Amodei says the AI industry lied about the risks for too long
In a CBS interview, Anthropic CEO Dario Amodei explained the worry behind his call to slow down. He called China the toughest dilemma and said he is uncomfortable that a private company is building the technology.
Anthropic CEO Dario Amodei spoke to CBS's "Sunday Morning" after publishing his call for an AI slowdown on Saturday. The interview was filmed at the company's San Francisco headquarters.
"I won't lie to you, there are real dangers," Amodei said. "And I think for too long the industry lied to people about the fact that this technology had risks."
"Not panic, a warning sign"
Amodei said AI development has been on an exponential curve that is starting to get steeper. "It doesn't mean we need to panic today. It doesn't mean we need to shut it all down. But I would say it's a warning sign. It's a warning sign that we need to slow down." He also admitted he had not fully appreciated what it would be like when progress was this fast.
Asked whether Anthropic will stop releasing more advanced models, he said: "It doesn't mean that." The aim is to make sure every generation of models is properly tested.
To that end, Anthropic will give independent evaluators permanent, "employee-like" access to its models. Amodei compared it to a food inspector: whenever a company builds a model, a third party can verify whether it is following the safety practices it committed to.
Yes to a kill switch, no to a ban
Amodei said he supports federal regulation. Proposals in the US range from an outright ban on superintelligence to a mandatory kill switch. "The AI kill switch could be a good idea," he said, but he opposed a complete ban for two reasons: giving up benefits such as medical cures, and the fact that the technology will be built elsewhere, including in authoritarian countries.
China, "the toughest dilemma"
Asked whether the US should slow down if China does not, he said: "This is the toughest dilemma. This is the toughest aspect of our situation." He pointed to nuclear arms agreements with the Soviet Union as a possible model, while keeping expectations low: "The more long-term thing would be working together to put a speed limit on the rate of AI progress. I think that's going to be very difficult, because the incentives to pull ahead and the military advantage that you get from that are so large. And honestly, I don't know if it's possible, but we should try."
As CNBC noted, Chinese President Xi Jinping on Sunday called for a consensus-based global AI governance framework at the BRICS summit in India. Trump and Xi are due to meet at the White House on 24 September and are expected to discuss AI.
"Very strange that a private company is building this"
Asked about his personal responsibility, Amodei said it has always been strange that this technology is being built by a private company: "People ask me that question all the time: why isn't this being built by government? And the strangest thing about it is, I agree with them. I'm uncomfortable."
Asked whether he would hand the technology to the government, he answered: "To the right combination of governments." He said a single government could abuse it as easily as a single company, and that some form of joint oversight by democratically elected governments would be better.
