"Stop making those harmful predictions."
Nvidia Chief Executive Jensen Huang delivered a pointed rebuke of Geoffrey Hinton, the University of Toronto professor emeritus known as the "godfather of AI." Huang said the bigger problem is not the risk of AI itself but the fear that groundless doomsday scenarios instill in the public.
"A Scientist Saying It Doesn't Make It Science"
In an interview on "The Ezra Klein Show" released by The New York Times on the 23rd, Huang voiced strong frustration with the AI risk arguments Hinton has been raising.
Hinton, a pioneer in artificial neural network and deep learning research, is called the godfather of AI. The 2024 Nobel Prize winner in physics has recently said that uncontrolled AI could bring about societal collapse and even threaten human survival, putting the odds at roughly 10% to 20%.
Huang took issue with those figures first. "A 10% probability is not based on science or research," he said. "A scientist saying it doesn't make it science. Those predictions are harmful." He said he would tell Hinton that "saying things like that is irresponsible," and criticized the professor on the grounds that his past predictions about the future of AI had also failed to pan out.
What worries Huang is the social fallout such warnings could produce. If people repeatedly hear that AI will take human jobs and ultimately threaten humanity itself, he argued, younger generations may grow pessimistic about the future altogether.
Citing cases in which young people hesitate even to go to college out of fear that AI will make it harder to find work, Huang said, "You shouldn't assume that sounding the alarm helps society." He added that "all of us need to be smarter, more mature, and grounded in objective evidence and science."
"Zero Chance the World Ends in 2030": Repeated Attacks on AI Doomsaying
This is not the first time Huang has publicly criticized AI doomsaying.
In a CBS News interview on the 18th, he dismissed claims that AI could advance to superhuman levels within a few years and wipe out humanity, declaring that 2030 will not be the end of the world and that the chance of that is zero. He went on to call frightening people unnecessary and irresponsible, criticizing such claims as grounded in neither fact nor science.
Huang was pushing back against AI risk arguments raised by former Anthropic researcher Jacob Coxon and others. As concerns spread that AI could slip beyond human control amid rapid technological advances, Huang characterized them as an excessive "doomsday narrative."
Huang does not deny that AI carries risks, however. In the CBS interview he said he was not arguing that their concerns were wrong, making clear that he shares an awareness of safety issues. At the same time, he stressed that developers should move as fast as possible but no faster than they should, warning against accelerating AI development at the expense of safety.
Hinton Calls It a "Wild Guess" as Views Split Over Risk Preparation
Hinton himself does not claim the 10% to 20% figure is the product of precise scientific calculation. In a CNN interview in August, he explained that there is not enough basis to calculate exactly the probability that AI will pose a serious threat to humanity, describing his own figure as a kind of wild guess.
Even so, Hinton maintains that society must prepare for the worst case. In a recent closed-door briefing for U.S. members of Congress, he is reported to have warned that humans could lose control of the technology if lawmakers fail to put AI safeguards in place within about a year.
The clash between the two men is less about whether AI poses risks than about how forcefully uncertain future dangers should be flagged. Hinton holds that risks tied directly to human survival warrant advance preparation even if the odds are low, while Huang counters that warnings built on probabilities lacking scientific grounding can inflict a different kind of harm on society.
In the CBS interview, Huang also stressed that companies should not release products that are not safe. His argument is not that AI's risks should be ignored, but that safety should be addressed on the basis of verifiable evidence rather than fear.
