• d@nny disc@@circumstances.run
        link
        fedilink
        arrow-up
        1
        ·
        48 minutes ago

        @dgerard it set legal precedent on taking a copyrighted work out of its original context. scanning texts for upload is very similar to the search interface provided by the google books case which is very similar to AI training

        • Cricket [he/him]@lemmy.zip
          link
          fedilink
          English
          arrow-up
          2
          ·
          30 minutes ago

          Oh, just watch, lawyers can find very creative ways of splitting hairs. “The output was sufficiently transformed from the training inputs to be considered an original work”. The one legal question that I think will be a tougher hurdle for them to overcome will be whether AI output can be copyrighted, although I don’t know how much tougher. I happen to think that that one will require houdini-like contortions to justify granting it copyright protections, but I’m not sure. Again, lawyers can get very creative.