@raph I don't agree with this framing.
It's not generating verbatim copies of the pages it trains from.
Isn't the work of ChatGPT more like how a person can read, learn from what they read, and then rephrase or recombine that knowledge in their own words?
I don't feel like I'm stealing when I do that.
It also seems a lot like a search index bot -- finding and processing information that's published online and using it to make a derived product.