Blog 10: Policies for Creative Works
Published on:
Case Study:
ARTificial: Why Copyright Is Not the Right Policy Tool to Deal with Generative AI
Todays’s case study takes us in the legal wonderland of GAI ethics, specifically how AI-genrated contents might make it even harder to apply copyright laws in the technological realm, an area that has already caused enormous expansion of the copyright doctrine.
The author explores the legal and philosophical dimensions that AI-generated content poses by answering to a number of questions that I will go into detail below. The main idea that the author explores is the difference between GAI and human content in asserting what is creative work and what is not.
To begin, the author focuses on how artists whose work is used to train AI should be compensated and in what circumstances. Compensating artists for the outputs that AI gives is complicated in the sense that the AI systems need an unimaginable amount if data to become fully functional. The dilemma that the author raises is how do we know which artist deserves compensation per output from billions of data? In my opinion, we need to think about compensation in a completely different way when it comes to artists whose work is being funnelled into AI models. I think what needs to happen is for the artists to be compensated in terms of shares of the AI companies they have helped build. However, I realize that the proposed method is a bit of a far reach in today’s capitalist world. The authors have little power because the big corporations take away the profits of their work. A practical step that can help steer into the right direction where the authors are properly compensated would be to unionize in order to make their voices heard. In rerms of the circumstances in which authors should be compensated, I think this depends on how influential their work is for a specific element of culture. This can be either in terms of quantity or quality. Quantity for example if as a photographer, you have contributed a large number of puictures that are then used to train image generation models. Quality-wise, we can consider specific works that are iconic in our society such as the Monalisa. Even if it is just one piece of art, it still holds a great value in what is considered an artistic painting.
In the heart of copyright laws is originality. Another million-dollar question that today’s author tackles is: Are GAI outputs original or derivative works? I really enjoyed going through the author’s discussion of the question especially their opposition of the commonly held belief that AI content is plagiarism. The author asserts that AI content is rather on a gradient that is contextual in the eyes of the beholder. As AI content becomes more and more undistinguishable from human content, it would be unfair to still treat both contents differently simply because one is machine-made. I completely agree with the author’s take on the dilemma of originality, but would want to add that in the legal context, it would make more sense to consider the data that AI gets trained on as inspiration rather than plagiarised content. This way we can then develop a scale of how much similarity between an AI and a human generated piece is too much for us to sanction a GAI system as plagiarising or not.
When copyright is involved in AI content creation, an ethical question arises that helps us reflect on what the future holds. Today’s case study asks if assigning copyright to AI works would create an imbalance in the public domain due to the massive scale and velocity at which AI can operate. The author raises a critical concern if copyright is extended to cover AI content. Considering the ability of AI to create a vast amount of content in a short period of time, assigning copyright law to such content could restrict a big part of our society who cannot afford premium content in what we can use and cannot use in our pursuit of knowledge. That restriction would be a heavy infringment on our rights to expression as well, especially to the small players who cannot afford paying for the rights of content they created using AI because the content is protected by copyright law. Furthermore I think it would be a shameful hypocrisy to assign the copyright law to content that was generated using copyighted content to which the AI companies had no rights to. For example, OpenAI has come under fire numerous times for using copyrighted material to train their AI chatbots without the consent of the copyrighed material’s owners. From this perspective I do not think AI content should receive any special treatment when it does not respect the copyrights of others in the first place.
There is also a discussion about the use of unlicensed works in the training of AI. The author sets themselves to explore if the use of such works can be considered a fair use or an infringing one. Despite the author’s attempt to present the question as complex as possible, I see a very clear response to the matter. One of the author’s points is the fact that AI models do a job that is far from a creative one when learning from data that it could be considered a fair use. Although I agree with this specific way of looking at it, the fact that the AI companies are privatized hence keep all the profits to themselves does not sit well with me, all the while, they give little to no compensation to the owners of the works they realy on to train their AI models.
From the discussion of the concerns above something has kept lingering in my head. How feasible is it for the media industry to come together and create an AI model that makes sure artists are fairly compensated and enjoy the benefits of their work? It is so sad that today’s AI companies are privatized and do not give much importance to artists whose work are the real backbone of AI systems. I kept thinking of the above question as I was trying to make sense of the complex discussion of the fair use doctrine so I think it would be a great reflection question in that domain.
All in all today’s case study was such an eye opener. I really enjoyed the author’s navigation such a complex legal system in the digital media realm and their extensive challenge to the status quo. My main takeaway is how copyright laws have stretched for a long period of time to fit new forms of media and how GAI content is no different.