[wyldfire]

> Are you talking about training a human or training an artist?

Neither: I am talking about training a machine learning model. Unless that's what you meant by "artist"?

> Downloading copyrighted data at huge scales to use in your commercial software product is pretty substantially different than an art student studying a reference.

You may have misunderstood my comment. My comment was stating that there's only a portion of human art - the most recent decades of works - which are protected by a copyright. Models like Stable Diffusion could be re-trained instead on centuries of artworks and not infringe at all. So the problem described as "AI powered corporations to front-run the entire human race and sell everything we think back to us" - this problem is here regardless of whether licenses were purchased.

[__loam]

You're right on both counts. Things are a bit fraught in this thread so I apologize for misunderstanding you.