Blog 10: Policies for Creative Works
Published on:
Copyright in Generative AI
Case Study:
Why Copyright Is Not the Right Policy Tool to Deal with Generative AI
Case Study Summary
This case study dives into copyright issues, specifically regarding Generative AI (which they shorten to GAI). The article also discusses whom this issue effects, mainly citing that artists are at risk by GAI. They explain how current copyright law is not very well fit for AI usage, and expanding copyright law to cover AI is not the right solution. Instead of trying to change copyright laws, the study discusses how changes to data and consumer protection, and labor laws would work better. They explain how discussing intelluctual labor and forgetting about the “ghost workers” -the data workers that train models- is unfair and demonstrates how policy interventions would help both sides while copyright wouldn’t.
Discussion Questions
How should authors be compensated, and under which circumstances? If an author would be economically disadvantaged by AI using their work, there should be some sort of compensation to make up at least what the author would be losing. For example, I have seen some really unique artists on social media that do commissions, and if a model was trained on their data, it would bring less revenue in from commissions, as people would be asking the AI instead. If an author’s work is used in a very broad training, something like manga style drawing, it is not really taking away from the author’s opportunities.
Is training with unlicensed works a fair use, or an infringing one? Going off of the manga topic, I think that unlicensed works could be fair use if there is less creativity in it. If thousands of artists draw a similar style man but each give that man a different story, the creative part does not translate into what AI generates. This leads to a fairly ethical generation, where someone is not economically detrimented. However, when AI uses people’s creative work, and allows people to disguise generated images as their own unique work, there is an unfair effect on intellectual labor.
Are the outputs generated by GAI original or derivative works? Outputs are derivative works, because they use data from multiple places and its training to meld together one generation. I was recently thinking about how “stealing” intellectual property works, because many times in my life I have used other people’s work as inspiration. Throughout school when you are stuck on something and need to get creativity flowing, sometimes seeing what other people have done helps you. This is very similar to what AI is doing, so one could argue that humans can only make derivative works. However, I think the distinction is that humans recognize certain things as cheating rather than inspiration. If you look too closely at someone else’s work, and figure out the exact structure and makeup of it, and use it constantly when making your own work, it quickly changes into cheating. If you look at their work to understand the essence of it, you can use your own creativity to make something that resembles you more. This difference is important because what GAI does is what I have classified as cheating, and not inspiration, meaning it is derived and not original.
Should the outputs be entitled to some form of copyright protection? If so, how should we deal with AI authorship? Would someone be comfortable with their own idea patented and monopolized by someone else who heard of the idea? This would be unfair and uses someone else’s creativity for your own benefit. I don’t think generations are any different, they should not be able to have protections and benefit from someone else’s hard work. Even if someone’s work is not insanely unique, like the manga artist topic, or if it is not for their own economic gain, it does not deserve protection because no one really put in any work. A model simply used an algorithm to generate a mutation of different works, which does not sound like any authorship. AI authorship should only be used for entertainment purposes, and not as a replacement for doing real work.
My Discussion Question
How could policies change to protect creators from an economic detriment, while recognizing that inspiration happens within humans and is not the same?
As I discussed earlier, human work could be argued as derivative, causing an issue with understanding what is classified as stealing creative work and what isn’t. Artists nowadays sample other people’s songs by using lyrics, a melody, or a backtrack. This is legal and ethical, as artists have to negotiate the terms and need licenses. This is inspiration and allows an artist to make something part of their work, and not be the whole catalyst for their work. However, a lot of other examples exist that aren’t as strictly written into copyright law.
Reflection
This case study demonstrated the lack of a simple solution to GAI and its effect on artists or people that make some sort of creative thing. Changing copyright laws is not the right solution, even though I probably would have thought that it would be. I don’t personally make any sort of creative work, but I can see AI being used to steal from composers and use their creative work and publish it as their own. The issues of copyright and GAI require awareness of what ways generation can impact different groups and different types of creative work.