2 ms·
They didn’t train it on every available copyrighted work though, but on a specific set of legal questions and answers. And they did try to license them, and onl
by echoangle 2y ago
They didn’t train it on every available copyrighted work though, but on a specific set of legal questions and answers. And they did try to license them, and only did the workaround after not getting a license.
- synthetic-echo 2y agoI think they were talking about the "model giants" like OpenAI you mentioned. Not saying they're correct, but I will concede the amount of copyrighted information someone like OpenAI would want is probably (at least) an order of magnitude more than this particular case.