AI slowdown: The only AI guardrails you need are Trump, apparently
In an earnest blog post, Anthropic CEO Dario Amodei has called for a slowdown on AI development. The chatbot is just too cool, see: [blog post] We must slow the pace at which we improve the capabilities of AI models. You might think it’s been slowing already. But Amodei told CBS News Sunday…
In an earnest blog post, Anthropic CEO Dario Amodei has called for a slowdown on AI development. The chatbot is just too cool, see: [blog post] We must slow the pace at which we improve the capabilities of AI models. You might think it’s been slowing already. But Amodei told CBS News Sunday Morning: [YouTube] Dario Amodei: What it means, it doesn’t mean we need to panic today. It doesn’t mean we need to shut it all down. What I would say — Jo Ling Kent: Why not? Amodei: It’s a warning sign. It’s a warning sign that we need to slow down. This is the punchline to last week’s Jacob Coxon media push, where the chatbot would definitely kill humanity — maybe. If you took these guys at their word, you’d shut down OpenAI and Anthropic tomorrow and put Altman and Amodei in jail. But it’s really important you don’t do that, see: Amodei: I just don’t see how it’s a stable solution to never build the technology. What we need to do is not stop. We need to slow down. Sam Altman from OpenAI, Elon Musk from xAI, and Demis Hassabis from Google concurred. Microsoft didn’t weigh in. [Guardian] If you were some sort of cynic, you might think the models were running out of steam and the vendors were running out of money. The chatbots ran out of room for growth a while ago. OpenAI, Google, and Anthropic admitted in November 2024 that their chatbots couldn’t scale any more. In September 2024, OpenAI had posted a graph showing their new o1 Strawberry model needed exponentially more training for small linear improvements — even on their own pet benchmarks. The AI vendors need to reduce expectations, because they’re all feeling the pinch. And they want to hit the stock market. SpaceX stock is almost back up to its IPO price. OpenAI and Anthropic are trying to get it together for their IPOs. OpenAI has announced that the most obvious prediction ever has come true — Altman tells Fortune they’re delaying their IPO past 2026: [Fortune; YouTube] We’re not rushing into an IPO. I actually think that, given IPO timing, profit pressure, and pausing for safety, everything happening with safety, this would be right now would be an ill-advised moment to go public. It’s the safety, see, not OpenAI’s account books being a car crash or the company haemorrhaging executives. Anthropic wants to go public a bit sooner, and they’re loudly telling their investors they’ve had a second profitable quarter in a row! Yaaay! [FT, archive] Of course, that’s “profitable” using adjusted accounting, not any sort of Generally Accepted Accounting Principles. And Anthropic claim annualised revenue on a basis that isn’t clear, but is not one of adding up the actual dollars coming in. If you use made-up accounting, then Anthropic’s going great! Amodei’s already pitching for a government bailout, by making governments part owners or controllers of Anthropic — so they might prop it up: Kent: Would you be willing to give up the technology to the government? Amodei: To the right combination of governments. So, the worry I have – Kent: You’re willing to hand over this company. Amodei: I am concerned that one single government could abuse this technology just as easily as a single company could. But I think a combination of democratically elected governments, I don’t know about handing over, but some kind of oversight, some kind of joint governance. The US government already uses Claude in armaments — the US Department of Defense loves the chatbot, and is delighted it can kill people on Claude’s recommendations. The desired end game for Anthropic is a large enough contract to keep the company going without its current venture capital subsidy, even as the retail price of the API — that’s what you pay — goes through the roof. President Donald Trump is not happy with slowing down the AI, and he’s not happy with Amodei: [BBC] The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades! The Trump Administration has stopped AI ‘people’ from doing bad, or potentially bad, ‘things,’ like Dario (Anthropic!), who is now pretending to be a ‘perfect little angel’ — and we will continue to do so! There you go. Trump can do this job too. The obvious reading of that Presidential message is that when AI’s main market is once more the US military, as it was through most of AI’s history, it’s OpenAI that gets the contract. And not Anthropic. It’s not clear how well this will work out in practice. The present administration’s military adventuring is not working out so well. But the Pentagon is already asking OpenAI for a chatbot with “minimal refusal rates”. The use case for AI is an excuse for what you were going to do already. [Intercept; Google, PDF] Video — PodcastSource: Pivot to AI — Published — Category: Business