Major media companies including Disney and Warner Brothers are suing providers of generative AI services like MidJourney on copyright grounds as they claim that these AI companies are using their ^intellectual property^ as part of their training data1 and that presumably they are somehow harmed when users of the AI tools generate images, text, or video.
If I am not explaining their complaints adequately then that’s because I don’t know why they think that they can use their copyrights to stifle innovation.
These media corporations are making surprisingly maximal claims about the power over thought that they are allowed to wield; it’s as though they think that no one is allowed to make a creative work that is even influenced by one of “their” properties in any way.
Perhaps Disney and a few other mega-corporations are trying to lock down the very ability to tell stories. I don’t know, and I am not taking a side between them and their antagonists like MidJourney and OpenAI in their litigation. Already the author’s guild and the New York Times are suing OpenAI, with more lawsuits likely to follow. But looking at this battle in terms of how it will affect copyright case law, I am very hopeful that the AI companies could use their financial heft to wear the media companies down into a humiliating ‘no-fault’ settlement or a too-late (or losing) court verdict a decade from now.
For now, the “AI boom” means lavish funding for generative/LLM entities like OpenAI and Anthropic2, and incumbent tech giants3 (especially Google and Microsoft) are much more on the side of artificial intelligence tech and its applications than on the side of copyright enforcement. By leveraging their extraordinary political and financial power, big tech could cripple copyright in the courts or even pass legislation that waters it down in the name of progress.
Maybe a copyright term of something like 7 years would be ok, but the current system has tilted the balance way too far in favor of corporations and profits and away from individuals and the public domain, and a de facto abolition of copyright law done by the AI companies in the courts might be the best policy option on the table in the near future.
“training data” is a slightly technical concept that IBM defines as “information that is used to teach a machine learning model how to make predictions, recognize patterns or generate content”
No posts

Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.