Gregory's Server@mmasnick layman legal hypothetical; if this model were trained on data that was SET to be public domain, but wasn't at the time of the model being created, but the model wasn't released until the sources were public domain.... Is it copyright infringement?
|
3 comments
Resuna
Pretty clear that the weighting produced by the training process is a derived work of the training data. |
@eikenberry well, i've been arguing that training a model does not implicate any copyright rights, so I will stand by that and say no. But, others might disagree. Courts will let us know...