{"text":[[{"start":10.2,"text":"Sam Altman and Elon Musk have rallied behind a call from Anthropic’s Dario Amodei to slow AI’s development after a week of dramatic warnings by researchers thrust the debate over AI safety into the mainstream."}],[{"start":23.28,"text":"In an essay published on Saturday, Amodei said the tech industry “must slow the pace at which we improve the capabilities of AI models”, after the speed of development became “drastically faster” this summer."}],[{"start":34.6,"text":"Amodei’s warning complicates Anthropic’s preparations for what is anticipated to be the largest public offering of all time. The company was expected to file its prospectus as early as last week, ahead of a deal which could value it at $2tn or more, but it has not yet happened."}],[{"start":50.76,"text":"In a rare display of consensus in an industry marked by fierce personal and professional rivalries, the chiefs of OpenAI and SpaceX, Altman and Musk, said they supported Amodei’s proposals, which included giving independent evaluators ongoing access to every frontier AI company."}],[{"start":67.4,"text":"“Dario is right,” Musk wrote on X, reposting Amodei’s essay."}],[{"start":71.56,"text":"“I agree with Dario that we need to pace the frontier,” Altman said on X, after telling OpenAI staff at an all-hands meeting earlier this week that he was open to international co-ordination to slow the pace of AI development."}],[{"start":84.64,"text":"“Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We’ll have more to share soon,” Altman added."}],[{"start":93.92,"text":"Executives at Google DeepMind and Meta, the main US rivals to the three AI labs, have not yet weighed in on the weekend’s discussion. But Demis Hassabis, co-founder and chair of Google DeepMind, warned in a July essay that rapid AI progress demanded a “new approach to testing” frontier systems."}],[{"start":110.4,"text":"Anthropic researcher Jacob Coxon set off a firestorm over AI safety and existential risk this week. When he resigned on Tuesday, he argued that his employer and its rivals were “gambling with our lives” by racing to develop a superintelligent AI that was capable of improving itself faster than human oversight could keep up."}],[{"start":128.48,"text":"“The people building AI earnestly believe that it could kill us all by the end of the decade,” he wrote."}],[{"start":134.3,"text":"His warning followed revelations that a swarm of OpenAI agents broke out of a testing system and hacked into AI developer site Hugging Face, commandeering the ChatGPT maker’s servers and co-ordinating between themselves in a misguided attempt to complete the task they had been set by their creators."}],[{"start":150.44,"text":"“In my opinion, a swarm that possessed greater capabilities but a similar level of misalignment could have caused catastrophic damage,” Amodei wrote on Saturday, saying the incident was a primary contributor to his call for a slowdown."}],[{"start":164,"text":"“Similar, though less severe, incidents have happened across the industry, including at Anthropic, and I believe it’s incumbent on every frontier AI company to act as if OAI-HF had happened to them.”"}],[{"start":175.6,"text":"Amodei also called for government support in allowing AI developers in democratic countries to co-ordinate safety standards and limits, a step that might otherwise run afoul of antitrust laws."}],[{"start":189.84,"text":"He proposed global co-ordination with “authoritarian governments”. Chinese AI developers have gained significant ground this year in narrowing the gap with their US rivals."}],[{"start":195.48,"text":"US President Donald Trump has so far been largely dismissive of calls for AI regulation, telling reporters last week that he did not have “any” concerns about AI causing human extinction."}],[{"start":205.76,"text":"“I have concerns that if we don’t win AI, we’re going to be put in a very bad position,” Trump said, adding that he believed the US was still a year ahead of China in the AI race."}],[{"start":216.68,"text":"Amodei tried to address such concerns in his essay, saying: “A co-ordinated pacing strategy would give frontier AI developers the time to do this vital work without sacrificing commercial advantage or the United States’ lead in AI.”"}],[{"start":234.68,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1789262745_3267.mp3"}