We’ve done this before. Generative AI doesn’t have to be unethical

When victrolas and other music players were first introduced, they weren’t very good. As the technology improved, they slowly replaced live performances in many but not all situations. We still have live orchestras and live bands even though most of the music we listen to on a daily basis is recorded. Players also made it possible to listen to music in situations in that were never practical with live performers, such as while commuting or waiting in line. But when Napster facilitated the theft of copyrighted music, it was, to put it mildly, a huge deal. In music, it is possible to recognize and protect intellectual property while acknowledging that new tech has outcompeted many individual professionals.

The poor performance of today’s AI art and text generators is temporary. I fully expect they will improve within the next few years. But what will not go away on its own is the ethical issue: They were made using stolen material, much of it copyrighted. While we live under capitalism, we should respect its rules: Intellectual property is property and taking it without permission is stealing.

It would be completely possible for OpenAI to negotiate with Studio Ghibli for a license to use its work and to go to Ghibli’s competitors if it should refuse. There is precedent for OpenAI to make a quarterly, annual, or other periodic payment to every author and artist whose work they used to build their model and draw in their profits.

If AI companies acted ethically, we would still have to face hard questions about the role of writers and artists in a changing century, and about their unconscionable fuel consumption, but they would have opened up a new revenue stream for those writers and artists, possibly making these career paths more viable instead of less so. And they might, like cylinders, records, tapes, CDs, and MP3s, open up new ways for us to experience that art.