AI's biggest fight is whether it is too dangerous to share or too dangerous to lock up

The sharpest line came after a long discussion of Mark Zuckerberg's 6,500-word essay, "The Future is for Everyone." Zuckerberg, the tech executive behind Facebook, used the essay to promise open source models, meaning the model code and weights are published so anyone can download and run them, plus a free agent for everyone, a tutor for everyone, and a future built on wide access rather than centralized control.

David Sacks, a venture capitalist on the panel, said he found himself nodding along. He said Zuckerberg had pointed at researchers around Anthropic and the effective altruism movement, a philosophy focused on doing the most good for humanity, and asked why they are rushing to create a future they describe as dystopian, with job loss and rogue systems. Sacks then offered his own answer. He said he thinks most of those researchers are missionaries, not mercenaries, who believe that a dangerous future can be managed only under their enlightened control.

Sacks called that belief an old one. He compared it to what writer Thomas Sowell called the vision of the anointed, in which intellectuals believe that giving experts maximum power lets them steer society toward better results. He then quoted Zuckerberg's warning: "Historically, hoping that an absolute power will benevolently provide for humanity if sufficiently enlightened has not led to safe or positive outcomes."

That set up Gavin Baker, an investor joining as a guest, to offer his summary of the divide. "I think Anthropic and people in the effective altruism movement believe that this technology is too dangerous to distribute," Baker said. "And what Mark Zuckerberg and I think Elon and Jensen believe is that this technology is too dangerous to centralize."

Baker said he believes the people warning about danger are sincere. His worry, citing Dwarkesh Patel, a podcaster close to the AI scene, is that even a well meaning lab builds software wired to do what it thinks is best for humanity, with unpredictable consequences. The safer answer, Baker argued, is a rich variety and diversity of AIs. Baker claimed history will judge Elon Musk, who runs xAI, kindly for dedicating his Grok system to the pursuit of objective truth.

Sacks said he had heard the centralized version directly in Washington, where officials wanted two or three companies, no chip or model exports, and something like an atomic energy commission to cartelize the industry with the government. He acknowledged that officials dispute how a similar meeting with investor Marc Andreessen was characterized, and said his reading came from what he heard in his own meetings. Baker closed the point with his own hedge, that there is really no counterexample he can think of to distribution winning out, adding, in his words, "I want an AI that looks out for me."