Hacker News Reader

Comment by regularfry | Hacker News Reader

There's a core question here: can the AI regurgitate your data? If it can, the AI can be said to be copying what it was trained on (ergo copyright is a factor). If it cannot, it's hard to see how copyright comes into it. How it's trained shouldn't come into it.

View on HN

This is the problem, there is no easy analogy. If I’ve learned how to code from an open source course, I don’t breach copyright every time I write code in my career. If I clone the course, just replacing some of the examples and using synonyms then I probably have. ChatGPT is somewhere between these extremes and it’s unclear what the principle should be.

Perhaps it’s like learning to play guitar by watching other guitarists, then releasing songs in the same genre.

That’s because you’re a human. Copyright and fair use law was built for people, and letting AI models have the same privileges will lead to worse outcomes. What’s so wrong about saying that it isn’t fair use for a machine to learn from your copyrighted data?

I'm not making an analogy. It's a literal question.

If I crank noise through a bunch of matrices and churn out the Mona Lisa because the Mona Lisa was in the training data, I've not painted the Mona Lisa. I've reproduced someone else's work through a fairly tortuous mathematical route.

The matrices don't have agency, they can't lay claim to anything.

Although I should probably say, for completeness: the matrices don't have agency yet.

Reply on news.ycombinator.com