Mark Zuckerberg and Jensen Huang, heads of Meta and Nvidia respectively, have rejected calls from the heads of Anthropic and OpenAI to commit to an industry-wide slowdown of AI development in the name of safety.
In separate comments Tuesday, the two executives argued that individual companies bear responsibility for their AI models and any potential dangers they pose to humanity.
Huang was particularly sceptical of claims that artificial intelligence could take over the world, noting that AI is simply a piece of software. “Safety is an engineering problem, not a legal one,” he said at Salesforce’s annual Dreamforce conference. “We’re developing software after all. We’re developing computing systems after all. It’s a complicated computing system, but it’s ultimately a computing system.”
He added that he does not believe that increased regulation is the answer, suggesting that safety rests with model developers: “If you build a product or a service, and you’re not confident in its functionality, capability, or safety, then don’t release it […] that’s a very obvious thing to do.”
In a post on X, Zuckerberg expressed similar views. He argued that there are strong incentives for companies to release models that do what people want them to, both to prevent liability and to attract users.
The pair’s stances sit in stark contrast to those of Anthropic’s Dario Amodei and OpenAI’s Sam Altman, both of whom have called for a slowdown on AI development this week.
On Monday, Amodei said he was arguing for a slowdown to allow safety measures to catch up to the maturity of AI models. In particular, he said, models have a growing ability to help train their successors in a process known as recursive self-improvement, which he suggests could threaten developers’ ability to control future systems.
Zuckerberg directly addressed this claim in his statement, saying he believes that “the significant majority of compute” should be committed towards “serving people rather than racing towards recursive self-improvement”.
Altman backed Amodei’s calls, suggesting OpenAI would also slow the development of its models. He additionally pledged that his company, which is believed to be losing millions of dollars a month by some estimates, would not pursue a stock market debut until safety concerns were addressed.
The debate around AI safety has intensified in recent months following a string of high-profile cyber security incidents caused by frontier models. Those outside the industry have weighed in, too, with Nobel Peace Prize laureate Maria Ressa describing its development as a public safety issue earlier this week.
Another concern for US AI leaders is that of Chinese models, which have proven to be cheaper to run and similarly capable compared with those from the US. In his essay, Amodei stressed that any slowdown of AI development among “democratic countries” should preserve their “lead” over China, fearing risks to national security.
China, for its part, has emphasised collaboration. China’s largest newspaper, the People’s Daily, said in commentary Wednesday that AI is not “a monopoly of great powers”, and urged the US to work with Chinese companies to manage risks stemming from the software.
