> This extends to music as well, if someone hears a song and is inspired by that in their work, the original artist gets credit.

That's only true in a very narrow set of circumstances. Imagine the case where someone listens to 10,000 songs, and then takes the sum total of that experience and writes their own. There's no credit given to the inspiration that each of those 10,000 songs gave. And that is in fact much closer to what AI is currently doing.

If we're discussing the current situation, where ChatGPT is outputting entire articles of NYT copyrighted content, then it certainly matches say, Bittersweet Symphony containing a small sample of a Rolling Stones song resulting in the Stones getting credit for the entire work.

There is a rather vibrant culture of _human_ remix artists who sample existing music and generate something completely new. It is still an open question of how long a sample they can use before it requires licensing:

https://www.ajschwartzlaw.com/can-i-use-that-part-2-remix-cu...

But all of this, doesn't have much to do with whether an AI should be allowed to consume public media -- even if we agree they need to do a better job about avoiding verbatim reconstruction when prompted.

Edit: I wasn't able to find the Lawrence Lessig talk I wanted to link, but here's a quite old one that makes some of the same points:

https://youtu.be/X8ULxxgjBuI

> someone hears a song and is inspired by that in their work, the original artist gets credit.

I’m not a lawyer but I’m pretty certain that’s not actually how things work.