zlacker

[return to "A federal judge sides with Anthropic in lawsuit over training AI on books"]
1. Fluore+bN[view] [source] 2025-06-24 20:51:07
>>moose4+(OP)
I'm surprised we never discuss a previous case of how governments handled a valuable new technology that challenged creative's ability to monetise their work:

Cassette Tapes and Private Copying Levy.

https://en.wikipedia.org/wiki/Private_copying_levy

Governments didn't ban tapes but taxed them and fed the proceeds back into the royalty system. An equivalent for books might be an LLM tax funding a negative tax rate for sold books e.g. earn $5 and the gov tops it up. Can't imagine how to ensure it was fair though.

Alternatively, might be an interesting math problem to calculate royalties for the training data used in each user request!

◧◩
2. dmix+VR[view] [source] 2025-06-24 21:21:11
>>Fluore+bN
That's a very different use case IMO. An LLM isn't generating a replica of a book for the users. At most we've seen people able to reproduce exact portions of stuff, but only with lots of prior knowledge of the material by the human in the loop and plenty of manual effort (aka not a direct commercial threat). And that was before more LLMs put effort into stopping that sort of hacking.

The last thing the world needs is more nonsensical copyright law and hand wavy regulation funded by entrenched interests.

[go to top]