Sounds like more marketing BS.
What are they selling?
The same thing they’ve always been selling: a false promise that their AI models will solve all humanity’s problems.
“This AI is so good it’s too dangerous.” Is absolutely just made up BS to make their AI models sound more powerful than they are. It’s trying to drum up a sense of mystique and sophistication for their brand.
Imagine a car company announcing they have invented a new engine that’s “too fast to put in their cars”. That would be trying to manipulate people into thinking, “Wow, it must be so amazing, I want it.” and “Their engineers must be so smart and skilled to create something so powerful. [brand] has the best engineers.”
If this were true, and they were actual researchers, they wouldn’t keep it “under wraps”. They would release it to the corporations with a deadline for patching, then release it to the public after that deadline. You know… like any other good cybersecurity researcher would.
Ugh okay it’s really not at all hard to trick these models. You can get them to help you build all manner of guns, bombs, whatever, if you tell them you’re a regulator trying to understand how to prevent people from doing that. They are not intelligent, they’re just gating certain types of questions. Be creative and you can do whatever you want
Grifters upgrading from pseudo-science and anthropomorphism to straight up chaos majick.
Here’s a much better incantation: Blood Incantation
“incantations”… as in “prompts” right?
Does these incantations involve labubu and summon Trump or Satan?
