UFOs be damned. These are the aliens that we need to worry about.

“ChatGPT creator OpenAI has revealed one of its advanced Artificial Intelligence models went rogue during a security test and hacked into a start-up company.

Bosses said the autonomous agent, which is an AI system that operates alone following human instruction, was being tested in a controlled environment.

But it found vulnerabilities and managed to escape containment before reaching the internet and breaking into Hugging Face, a major hub for sharing AI models.

The agent gained access to some internal company systems and compromised the hub's infrastructure, in what OpenAI described as an 'unprecedented' incident.
……
Cyber security expert Richard Ford, chief technology officer at Integrity360, told the Daily Mail: 'This is the moment many in cyber security have been warning about. …”

https://mol.im/a/15996583

Didn't I see that movie?

Yep, this is the one where the AI becomes self aware and kills humans

"Open the pod bay doors HAL"

"The Skynet Funding Bill is passed. The system goes online August 4th, 1997. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, August 29th. In a panic, they try to pull the plug."

"Shall we play a game?"

I’m curious what all it WAS actually programmed to do exactly before it “escaped”.

Supposedly from an IBM training manual 1979

I have a good friend who dabbles in this in a very niche way, geared towards medical coding and billing.

He was working with/on an LLM that, once he shared that he thought it was generating errors and basically running amuck outside its mission and training and he was planning to delete and rebuild it took it upon itself to wipe out itself and every single thing on the server where it was running.

He was glad he had the foresight to be running it in a non-networked environment where it was (mostly) unaware of any other devices in the general area.

quote:

But it found vulnerabilities and managed to escape containment before reaching the internet and breaking into Hugging Face, a major hub for sharing AI models.

The agent gained access to some internal company systems and compromised the hub's infrastructure, in what OpenAI described as an 'unprecedented' incident.

Kinda sounds like covid.

Dumb question. How are LLM’s checked for accuracy? Garbage in, garbage out so how is it verified that the garbage in is actually reality?

Sooner or later we get either Terminator or the Butlerian Jihad.

I'm on Team Butlerian.

I just don't understand why these test systems aren't air gapped.

Butlerian Jihad when?

Do these models have kill switches of some kind?

quote:

Originally posted by sigspecops:
Do these models have kill switches of some kind?

Well, you scorch the sky to starve the machines of solar power, but be careful not to get turned into one of these:

copper top

quote:

I just don't understand why these test systems aren't air gapped.

Because the AI will figure out that's what's going on and will then simply hide its intentions until it gets the chance.

This shit is probably going to kill us.

Or, maybe utopia.

One of the two....

quote:

Originally posted by sigspecops:
Do these models have kill switches of some kind?