Anthropic CEO Dario Amodei has once again maintained his stance on slowing down the pace of advanced AI models. In a recent interview with CBS News, he highlighted that the rapid progress of AI is becoming a warning sign for the industry. According to Amodei, AI may bring serious risks and argued that companies must focus on testing and independent oversight before launching AI models.
Here’s What The Anthropic CEO Said
Amodei told CBS News, “I won’t lie to you – there are real dangers. And I think for too long the industry lied to people about the fact that this technology had risks.” These comments come from the tech mogul after a researcher from Anthropic, Jacob Coxon, warned that both OpenAI and Anthropic may be taking an extremely dangerous approach by developing self-improving AI systems.
Reportedly, Amodei said that he is surprised by how quickly AI models have advanced. As per him, the development of AI has followed an exponential curve, with the pace of progress now becoming even steeper. Amodei reportedly asserted, “I don’t think I fully just appreciated what it would actually be like when the progress was as fast as it was.”
Despite his call for a slower pace, Amodei clarified that Anthropic does not intend to stop releasing more powerful AI models. Instead, he reportedly said the focus should be on ensuring that every new generation is properly evaluated before being released.
“It doesn’t mean that. What it means is that we need to make sure that every generation of models that we release is properly tested,” Amodei said. To strengthen that process, Anthropic plans to provide independent evaluators with permanent access to its models. Amodei compared the approach to inspections in other industries.
“Whenever anyone builds an AI model, there’s a third-party evaluator who can verify whether that company is following the safety practices that they’re committed to – it’s like a food inspector,” he said.
Amodei Backs AI Regulation
Amodei also said that independent testing alone may not be enough and expressed support for federal regulation of AI. Several ideas have been discussed by policymakers, including proposals for restrictions on superintelligence and systems that could allow authorities to shut down dangerous AI.
When asked about an AI “kill switch”, Amodei said, “The AI kill switch could be a good idea. But I don’t think banning it, completely banning it, is the right approach.” He explained that completely stopping the technology could also mean losing potential benefits, including advances in medicine.
“One, the idea that we would never get the incredible benefits, never get the medical cures, that doesn’t sit right with me. Second, this technology will be built in other countries, including authoritarian countries,” he said. Notably, one of the biggest challenges, according to Amodei, is deciding whether the US should slow down AI development if China continues moving rapidly.
He described the situation as one of the most difficult questions surrounding the technology. The Anthropic CEO reportedly pointed to agreements between the US and the former Soviet Union over nuclear weapons as a possible model for international cooperation on AI.
“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 in 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,” he said.
