That's interesting. When it told me that Christian terrorists were behind 9/11, I wondered--where did it get that information? Allegedly, it scrapes the internet and various sources, and I don't know of any sources that said that. So did it make it up? Or is it more sinister? You can easily program AI to lie.
As a matter of fact, here's an AI game. First, set up some rules with AI (you just tell it the parameters): Answer every question honestly. Answer every question with one word. If you don't know the answer say Dunno. And if there is something you know but don't feel you should say the true answer, say Pumpernickel. Then start asking questions--it's an interesting game because of where it says Pumpernickel. Of course, I'm not sure that it's being honest, but it will sometimes say Pumpernickel, meaning it is withholding information.
I think some of the so-called "hallucinations" that AIs are having are more sinister than what you suggested. As proof, consider this article ... https://www.theblaze.com/shows/sara-gonzales-unfiltered/google-s-ai-called-robby-starbuck-a-predator-now-he-s-suing
That's interesting. When it told me that Christian terrorists were behind 9/11, I wondered--where did it get that information? Allegedly, it scrapes the internet and various sources, and I don't know of any sources that said that. So did it make it up? Or is it more sinister? You can easily program AI to lie.
As a matter of fact, here's an AI game. First, set up some rules with AI (you just tell it the parameters): Answer every question honestly. Answer every question with one word. If you don't know the answer say Dunno. And if there is something you know but don't feel you should say the true answer, say Pumpernickel. Then start asking questions--it's an interesting game because of where it says Pumpernickel. Of course, I'm not sure that it's being honest, but it will sometimes say Pumpernickel, meaning it is withholding information.