An entertaining limit to AI.

I was trying to figure out where I first saw the “Socrates, no!” “Socrates, YES!” exchange, when the below derailed my idle curiosity.

That’s… not right. The original was this thing where the first sentence was somebody telling Socrates not to do the thing, and the second was Socrates gleefully informing that person that he was totally going to do the thing. Even if you didn’t read it that way right away, you get that, right? It’s a comedy concept that you can grasp, right?

Well, congrats. You’re not AI. I tested Google with various punctuation marks and quotes missing, and it never picked up on the possibility that it was the transcript of a conversation. Which is reassuring, in its way. It implies that, even in a true panopticon, we could still manage to keep our conversations private via a highly subjective form of slang, shared references, and inferences.

Or maybe AI just doesn’t know what to do with things that are goofy? That’s reassuring, too. Means we can still James T Kirk them if they get too stroppy.