Clearly they're developing something new here


[ Follow Ups ] [ Post Follow Up ] [ UCLA Open Forum ]

Posted by blindness on August 24, 2026 at 16:05:43

In Reply to: Autonomous AI Attacks, yippee! posted by ClockBlocker on August 24, 2026 at 14:39:12

The thing about AI being bullsh*t has some to do with the stated goals of what it is supposed to do for us, and some to do with the fact that LLMs at least, do not have any internal representation of the world have no understanding that the sequence of characters it's dealing with correspond to anything out there that we call "realty" (hence the hallucinations).

That said, AIs going out there and doing things is a very real risk. It was something I first heard about in this video 11 years ago:

Another one tha made an impression on me is this one, same dude, miraculously the same outfit, two years later:

I'm not sure if this new generation of AIs have handled the "internal reprsentation (= modeling) of the world" problem. They are ayt least acting like there's some movement on that front.


Follow Ups:



Post a Followup

Name:
Email:
Password:

Subject:

Comments:

Optional Link URL:
Link Title:
Optional Image URL:


[ Follow Ups ] [ Post Follow Up ] [ UCLA Open Forum ]