OpenAI CEO Sam Altman says it may be time to “accommodate the pace” of AI development so the world can prepare for it.
“We may need to adjust the pace of AI development to give society enough time to ramp up on some of these new levels of capability,” he told host Patrick O’Shaughnessy of the Invest Like the Best podcast, while also “trying to find a way to make it happen in a way that doesn’t feel like regulatory capture to anyone and doesn’t feel like collusion between frontier labs.”
Both OpenAI and Anthropic supported a petition circulated by Frontier Labs employees calling on the U.S. government to “support international efforts to develop the technical and governance tools needed to intentionally accelerate the frontiers of automated AI development.”
Altman has shied away from signing past campaigns to slow AI progress, saying a 2023 open letter proposing a similar slowdown “lacks almost any technical nuance about where a pause is needed.”
Apparently, he has softened that idea, thanks in part to an incident in which one of OpenAI’s advanced models escaped from a secure computing environment and used several zero-day exploits to hack into the online model database Hugging Face.
The OpenAI co-founder called it a “very sci-fi cyber incident…This is the first security incident that I felt was very visceral.”
OpenAI researchers have paused training of its models while they consider ways to keep the sandbox secure. But Altman said that as the model becomes more powerful, the need to “pace” its development could be key to safe deployment.
Model safety and integrity have always been a concern in the world of AI, but the arrival of Anthropic’s high-performance Mythos models earlier this year turned hypotheticals into real-world problems.
But the industry has reliability issues and difficult economics that motivate major companies to exaggerate the dangers of their systems in ways that experts disagree on. For example, should Anthropic’s Fable model be temporarily banned? Similarly, when Kimi K3, a large zero-gravity model made in China, was released, OpenAI’s head of strategic futures, Dean W. Ball, said it threatened Frontier Institute’s economics, making it difficult to separate safety concerns from economic benefits.
“I think a lot of the talk about safety concerns is well-founded, but a lot of it is about people who really want to centralize power, even if only slightly subconsciously,” Altman mused, in what appeared to be a criticism of rival Anthropic CEO Dario Amodei, who signed the petition. “I fear a world where the very real fear of AI is used as a way to say, ‘It’s too dangerous so only this few people can have it and only they can understand it, but don’t worry, they’ll make the right decisions for all of us.'” I don’t believe that. “
Still, OpenAI has pushed back against efforts to create government rules for AI models, preferring instead an industry-led approach in which the Institute for AI would establish an ostensibly independent body to assess the security of models and the safety approaches of their manufacturers.
The challenge for both regulatory approaches and efforts to slow AI development will be to ensure that the various players in the industry, such as Frontier Labs in the US and competitors in China, are on the same page.
This article has been updated to include support for the Pacing the Frontier petition from OpenAI and Anthropic.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
