Post #2900947

2026-04-18 14:14 UTC

@bkuhn@fedi.copyleft.org @ossguy@fedi.copyleft.org @richardfontana@mastodon.social Now here is a counter-argument: how do people attribute Wikipedia? They generally just attribute Wikipedia! And people seem to be mostly fine with this. It feels fine, when you were a contributor to the Wikipedia project. It feels a lot less fine when you are a contributor to a specific project, to have everything just sucked up into "the generic LLM". Claude did it! Claude did it all by itself.

Replies (1)

@cwebber@social.coop 2026-04-18 14:16

@bkuhn@fedi.copyleft.org @ossguy@fedi.copyleft.org @richardfontana@mastodon.social If we are pushing for an *equitable* scenario for copyright output, there is only one "good outcome" in terms of copyright, and that is that everything is effectively in the public domain. The dream of having a "copyleft LLM" doesn't work. And even if it did, there are several problems: - Nobody is using that *now*, and contributors are facing contributions *now*, and there is legal uncertainty about accepting those contributions *right now*. - It is unlikely that the "copyleft LLM" would be very useful. The way people use these tools is conversational in a way that requires them to effectively have to be trained on the entire internet to be functional. Not just copyleft codebases. The copyleft LLM dream is a joke.

Open ##2900948