After reading Kevin Roose’s excellent piece in the NY Times about the AI chatbot Claude 2, I decided to finally try the ChatGPT alternative. Of course, I started with my favorite subject: me.

So, what did the “next-generation AI assistant for your tasks, no matter the scale” tell me about myself? Here you go:

Not surprisingly, the results are a mixed bag. Yes, I’m American and have written the books mentioned, including The Age of the Platform. Yes, I’ve written for HBR and MIT Sloan Management Review, but never for The Financial Times.

Whoops

As for hallucinations, get ready:

• I neither live in San Francisco nor teach as an adjunct at a university there. (In fact, I recently passed on several opportunities to return to academia.)
• I never taught at the graduate level at ASU, nor have I taught coures on innovation.
• I had never even heard of WiredPen, but I suspect that Claude “knows” that I wrote a bunch of pieces for Wired a few years back.
• Although I cover generative AI in one of the chapters in The Nine, I’ve never penned an AI book.
• I never studied philosophy, and I have never skied. I haven’t hiked in years.
• Worked in politics? Not in this lifetime. Maybe the other Paul Simon?

Maybe Claude’s new version is confusing me with one of the other Phil Simons, like this one.

Español

Claude did mucho mejor when I tested my Spanish skills. (I used to be fluent.) Note how it responded below when I asked it to stop showing me answer-related emojis and only give me verbs:

Simon Says

Are generative-AI tools impressive? Hell, yeah. No doubt that they’ll rapidly improve as well. Still, I won’t be making any key life or business decisions based on bots’ recommendations and answers anytime soon. Ditto using it to write for myself and my clients.

Nor should you.