Mark Zuckerberg Says U.S. Should Accelerate AI Development, Not Restrict It
Meta CEO says discourse regarding artificial intelligence isn’t optimistic enough
- Meta CEO Mark Zuckerberg said AI discourse should be more optimistic and warned against restricting access to the technology.
- Zuckerberg said a voluntary U.S. review of AI models should be efficient and doubted that banning foreign open-source models would be productive.
- Meta has projected up to $145 billion in capital spending this year, largely to buy AI chips and build out data centers.
Mark Zuckerberg thinks artificial intelligence is working out great for humanity. His fear: that the people who think otherwise will influence who has access to the fast-evolving technology.
Optimism about AI’s impact on the world “should empirically be the default assumption of how this is going to play out,” the Meta Platforms META -0.08%decrease; down pointing triangle chief executive officer said in an interview. He cited historical analogues, recent jobs data and his own personal use of AI for parenting and fitness as informing his positive outlook on the technology.
“I get that it’s always hard to debate about the future because it hasn’t happened yet,” Zuckerberg said. “But I do think we have a lot of data points at this point, and that should point us to be much more optimistic than I believe the current discourse reflects.”
In an opinion column in The Wall Street Journal on Tuesday, Zuckerberg said “superintelligence”—his company’s preferred term for smarter-than-human AI—“will be the most profound technological advance we will see in our lifetimes,” as long as its power isn’t “restricted to a few institutions.”
He called it “surprising that the discourse from many of those who are developing artificial intelligence is so filled with doom,” an apparent reference to warnings about mass job displacement by leaders at companies including Anthropic and OpenAI, makers of the top-rated AI models.
Zuckerberg, whose company is racing to catch up to rivals including Anthropic and OpenAI, lends a high-profile accelerationist voice to America’s debate over how to develop and regulate increasingly powerful AI models. Meta shares have fallen about 17% during the past year.
“There are new issues that we need to think through,” he said in the interview. “But at the same time, as we think through those issues, distributing the technology seems like it is creating more value than harm by quite a margin on all these fronts.”
The Trump administration is in the process of implementing an executive order establishing a voluntary 30-day review period for new models. Zuckerberg said he thinks any such review should be done as efficiently as possible.
“I think saying, ‘Hey, we want to review stuff for’—whatever it is—30 or 60 days, it may not sound like that much,” he said. “But the field is moving so quickly, that actually is quite a meaningful amount of time.”
With Meta’s AI model development lagging behind that of Anthropic and OpenAI, any across-the-board regulation on the speed of development could potentially lock in those companies’ leads.
Some Trump administration officials have also discussed restricting U.S. companies from using open-weight models developed in China, citing allegations that Chinese companies have unfairly copied American models’ technology. Critics of such restrictions have said access to lower-cost open models levels the playing field for companies and AI developers.
In the interview, Zuckerberg, whose company recently signed a letter urging the government to avoid banning open-weight AI models, said he doubted “banning open-source models from different countries is going to be productive.”
American policymakers should concentrate their efforts on accelerating the domestic AI industry and eliminating bottlenecks that would slow down AI-bred innovations such as new drugs, he said.
Meta has projected up to $145 billion in capital spending this year, largely to buy AI chips and build out data centers. It laid off 8,000 employees in May, attributing the cuts to the cost of its AI initiatives.
In June, the company announced it would create a workforce academy to train skilled workers to help build its data centers. The free, five-week program guarantees participants a job at a Meta site upon graduation.
“So far, I think that all the work around AI has net created a lot of jobs because there’s all this infrastructure that needs to get created,” he said. “So that concern hasn’t exactly played out the way that people feared it might.”
The company will give investors an update on its business Wednesday when it reports second-quarter earnings after the bell. Analysts are expecting it to go free cash flow negative as it steps up its spending on AI infrastructure.
On past earnings calls, Zuckerberg has touted his company’s vision of personal superintelligence. In the interview, he defined that as “basically the combination of really strong and intelligent AI with context about you.”
In his own life, Zuckerberg said he uses AI to pull recipes together and order all the ingredients needed to bake with his 3-year-old daughter every weekend. He also uses it to help him stay in shape.
“I put cameras in my gym, and it watches me work out and can give me feedback on that,” he said. “It just helps out with all the stuff.”
News Corp, owner of the Journal, has a content-licensing partnership with Meta.