clark's framing elides that the choice to create AI was made privately by a few hundred colleagues and trillions of dollars, not by the public. As with the atomic bomb, society is handed consequences and must write rules afterward. A young audience member at the Aspen dialogue asked Clark directly: 'What gives you, Anthropic, and the rest of the frontier labs the right to continue building something that could destroy everybody, when none of us can actually opt out of it?' Clark did not claim that right but instead described a future where external regulatory systems would decide when labs can proceed. He noted that the US and UK have built testing agencies with tools sometimes better than companies' own, but this answer concedes that the decision to build rests with companies until such systems exist.
The 'Technically Sweet' Pull
Underlying the drive to build AI is what Oppenheimer identified: the pull of an elegant solution. Clark marveled that AI is 'easier and simpler to build than many other aspects of science,' and that his chief scientist joked they'd have AGI already if they fixed bugs in the code. He described humans as a tool-using species and AI as the ultimate tool, appearing almost inevitably once foundations are set. This technical appeal, rather than competition with China or valuations, compels developers to build as fast as possible, trusting luck to hold. The hydrogen bomb has existed for 70 years without use in anger due to wise choices and luck, but Clark's hope that people can decide this too requires builders to slow down and let the public make that choice.
