00:00The people building the world's most powerful AI systems are now calling for stronger safeguards
00:06as concerns grow that the technology is advancing faster than we can control it.
00:11Antropic CEO Dario Amodeh has proposed independent audits of the most advanced AI models
00:17following repeated warnings from the company's own researchers
00:20that the technology could kill all of humanity by the end of the decade.
00:25Is artificial intelligence moving too fast for us to control?
00:30Because it's now going so fast that the researchers themselves have become terrified of it.
00:33And that's what we've seen in the last two or three weeks.
00:35It's the people who work at the companies who are coming forward.
00:38And the call is inside the house saying, like, what we're doing is risking catastrophic damage.
00:44So I think we have to take them seriously.
00:46It doesn't have to kill us like a robot doesn't have to kick in your door and shoot you.
00:50It has to simply destroy the systems that keep us alive in modern society.
00:54So imagine that all of those systems go down simultaneously, right?
00:58That's a mass casualty event, even in the United States, much less the developing world.
01:02So that's one way.
01:04A second way people talk about is in the next few years,
01:07someone vibe-coding some kind of dangerous super virus.
01:11Again, that seems kind of preposterous.
01:13It does require a few links in the chain to still be created.
01:17But that's something that many people fear.
01:20Amodeh wrote in an essay published on Saturday that the US government should write regulations
01:25requiring independent audits of the most advanced AI tools known as frontier models.
01:31He outlined a three-step framework intended to slow the pace of AI development
01:35and create more time to manage its risks.
01:38The proposal was immediately endorsed by several tech leaders,
01:42including Elon Musk, who runs ex-AI, and OpenAI CEO Sam Altman.
01:47I think Washington, D.C. needs to assert its control over the industry.
01:51And it needs to be regulated minimally like we do banks.
01:54So banks that have systemic risk to the financial system,
01:57we have embedded examiners there, making sure they don't make risky decisions.
02:01We need that for the frontier companies.
02:04OpenAI, Anthropic, Gemini, the Google product as well,
02:08perhaps Meta and SpaceX AI too.
02:11So that's the minimal thing we should do.
02:13US President Donald Trump, however, has so far resisted concerns about AI
02:17and said what was most important was keeping the US in the lead
02:21against China in the intensifying AI superpower race.
Comments