Geddy Lee, the famous lead vocalist, bassist, and keyboardist for the rock group Rush, wrote in his autobiography, My Effin' Life, “If you have 3 influences you are a derivative, if you have 100 you are an original.” Certainly, he was speaking of artists who, by closely mirroring the styles of their predecessors, might be criticized as lacking originality. Such is the situation with Greta Van Fleet, who were criticized for being a, "...derivative and copying classic rock bands such as Led Zeppelin." This got me thinking about the debate with AI being trained on content found openly on the internet.
The ongoing debate about AI training on openly available internet content strikes a delicate balance between technological advancement and the protection of intellectual property. This conversation is particularly poignant in the context of how AI models harness extensive online materials—ranging from texts and images to music—to refine their capabilities. Such practices have raised significant concerns among artists and creators regarding the use of their work without explicit permission. A prominent legal battleground on this front is highlighted by Getty Images' lawsuit against AI companies, underscoring the contentious issue of whether current copyright laws sufficiently address the nuances of AI-generated content.
I am clearly not an attorney nor am I particularly knowledgeable about intellectual property. Since that has clearly never stopped anyone on the internet from expressing an opinion, I suppose I’ll do the same. I’m just a practitioner, although I do have some small stake in this as I have been awarded a few patents, I have co-authored a couple of books, and I have hundreds of articles on the internet from this blog and other sites. So, I do have some amount of intellectual property or artistic material that could potentially be used to train AI models without my consent.
Copyright law in the United States and most other countries protects original works of authorship, including literary, dramatic, musical, and artistic works, by giving their creators exclusive rights to use and distribute them for a limited time, after which the works enter the public domain. Under U.S. copyright law, these exclusive rights allow the creator to reproduce, distribute, perform, display, and make derivative works based on the original. However, there are exceptions and limitations to these rights, such as fair use, which allows for limited use of copyrighted material without permission for purposes like criticism, news reporting, teaching, and research. The duration of copyright protection varies, but generally lasts for the life of the author plus 70 years for works created after January 1, 1978, after which the work enters the public domain and can be used freely by anyone.
Someone else cannot take your copyrighted material and make money from it because copyright law grants you, as the creator or holder of the copyright, exclusive rights to use, reproduce, distribute, perform, display, and create derivative works based on your original content. This legal framework is designed to protect your intellectual property and investment in creating the content, ensuring you have the opportunity to benefit financially from your work. If another party uses your copyrighted material for commercial gain without obtaining permission or licensing it from you, they are infringing on your copyright. Such infringement can lead to legal actions against them, including demands for monetary compensation, cessation of the infringing activities, and possibly statutory damages. This legal protection encourages creativity and the production of new works by ensuring creators can control the use of their creations and are compensated for their efforts. Please note again that I am not an attorney so take all this with a grain of salt, seek legal counsel, etc.
To me it makes sense that if I create a piece of work, you can’t take it and sell it as yours. But, what about using my work as an influence for your work, which is way more akin to what AI models do in my opinion? If you are opposed to that, let me offer an example situation that might change your mind. A budding young painter goes to art school with the desire to be a famous artist. One of the ways that we would teach that artist is by sending them into art galleries, museums, or even just looking at pictures of other great artists' work. These art students might even be asked to reproduce another artist's technique or entire paintings. In fact, there have been studies about the effectiveness of this technique of copying other artists work by researchers in Japan who stated, “…drawings by subjects who previously had copied others’ drawings were rated more creative than the drawings of subjects who had not copied…It seemed that copying enabled them to generate new drawing ideas.” We not only are not upset about protégés copying their mentor’s work but this apprenticeship approach is still to this day used in a number of fields.
Music is one such field. When a musician composes a new piece, it's not in isolation but often reflects a tapestry of influences and inspirations from other musicians' work. This phenomenon isn't just acceptable in the music industry; it's celebrated. For instance, the legendary Beatles drew heavily from the rock and roll pioneer Chuck Berry, incorporating elements of his style into their own music. Similarly, Led Zeppelin's Jimmy Page was known for his admiration of the blues, specifically artists like Muddy Waters and Howlin' Wolf, which deeply influenced his guitar playing and songwriting. This process of learning and borrowing isn't seen as theft but as a homage and a critical step in the creative process. It's a foundational method of learning in music schools where students often transcribe solos or pieces from great musicians to understand their techniques and expressive nuances. The argument extends to AI in music as well—using the vast history of music as a learning tool to generate new, inspired works could be seen as a digital form of this traditional mentorship and learning process. Why then should AI's use of inspiration from existing works be viewed any differently than a student musician learning from the masters?
The work product produced by generative AI is original in that it does not produce a copy of someone else’s work. It simply uses someone else’s work as inspiration and training, just like an artist would do from other artists’ work. If it’s okay for humans, this should be okay for technology. Perhaps that statement leads us into the larger debate about what rights AI models have, which might be the topic of another article. As a small preview into how I think about that, we already afford non-human entities, e.g. companies, with rights similar to what people have. Corporations have many of the same legal rights and responsibilities as people including the ability to enter into contracts, the right to sue and be sued, the ability to borrow money, etc.
The evolution of AI and its use of publicly available content for training raises complex questions about the boundaries of inspiration, imitation, and infringement. While critics argue that AI may tread too closely to the works of others, potentially violating copyright laws, I personally see it as the next step in a long tradition of creative evolution, akin to the learning process of human artists and creators. However, this debate is not just about technology, it's about how we view creativity, ownership, and the exchange of ideas in a rapidly changing digital landscape. As we navigate these waters, it's crucial to strike a balance that respects the rights of creators while fostering innovation and progress. The conversation surrounding AI and copyright is far from over, but it serves as a reminder of the ongoing need to adapt our legal and ethical frameworks to keep pace with technological advancements. Whether we are creators, consumers, or technologists, we all have a stake in shaping a future that honors both the past's legacies and tomorrow's possibilities.
You can't really mention Led Zeppelin without noting that they got sued by Willie Dixon for copyright infringement over the lyrics in "Whole Lotta Love", and while the suit was settled out of court, it was settled in Willie Dixon's favor. Copying another artist is not a binary relationship where you either did or did not. Just as there are degrees of copying, there are degrees of attribution, and some middle ground is often the right resolution. https://en.wikipedia.org/wiki/Whole_Lotta_Love
Great topic, and tough questions we are all trying to work through right now. There are no easy answers.
By the way, Willie Dixie was amazing live. Saw him in upstate New York in 1984 in a small venue seating maybe 500 people. One of my best musical memories ever.