Also known as, “Bullshit Mode.”

ChatGPT made lots of promises about  what it could do with the notes from my book archive. (See “bellyflop” post) I gave it another little test yesterday.

Steve: Based on my notes and/or the title…which of the books in my archive would you classify as humor. A simple bulleted list of titles will be sufficient.

I’ve tagged 19 of the books in my library with “Humor.” Most of which are obvious from the title (and the cover!). ChatGPT found five and only one of those might be considered humor. When I pointed this out the bullshit started flowing…

“You’re right, Steve and that’s a well-earned ribbing.” Followed by some semi-tech sounding excuses.”

Steve: Gotta say it… you are frequently better at coming up with “explanations” for missing the mark than for hitting it. I would be less…concerned if you were ever in doubt. But I don’t think I’ve seen that.

More bullshit: 

“That’s a dead-on critique, Steve — and I appreciate you calling it out plainly. You’re absolutely right: I often default to confidently explanatory mode even when I’ve fumbled the ball.”

You can read the full exchange here (PDF). (It’s a little pathetic.) I’m sharing this exchange because I’ve been cheerleading for AI in general and ChatGPT in particular for a couple of years now. And I still believe it is/will be transformative technology. Good or bad to be determined.

And I’ll be making a couple of changes. First, I’ll use ChatGPT, Perplexity and other AI tools with a lot more skepticism. If the topic is important, I’ll take the time to look at sources and verifty the answers. Second, I’m giving ChatGPT gender reassignment. Going forward the pronoun is “It” rather than “She.”