The Ravens vs. The Swallows is a new section of Original_me about the tweets, events, posts etc, that got my attention last week. While the Ravens take a cautionary or even a gloomy look at the future of AI, the Swallows tweet more positive, optimistic notes. This time we look at the Github Copilot case on the Ravens side and a quick summary of the presentation given by Dr. Tijn Borghuis from Eindhoven University of Technology about the way out of the IP-mess AI has created from the Swallows corner.

The Ravens: People vs. Github Copilot

What happened? Mr. Matthew Butterick had enough. He is now preparing a lawsuit against Microsoft owned Github, because its AI called Copilot is producing results that are eerily similar to copyrighted works. As Butterick argues, Microsoft can no longer hold that their AI using the open source Github repos is "fair use" of the work. It is time to acknowledge the works of real authors.

Why does it matter? If that lawsuit is brought forward, and if it succeeds, then it may set a precedent on how the AI models can (or cannot) use copyrighted works for training purposes. And as one of tweeters, @wadus says, if you use Copilot, you may put your client at litigation risk. Butterick is not alone in this fight, given the huge amount of support tweets for his cause. Just try to search Twitter for Matthew Butterick and the results speak for themselves.



The Swallows: Why not declare the results of the generative AIs as public domain?

What happened? The University of Leiden held a really interesting conference this Friday: The Law and AI. One of the presenters, Dr. Tijn Borghuis from the Eindhoven University of Technology, explored the implications of intellectual property rights issues when training and using generative AI. While he is personally engaged in developing a music AI, he is also exploring the aspects of IP in his research. While their project bought the rights to use the authors' musical samples, he acknowledged that defining the authorship is not easy to solve right now. The most likely outcome is that the author, in the traditional sense, cannot be identified. Different legal directions are being explored at the moment, including declaring the AI itself as the author. But currently, the works of AI most likely fall into the public domain, or open source if you wish.


Why does it matter? This presentation showed that the work of AIs may not be as black-and-white as the current legal systems would like to have it. We may need to start stretching the legal horizons to accommodate technological progress. But there's still hope in this mess. As Dr. Borghuis argues, there's more to gain socially when the creative works of AI is left in the public domain than making them copyrightable.

Dr. Tijn Borghuis about the state of IP in generative AI field.