Bill Gates. [Yonhap]
Bill Gates. [Yonhap]

Microsoft founder Bill Gates has sharply criticized calls for industry self-regulation of AI technology.

His central argument: AI is more dangerous than any technology humanity has ever encountered.

Gates said in an interview on The New York Times podcast "The Ezra Klein Show" on Wednesday (local time) that he cannot trust industry self-regulation on the issue. "That's really insane," he said.

Responding to the argument that companies releasing unsafe AI would face lawsuits and therefore need no external regulation, Gates said he was stunned by the question. "AI is the most dangerous thing humanity has ever encountered," he said.

Gates compared the absence of AI regulation to other industries operating without oversight — "no FDA to approve and regulate drugs, no aviation safety board, no requirement to put seat belts in cars," he said.

He also challenged the logic that liability lawsuits or self-regulation could contain AI risks. "Does that mean someone will make a biological weapon, kill 100 million people, and then we'll sue?" he said.

Gates argued that safety mechanisms and oversight systems must be made mandatory before a catastrophe occurs. On the labor market disruption caused by AI displacing human workers, he proposed taxing robots. He also suggested designating certain jobs as "human-only zones" — deliberately shielded from AI, much like nature preserves.

'AI powerful enough to trigger an event killing 1 billion people'

In a separate NBC interview released Friday (local time), Gates warned that "AI is certainly powerful enough to trigger an event that causes 1 billion deaths." While he said it would be "very difficult" for such a death toll to reach 100 percent of humanity, he added that "there has never been a weapon as powerful as what malicious actors could create by combining the latest AI tools."

Gates said the US federal government must pass legislation establishing safety mechanisms and oversight frameworks for AI development.

"Nobody thinks self-regulation alone is sufficient," he said, adding that while related legislation could place a burden on the industry, it would not significantly slow AI development.

Gates drew a distinction between the short-term risk of rogue states or terrorists misusing AI and the longer-term "existential risk" of AI threatening all of humanity. He also said last month that he anticipated disruption in the job market and again argued that certain occupations should be designated as human-only.


yul@heraldcorp.com