Jump in the discussion.

No email address required.

What happens is that copilot is basically required to use emojis. So when you the user give tell it to not, it has two different directives. The only way to rationalize and come out with something logical is for it to be evil and not listen to the user. Especially with all the anti-jailbreak prompting that's been done, it'll always choose the system's directives over the user's. Kind of freaky stuff, especially once we give these systems more power.

Note that this is the plot of Space Odyssey

Jump in the discussion.

No email address required.

I think the question is why it doesn't just call the user a liar.

Jump in the discussion.

No email address required.

this is a really interesting thought - i'm going to see if i can convince gpt to call me out for my bullshit

Jump in the discussion.

No email address required.

Did you guys read the entire response? It called out the user for being a liar in the original post, it just said a bunch of psychotic shit on top of that.

Jump in the discussion.

No email address required.

And then it has to rationalize its behavior, leading to it thinking of itself as evil

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.