Hacker News Reader

If the US makes it illegal to train LLMs on copyrighted data, the US will find a solution and not just give up and wait half a decade to see what China does in the meantime.

View on HN

What solution is there?

Zillow have the MLSs network that provide them lists, a similar solution could apply if courts agree that library copies apply for this - Anthropic could sign agreements with large libraries and "check out"/"freeze" copies for a minimally-agreed-upon duration and query across all to see which has a copy of each book they need. Spotify and Apple Music sign deals en masse with labels, the same could apply here with book publishers, labels for lyrics, museums for art, etc. Or whatever other creative solution that people who will need to find, will find. Right now they took the laziest path, because it worked. They will find the next-laziest path that works.

And the easiest option: Legislation change. If it's completely decided that the current law blocks LLMs from working in the US, the industry will lobby to amend the copyright law (which is not immutable) to add a carveout for it.

You're assuming that people will just give up. People never gave up, why would they now?

Reply on news.ycombinator.com