Discussion about this post

User's avatar
Marc Friedman's avatar

Nice article. I have been experimenting with ChatGPT. Some things have been very good, some fair, and some awful. As you may know, my book reviews are published on Substack. After I finish a review I ask Substack to write a review of the book. It disclaims knowledge of books published in 2022-23. But for earlier books it will write a review. They are terrible. In many cases ChatGPT just makes up stuff like inventing characters and plots.

Expand full comment
Andrew Bee's avatar

That was fascinating. It feels as though Chatgpt has been taught that any answer is better than no answer.

I kept waiting for you to ask it why it said one thing then contradicted itself in the next answer although I suspect that, had you done so, it would have informed you about the impending failure of the AE35 unit and you would be off to Mike’s scenario ;)

Expand full comment
5 more comments...