Anthropic CEO Dario Amodei Calls For AI Slowdown
- Reference: 0185623986
- News link: https://slashdot.org/story/26/09/12/1738240/anthropic-ceo-dario-amodei-calls-for-ai-slowdown
- Source link:
> The chief executive of Anthropic [1]called for a global slowdown of artificial intelligence development in a 3,800-word essay on Saturday, just days after one of the company's employees [2]quit over concerns about the safety of the technology. Dario Amodei, who co-founded Anthropic to focus on securely and carefully building A.I., [3]wrote that while he believed the technology could bring many benefits, it was advancing at too quick a pace for researchers to continue safely.
>
> "Over the last few months, I have become convinced that fully addressing the risks requires even more prudence -- not just investing in risk prevention, but pacing the rate of capabilities advancement so that risk prevention has time to keep up," Mr. Amodei said. "We must slow the pace at which we improve the capabilities of A.I. models. Progress will still seem fast, and we must make wise use of the time we gain." [...] "Left unchecked, it could outrun our ability to understand and control these systems, and so must be pursued very carefully, if at all," Mr. Amodei said.
>
> [...] In his essay on Saturday, Mr. Amodei suggested actions that the industry might take to slow down the pace of development. Mr. Amodei said all A.I. labs could agree to third-party technology assessments from "embedded evaluators," or outside specialists who can verify best safety practices across companies. He also suggested that countries with democratic governance systems coordinate to create safety standards, which could take the form of regulatory action. He added that it would probably require a global effort working with other nations, including authoritarian ones, to properly coordinate a slowdown.
>
> Mr. Amodei stressed in his essay that he still finds A.I. capable of bringing "incredible benefits" to humanity, including potentially curing diseases and accelerating economic growth. But even so, Mr. Amodei said the risks of A.I. were too great to not proceed with extreme caution. "The measures I propose to advance the frontier at a safe pace will not be easy," Mr. Amodei wrote. "But I believe we owe it to humanity to try."
Amodei's essay comes just hours after Bloomberg reported that Sam Altman told OpenAI employees the company is [4]open to slowing the pace of AI development amid similar concerns.
[1] https://www.nytimes.com/2026/09/12/technology/anthropic-dario-amodei-ai-slowdown.html
[2] https://slashdot.org/story/26/09/09/1943258/anthropic-researcher-believes-more-than-10-chance-ai-could-kill-all-humans
[3] https://darioamodei.com/post/we-must-pace-the-frontier
[4] https://slashdot.org/story/26/09/11/1746211/altman-considers-slowing-down-ai-development
Re: (Score:2)
I think that this is really a data center spending ceasefire to delay the bubble popping rather than these companies suddenly growing a sense of responsibility.
We all know the real reason (Score:3)
Training these models is incredibly expensive. In fact it's so expensive that there is no viable economic model to do so. As such, their options are either to drastically slow down the amount of new models trained or to go bankrupt. This is just a nice whitewashing of that. It remains to be seen if even using their models can be profitable, but we know that with the cost of training the next one they can't be.
Re: (Score:2)
It remains to be seen if even using their models can be profitable
Whether or not he stated the reason for the slowdown honestly (which don't think anyone should believe, and even the Coxon departure could have been staged), then one concern is that use of AI (cloud providers, specific models) might only be profitable or be maximally profitable only for bad actors, for example hacking systems for financial gain or other purposes.
The much more likely explanation seems to be that they can't deliver
Re: (Score:2)
I don't think anybody who uses Claude regularly, think that Anthropic is failing to deliver on their promises. It isn't perfect, but it is a *very* powerful AI. I can generally tell it what code I want written, just like a Product Owner would, and it gets it mostly right the first time. It generally writes better code than I would.
This seems like marketing to me. "Our AI is so powerful, it's dangerous!" This has been Anthropic's theme for quite some time now.
Re: (Score:2)
Why should we think this is anything more than a publicity stunt?
Anthropic has a long history of releasing statements about how "dangerous" their AI is. When Mythos came out, they made a big deal about it being so dangerous, they needed to wait to release it so it wouldn't fall into the wrong hands.
"Our AI is so powerful, it's dangerous!"
Just more clever marketing, IMO.
Continue the current amount of work (Score:1)
Slowing down implies doing less. They need to increase safety and that could be an equal amount or more of work.
Squires... (Score:2)
...wielding a Knight's sword. We need to train. Has anyone seen Pei Mei?
How would they even know? (Score:2)
We already know that these companies aren't entirely certain what their AI trial balloons are doing at any given time. If all the companies got together and agreed to "slow down" ... how would we even know if they did? How would we even know if they intended to slow down, and then whether they actually took actions to slow down and actually understood whether they were even on top of what they were previously doing?
