Merely because some regurgitation of content can be achieved by overly narrow propmting does not make the generation of the training to be infringing on it's own.
Could OpenAI be guilty of making that overly narrow prompty too easy? Maybe. But it's that narrow case, not the broader argument that "all training is infringment".
The copyright situation around music is a lot more restrictive because there is a lot more actual law around music. Specifically around the rights of song writers having different rights then performers and the recordings of their performances. The result is that the window of fair use and non-infringing creation is a lot more constrained.
Techdirt has not posted any stories submitted by yokem55.
Merely because some regurgitation of content can be achieved by overly narrow propmting does not make the generation of the training to be infringing on it's own. Could OpenAI be guilty of making that overly narrow prompty too easy? Maybe. But it's that narrow case, not the broader argument that "all training is infringment".
Music has a lot of other protections
The copyright situation around music is a lot more restrictive because there is a lot more actual law around music. Specifically around the rights of song writers having different rights then performers and the recordings of their performances. The result is that the window of fair use and non-infringing creation is a lot more constrained.