Monologue: Shut The F*ck Up, Dario!

Ed Zitron (0:01)

Causome Media Hello and welcome to this week's Better Offline monologue. I'm your host Ed Zitron.

I wanted to do a more reasonable monologue, one where I carefully pick apart the AI safety debate, skewering the fact that nobody seems to actually be talking about making things safer, or curtailing the actual harms of LLMs, but I reached the point where the only thing I wanted to say was that Dario Amodei, CEO of Anthropic, needs to shut the fuck up. Dario Amodei, you have spent years threatening people's livelihoods, suggesting that AI would wipe out 50% of white collar jobs, and when one of your nasty little rationalist goblins quits your company and scares everybody, you've doubled down, saying that for too long the industry lied about AI's risks, without saying what they were, just that progress is fast and we've hit an exponential. Shut the fuck up, shut the fuck up. You build cloud software, you're in full control of what gets built or not. Just because you don't fully understand how LLMs make their decisions doesn't mean they're out of control or have their own intentions, and your unwillingness to explain the difference and further fuel anthropomorphism only makes me more certain that you need to shut your fucking mouth. I read in the Financial Times that you're telling investors that you have 80% gross margins if you leave out the cost of training, revenue share with Amazon, Google and Microsoft and stock-based compensation, to which I have to say shut the fuck up, Dario. Shut the fuck up. I am sick of it. I am sick of hearing these fucking stories where you mislead people about your company, about your economics and about your damn software. You make cloud software. Shut the fuck up, man. You are running your company in a slimy, dishonest and nasty way, conning investors and the general public with the full knowledge that the tech and business community won't bother to push back.

How about the fact that you are constantly changing the rate limits on your users' subscriptions, or that there are tangible reports that Anthropic overcharges its customers? Does that matter to you? God no, you know it doesn't. You have created a cult of personality on top of the actual cult that exists within your company, an Arnold Palmer of effective altruism and rationalist types. Where is your name in Yudkowski's Harry Potter fanfiction? That's super strange. You are a strange guy, Dario.

I realize that as the CEO of a multi-billion dollar company, you probably don't hear direct feedback like this very often. So let me repeat myself. Shut the fuck up. You make cloud software that's both unprofitable and unreliable. While it's got better, I'll absolutely say that. I don't really give a fucking shit. You've made all of these promises over $500 billion worth. You've told people that you're birthing the machine god. You are not birthing the machine god, Dario. And to act as if you are is either deeply cynical or completely deranged. If you truly believe that you're building dangerous AI, you should treat it as such. Shut down Claude or heavily restrict it so that it can do very little other than the operations you consider completely and utterly safe. But we both know that your definition of safety is entirely aesthetic. You don't give a shit about anyone. If you did, you'd recognize that you, along with OpenAI and Meta, have allowed your dangerous little chatbots connected to near infinite resources and some sort of harness to run endless scripts to hack things at scale. And then you have the fucking temerity to try and frame it as if this is AI making its own decisions versus LLMs prompting each other.

Again, if you were truly concerned about safety, you'd blow Claude's brains out. You'd annihilate that French motherfucker before he could yell Alonzi and hack another website. Non, Monsieur Amodei, I'm just Claude Software, he cries. He said, no Claude, I'm afraid you're too dangerous and send him to hell.

The way I know you're in here is that you're not actually taking any meaningful steps to mitigate the harms of LLMs, nor are you calling for anyone else to do so. You could really easily get a PR win by saying, hey Sam, why does your LLM make people kill themselves? Because it's absolutely fucking happening. The quickest step would be to call an immediate halt to any and all cyber security model training runs. Except, you know, you can't do that because Mythos is your big sexy enterprise product. The multiple sources tell me is just fine and way too expensive. It would be so easy to call out Clammy Sammy and say, it's time to stop doing these experiments. Except your best idea is to get fucking METR to order everybody's models. Wow, the independent regulator that you help get funded that has multiple former Anthropic staff on board and is piloted by any number of different less wrong posters. You won't get away with this bullshit in any kind of sane or actually rational media climate.