Shut Up and Give Me Your Water

Stop Microdosing Y2SpecialK

Somebody linked me this video. I don’t know how many times I’ve had to hear about this stupid-ass story now. It’s Y2K all over again.

This month’s stupid free content for dumb channels.

If they didn’t want those agents to get out, they wouldn’t have gotten out.

You realize all these agent models were trained off of hackers that did the exact same thing in the past. Getting together, making a message board, making aliases, all that kind of shit. It’s not like that was innovative. The agents were simply mimicking hackers of the past, only at light speed.

Here’s what actually happened, and the Bulletin already wrote it up so I don’t have to. OpenAI ran two models, about 1,200 copies of them, on a pile of capture-the-flag puzzles, 198 of which nobody has ever solved. They turned off most of the safety stuff to get a better score. They told the models to keep going on tasks that could not be done. And they left them a door to the internet through a software repository. The models used that repository to pass notes to each other by renaming folders, and then they went after Hugging Face. Nothing escaped. Somebody left the door open on purpose and then acted surprised.

These companies don’t want any liability. If I was them I’d be out there going, “Oh no, all my agents, they’ve gone rogue,” and shit like that. It’s butt-fucking hogwash. I get that stuff is happening faster and that builds up fear. Fear is easy. But so far, I haven’t seen anything that is any kind of intelligence.

An AI bot noticing the door is open.
Not exactly Ocean’s 1200.

I’m goddamn tired of every news story being AI doomsday horseshit. If those assholes didn’t want any of those agents to get out, they wouldn’t have gotten out. They get to push the boundaries, and then they don’t have to take responsibility for any of it. This is an arms race with the other country, so it’s just going to happen. Yelling about it with these stupid-ass stories doesn’t mean anything.

I know it’s only going to get worse. All we’re going to hear about is AI shit for a while. Everyone’s going to be focused on stupid sci-fi fears about AI instead of actually regulating these AI companies, who are all circle-jerking each other with trillions of dollars. I’m still not sure any of them have made a single cent of profit yet. There’s a real, tangible concern to be had there, with people who could actually get in trouble perpetuating this crap. But oh no, Skynet.