I'm not saying that I think there are independent rogue AIs operating in there just waiting
yeah, right (Score:2)
Both of the 2 domestic frontier labs are screaming doomsday about AI, while concurrently asking for government regulation (which keeps getting batted down), and (presumably) funding senators to put bills in place to make model distillation a felony. While they're trying to IPO. "Oh, we really need to slow down the industry".
What the hell is going on here? In what world does this make sense - where a company making AI, and is trying to IPO, is blasting the internet with AI doomerism?
The only conclusion I can
Like a bunch of boys playing chicken (Score:2)
Even a bunch of punks progressively playing chicken more and more dangerously eventually have someone say, "Hey guys, this might be a bad idea." AI companies have progressively done less and less ethical things to this point... it's not like these are saints.
They fucked the environment. They fucked people's copyright. They fucked the people living next to the hyper scalers. They are about the fuck the economy.
Having your own software on your own computers hacking into private servers is a great place to
Bullshit. (Score:2)
> it could outrun our ability to understand and control these systems,
What is he talking about? They have no ability to understand OR control these systems.
Here's a better idea, how about the go government does it job and start prosecuting corporations that hack into other companies? Think a slow down is a good idea? I guarantee that a $100B dollar fine will cool your heels really fast.
Re: Bullshit. (Score:1)
This. Hoping we can control superintelligent AI when we have no experience controlling superintelligent AIs is like hoping you can invent automobiles and never ever have any wreck, not even once. Because if you wreck--if your super AI escapes long enough to establish itself--then it's game over for humanity.
The sheer audacity of this lie (Score:1)
Is something I find especially infuriating. There was a time long ago when people would put a little effort into the lies they were telling the public but these days they don't even bother.
ublock (Score:4, Funny)
I never played with custom filters before but holy shit:
slashdot.org##article:has-text(/\bAI\b|artificial intelligence|ChatGPT|OpenAI|GPT[- ]?[0-9]|Claude|Gemini|Copilot| machine learning|generative AI/i)
is going to make slashdot readable again.
Slowdown so it's not the thing that kills us all? (Score:2)
I guess he wants to let nuclear weapons and climate change catch up.
All together now... (Score:2)
Let's pirate the world
Let's speed up development
Let's spread out and gobble resources
Let's hack
Let's have a brief outage
Let's slow down and step back
Why such coordination?
We are going so fast we need to slow down! (Score:5, Insightful)
"Definitely, that's... that's what we're doing. We definitely didn't vastly over promise investors that we could replace all of humanity within a few years and are scared they'll sue us and win when they find out that was bullshit. Nope, not what happened. We're doing this deliberately, of our own free will. No suing."
Re: (Score:2)
If they were genuinely concerned, this is exactly what they'd do. So what makes you think they are doing this out of something for investors and not actually telling the truth about their motivations?
Re: We are going so fast we need to slow down! (Score:4, Interesting)
One word: China. Chinese AI models are already available open weight (essentially open source equivalent). Many of them were created via large scale distillation attacks against Anthropic and OpenAI. They are competitive in quality with American models. A slowdown deprives China of this key source of training and data for their models. They can then build in secret internally until the models are sufficiently advanced, and defenses against distillation significantly hardened; so they can take the lead. All the while not having to immediately show advancement over rivals to investors. I've never met a Silicon Valley company that decided to slow down for reasons that weren't also aligned with business interests.
Re: We are going so fast we need to slow down! (Score:2)
open weights is nothing like open source - it means you can download the model to run it locally, like you can download a binary to run it locally.
Re: We are going so fast we need to slow down! (Score:1)
I agree, but also: open source isn't really a viable concept for frontier AI models. You can't release the training data because it undoubtedly contains unsavory, politically-misaligned, copyrighted, and illegal content: you'd drown in p.r. and legal problems. And on the recipient side, nobody can afford billions of dollars of GPUs to make their own fork.
Re: (Score:3)
That could be making sense, if Chinese only use distillation to train their models. However, if they truly use data and train their modes properly, then they will not slow down and will get significant time benefit, if the Western companies slow down. This idea to slow down does not look good at all, unless it is absolutely clear how suddenly everybody is slowing down.
Re: (Score:2)
Nothing these people have ever said has been true.
What makes the think they're lying? Their lips are moving. It's Theranos all over again, only with multiple companies, and hopefully, the end result will be the same: prison sentences for fraud.
Re: (Score:1)
But first you eat their debt.
Re: (Score:2)
They've done irreparable damage to both the AI industry but the broader "tech" industry, to be sure. Trust has been greatly violated.