Post

“The monster is so hungry it might attack, unless you sacrifice one of your villagers” is a classic fairytale

484 reposts1.6K likes

Why is every media organization reporting on this absolute nonsense without checking first that Amodei isn’t just talking out of his ass?

Imagine if a self-driving taxi company CEO said, "oh, by the way our cars might start mowing down pedestrians and baby strollers when they see them?" as some sort of weird flex to investors about how powerful and "alive" their product is.

They keep warning about things they could just stop doing. Sure someone else might do them but at least they wouldn't.

What kind of idiot... adds their OVEN to their phone? Who needs remote access, to the appliance in your house that can burn it all down, if unattended?

The selling point was, if you're on vacay in the south of France & you forgot to turn off the burners on your stove, you can turn it off remotely to keep it from burning down the house. The fine print that makes it your fault if your phone pranks you by turning it back on? Priceless.

It’s such shameless fearmongering because if this is real, you can simply unplug the appliances, turn off your WiFi, block the IP address from your router, or a million other things. Also do they really think a fridge would have the storage available for an AI to run smoothly on it??

the headline is confusingly phrased as though it might be saying "... UNLESS we retain access to data centers", but I believe it's actually arguing "... REGARDLESS OF access to data centers". so it's still totally good to: deprive them of data centers, and regulate the hell out of them.

FWIW they aren't saying that data centers would help prevent replication. They're just saying that AIs can replicate even without access to a data center. They're warning that the monster is going to attack regardless of what the village does, in other words.

The way I’m reading the statement the company is making a direct threat. Will we have to endure two more years of this or can a Dem Congress do something?

The argument they have right-wing religious leaders & lawyers working on constructing is that the combination of being evil & self-replicating is proof they're human & therefore legally liable like all the rest of us, so techbro developers and corporations should be untouchable.

I can't get this "out of control AI" narrative to make sense. Cool, you put a chat-bot on software. It doesn't make it alive. Now you gave it a bunch of automated hacking software? Now you just let it re-program itself? You're the A-hole than... This is buggy software.

They're gonna make these things, because they don't give a shit about whether or not it should exist - just whether or not they can make one. So they will. And then it'll hurt people, and they'll go "Oh, but /we/ didn't hurt anyone! We even warned everyone!" and expect to get away consequence free.

If this is true, the only action moral human beings can take is to start blowing up AI company servers. I’m not kidding. If these goddamn engineers can’t keep their own monsters under their own control, we are going to have to destroy those monsters before they hurt all of us

Alas, delusional thinking in AI world is a crisis: bsky.app/profile/pile...

pileque.bsky.social@pileque.bsky.social

“It’s obvious that these AI models understand what they’re saying. You couldn’t possibly answer all the questions without understanding the question. It’s ridiculous to say they don’t understand.” - Geoffrey Hinton, "Physics" Nobel winner. The man is delusional. www.theguardian.com/us-news/2026...

Actually, every cyber threat assessment meeting I ever attended at the Cyber Command brought up the danger of what happened if the AI found it could "live and reproduce" uncontrolled on the internet. And it is still the greatest danger, if only because doing so would cripple essential services.

If an AI needs a whole data-center to operate on, then it can't fully self-replicate into smaller platforms -- but it can spin off bots and worms to infest and commandeer such machines. Best response: shut down data centers. Or, you know, maybe bomb them? Clear and present danger and all that.

They have built the perfect thief. That damned Claude stole $200 from my PayPal account. Both PayPal and Anthropic have no humans in the customer service department. Enter the Minnesota Attorney General and the Star-Tribune for some crowd justice. I see many others have been ripped off too.

I ask chat gpt to create me a list of reported a.i hacks around the world over the last 30 days. It covers GOVERNMENT S, banks, grocery stores and software companies. It's pretty long. Then i ask if hacking was crime. You already know the answer

sounds like anthropic want to scare everybody away from the technology so they can profit from it all by themselves.

This is very short term thinking. "When we cut the grid" means agreeing, for instance, to sacrifice the most vulnerable villagers (e.g. babies in neonatal ICU) ourselves, in advance of the monster doing it for us.

Were Anthropic a private citizen, law enforcement would be treating this as a terrorist threat. So, despite the John Roberts granting corporations (which does come from the Latin corpus, body, whatever) personhood, that doesn't extend to criminal liability, of course.