Mark Zuckerberg Blasts Centralization of A.I. Power

Mark Zuckerberg came out swinging against Anthropic and OpenAI, two of the world’s most valuable artificial intelligence start-ups, saying the way they developed A.I. software could lead to a centralization of power that would narrow people’s access to the transformational technology.

In an interview on Tuesday, the 42-year-old chief executive of Meta did not call out Anthropic and OpenAI by name. But he said leading research labs that want to develop A.I. technology in a tightly controlled manner — as Anthropic and OpenAI have done — would be akin to “abandoning our values” in American tech development and would stifle innovation.

“So much of the discourse from a lot of the other labs that are developing this is overwhelmingly filled with doom,” Mr. Zuckerberg said. “There needs to be a voice or several voices that are bringing realism to this debate.”

The comments from Mr. Zuckerberg, who rarely gives interviews to traditional news organizations, escalate a fraught debate over how artificial intelligence software should be created that has divided Silicon Valley.

Technologists like Dario Amodei, the chief executive of Anthropic, and Sam Altman, the chief executive of OpenAI, have said some of today’s leading A.I. models are too dangerous to be shared and developed without tight controls. But Microsoft, Nvidia, Google and Meta have disagreed, contending that many A.I. models must be developed largely in the open for people to advance the technology and build new businesses.

The discord heightened this month as breakthroughs in Chinese A.I. models appeared to erode the head start that American companies like Anthropic and OpenAI have enjoyed. The companies have lobbied regulators in Washington, raising concerns about Chinese open-source A.I. models.

On Friday, Jensen Huang, the chief executive of Nvidia, posted a letter supportingopen-source software development, which others in Silicon Valley rallied around. This week, Mr. Huang announced a coalition to work on open-source safety and cybersecurity tools for A.I., joined by other tech giants like SpaceX and IBM.

Tech executives including Mr. Huang and Mr. Altman are traveling to Washington this week to further make their case to the Trump administration, which is mulling how to deal with the issue.

Amid the battle, Mr. Zuckerberg has repeatedly argued that more people should have access to A.I. technology rather than less, a tacit criticism of Mr. Amodei and Mr. Altman.

Mr. Amodei, in particular, has called out the dangers of “alignment risk,” an issue with programming artificial intelligence in a way that is safe toward humans.

“I think a lot of people have this notion that if you build some kind of singular A.I. you can, through some idealized form of alignment, make sure it is benevolent to humanity,” Mr. Zuckerberg said in the interview. “I’m personally skeptical of that path.”

He instead called for “personalized superintelligence” for everyone, which would mean configuring A.I. bots to suit people’s needs and desires. A central “benevolent” A.I. superpower, he argued, would be untenable because people have different views and value systems.

“I think it is literally impossible to have a single benevolent superintelligence that is simultaneously aligned with everyone at once,” Mr. Zuckerberg said.

Representatives from Anthropic and OpenAI did not immediately respond to requests for comment. In previous statements, OpenAI has said it supports open-source software and is working with the White House on safe ways to support A.I. development.

In a blog post on Monday, Mr. Amodei said Anthropic was focused on “keeping powerful chips out of authoritarian hands, stopping industrial-scale distillation, and requiring safety testing of all sufficiently capable models, open and closed.”

(The New York Times has sued OpenAI and Microsoft, claiming copyright infringement of news articles. The two companies have denied the claims.)

Meta’s own philosophy around A.I. development has fluctuated. For years, the company championed open-source A.I. and gave away its A.I. models for free. Last spring, after falling behind Anthropic and OpenAI in the A.I. race, Mr. Zuckerberg took action.

He invested $14.3 billion in ScaleAI, an A.I. start-up, and hired its founder, Alexandr Wang, to revamp Meta’s A.I. division, which was renamed Meta Superintelligence Labs. Meta spent billions to assemble a team of all-star A.I. researchers around Mr. Wang and they began developing an A.I. model that was not open, but closed.

The move gave Meta the potential to sell access to new models for more revenue. This month, it began selling access to its latest A.I. model, Muse Spark, for the first time. Meta also continues to develop open-source models.

Mr. Zuckerberg has long believed that A.I. will transform society in positive ways that are yet to be understood. This month, Meta released a national advertising campaign saying that investing in A.I. tools and releasing them broadly meant the company was “betting on people.”

The Wall Street Journal published an opinion piece by Mr. Zuckerberg this week in which he criticized “centralizing” of A.I. power and said he believed in “individual empowerment, invention and balance of power.” In the article, he did not name Anthropic and OpenAI.

In Tuesday’s interview, Mr. Zuckerberg acknowledged the A.I. debate was contentious and noted that talks with different parties in Silicon Valley and Washington had been productive.

“I don’t know if the goal is to have everyone agree on everything,” he said. “Almost by definition, if you believe what I do, you don’t need to convince everyone of everything.”

Looking back over the course of technology, he said, democratization of tech tools has tended to eventually win out.

“I believe the general arc of the industry has been toward more openness and putting the power of tech into more people’s hands, not fewer,” he said.