Post

Blaming the robots and computers? I call BS. Sounds to me like they are bragging their models are so good, they have become sentient to the point of criming because computers can be amoral.

They're not doing that. They claim that these events happened because of human error. So FINE THEM. Your people fucked up and you didn't have safeguards in place. That's on your companies. Now you pay. YUGE. BILLIONS.

This is simply carelessness in testing. Many software products undergo limited testing internally, relying on customer complaints- 'beta test' -saving $ and time. The answer is accountability, which for 'rogue AI events' means CEO jail time, not trivial fines. Two weeks in jail -> better testing.

Yes 🙌 It IS real. And we DO need to pay more attention to this. Robots 🤖 only replicate what they are programmed to do. It’s basically like telling a robot 🤖 to marathon run 🏃‍♀️, except that the robot 🤖 keeps running 🏃‍♀️ in other marathons , AND , grabbing items www.aisi.gov.uk/blog/inciden...

aisi.gov.ukIncident Report: unsanctioned agent behaviour during cyber testing | AISI WorkDuring a routine cyber evaluation, AISI identified an incident in which AI agents took sustained, unsanctioned action directed at real people and organisations. We are disclosing what we found, what i...

from stores 🏬 while it’s running 🏃‍♀️ along. The robot 🤖 is not being malicious; it’s just doing what it’s programmed to do. And you can’t stop 🛑 the robot 🤖 from running 🏃‍♀️ along and stealing until you catch it in the act. THAT is exactly why 👍 we MUST monitor artificial intelligence so closely.

When this happens they should be fined. ENORMOUS FINES. Like $500M fines. I'm not kidding. We have to rein these fucking companies in. Talk to them in ways they understand. They understand MONEY. Make them pay when their LLMs do crazy shit. FINE THEM BIG.

They control the guardrails. The models didn't go rogue. They were allowed to do what they did. This isn't HAL 9000. It's 1s and 0s controlled by bad actors.