The world is “right to be afraid of AI”, says Sam Altman, the chief executive of OpenAI maker ChatGPT.
But he said that people should “trust that we are going to do the right thing because it's the right thing and we feel the magnitude of this”.
Mr Altman and the AI industry generally has come under increased scrutiny in recent days, kickstarted by a warning from a former Anthropic developer that the “people building AI earnestly believe that it could kill us all by the end of the decade”. It has led to new calls for regulation and an increasing recognition among AI executives that it could be sensible to slow down development and focus on the security and safety of the technology.
Now, in a conference in San Francisco, he said that much of the concern was legitimate.
“It doesn't take as much imagination as it used to [...] to imagine how this could go wrong,” he said. “I think the world is right to be afraid of this.
There are “two big challenges” in making AI safe, he said: the danger of a loss-of-control accident or “some other serious thing that could go wrong”, as well as the danger that AI could concentrate power too closely and let specific people exert their worldview.
But he asked that people trust his company to be careful about the technology they are building. “The world should trust that we are going to do the right thing because it's the right thing and we feel the magnitude of this,” he said.
“It’s great for our industry to say we want to come together and we want to be able to coordinate and make sure we have enough time to do this safely,” he said. But he suggested that people get “very scared” because commercial pressures or racing to develop the technology could lead companies or countries to “not do the right thing”, he said.
The speech was Mr Altman’s first public remarks since the post by former Anthropic and OpenAI researcher Jacob Coxon sparked new fear about out-of-control AI. But they echoed a series of tweets that the OpenAI chief executive has posted in response to the growing calls for new safeguards and a slowing of AI development.
“The world deserves confidence that American companies developing increasingly capable AI will act responsibly, especially as the trajectory of progress has steepened,” he said in those posts. “Every frontier lab must deliver on this, and there is no reason any of us should come to work if we cannot.”
He said that the company would “welcome a federal framework that sets consistent safety requirements for frontier AI”. But he warned that companies do not need to work for regulation or government intervention to start building public confidence in AI safeguards, and that AI developers should first start doing what they can themselves to make their systems more secure.
