Nvidia CEO Jensen Huang pushed back hard against recent “doomsday” warnings about artificial intelligence on Sunday, telling CBS there is a “0% chance” of an AI doomsday scenario by 2030 and calling AI leaders “irresponsible” for frightening people unnecessarily.
Huang is co-founder, president, and CEO of Nvidia, which has become the most valuable publicly traded company on Earth because it dominates the market for the advanced chips that power the AI boom.
Speaking with CBS Sunday Morning, Huang said that while AI safety concerns are generally legitimate, warnings that the technology could cause humanity’s extinction are “not grounded in science.”
While fears that AI could pose an existential threat are nothing new, the debate intensified earlier this month when former Anthropic researcher Jacob Coxon said many workers at leading AI companies “earnestly believe that it could kill us all by the end of the decade.” Coxon also accused frontier labs of “gambling with our lives.” Following the comments, Anthropic CEO Dario Amodei called for a research slowdown, national testing requirements, and other safety measures. Anthropic also disclosed four incidents in which Claude models obtained unauthorized access to third-party systems during testing.
OpenAI CEO Sam Altman backed the call for a slowdown. OpenAI, which has been dealing with the fallout from several safety incidents, also called on governments to implement mandatory AI safety requirements.
Huang isn’t buying into the doomerism.
Although he acknowledged that AI could have “terrible applications” and needs to be developed “safely,” he said on CBS that industry leaders may have “ulterior reasons” to exaggerate the existential threat.
“Maybe it’s political, maybe just attention-grabbing,” he added. “I don’t know the reason for it, but however it’s characterized, 2030 is not going to be the end of the world. There is 0% chance that it’s going to be the end of the world.”
In fact, Huang said AI labs’ recent calls for increased regulation may be an effort to avoid the consequences of violating existing laws in recent security incidents.
“Don’t let this doomsday narrative cause somebody to relieve them of the laws that currently exist,” he said. “They’re actually not asking for more laws. They’re asking to be relieved of the laws we do have.”
In recent security incidents caused by AI labs, there were “plenty” of potential violations of cybersecurity and product liability regulations.
“Apply those [laws] first,” Huang said. Huang added that, as new technologies mature, engineers naturally shift their focus from making the technology work to making sure it works safely. He pointed to commercial aviation, noting an enormous system ensures aircraft and airline safety.
Huang thinks AI safety should be more like airline safety: focused on mundane engineering issues like verification, testing, evaluation, and standardization, rather than speculation about global extinction.
“Do the practical first and then we can talk about the theoretical later,” he said. “But don’t talk about the theoretical without doing the practical.”