The Intellectual Property Dilemma of AI Art

As many of you may know, an AI-generated artwork called “Théâtre D’opéra Spatial” recently won first place in the Colorado State Fair’s fine art competition. Its author Jason Allen didn’t actually draw it, but automatically generated the piece using a text-to-image tool called Midjourney and printed it on canvas at his local print shop. Allen is not an artist in any form, and he’s even just getting started with AI painting. Like me, Alan only heard about AI painting a few months ago. But after a few months of amateur exploration in AI art, he beat out other artists with years of training in painting and digital art.

Before I started writing this article, I used another more user-friendly AI painting tool named Wombo Art (Allen started his journey with this tool) to create the following artworks within 5 minutes.

What I did:

1. Opened the website https://www.wombo.art/
2. Clicked the Create button in the upper right corner (no need to register an account unless you want to save the work)
3. Entered a few keywords as the prompt (e.g., big white castle in tranquil forest, by Hayao Miyazaki, a hyper-detailed C4D render, surrealism, surreal concept art, complex, aesthetic, smooth, sharp focus, artstation hd, virtual engine — use them in any combination)
4. Selected an art style
5. Hit Create button — voila~ my work was ready in a few seconds!

To be honest, compared to other AI artists, “my” paintings are simply rudimentary, but I’m already very satisfied. After all, my entire experience with AI painting is only a few hours. Around February and March of this year, an AI painting tool called Disco Diffusion suddenly went viral on social media. At that time, I was preparing for an exam. I did a little bit of exploration with this tool and drew a few da Vinci style manuscripts, and the results weren’t so good. Then I forgot about it. But only after a few months, I suddenly found AI painting has improved by leaps and bounds, not only more tools have appeared, such as Google’s Imagen and Parti, OpenAI’s DALL-E 2, Midjourney via Discord, Meta’s Make-A-Scene, Microsoft’s NUWA, and other smaller and more user friendly tools like NightCafe, starryai, JasperAI, etc., and the capacity of AI painting has been greatly improved. At first, AI painting relied on textual prompts and could only generate images. Now, you can choose your own picture as a reference style or a based template for improvement, and you can even generate animations with one click. Hmmm, AI has learnt a lot more than me in a few months.

There are a lot of discussions about AI painting. Some artists cheer that AI can help them do their jobs more efficiently, but some others worry that AI will cause many junior artists, especially illustrators, to lose their jobs. Some even argue that AI should not be allowed into the arts, just as robots are not allowed in the Olympics. But I would like to discuss this topic a bit further from an intellectual property (IP) perspective. At present, our legal system has not considered IP disputes around AI art, but whether AI works are protected by IP law, and whether AI works themselves infringe other people’s IP rights, are all essential issues worth exploring.

First, let’s think about a question: Who is the author of AI painting? Is it someone using AI tools? But by just providing a few keywords, can I claim that I am the author of a typical Leonardo da Vinci sketch? Well, I’m embarrassed to say that because I don’t even know much about digital art or Leonardo da Vinci. This is why many artists are outraged and think that Allen was using AI to cheat in the art competition. They think that Allen did not paint that work at all, he never picked up a brush, and he should not be the author of that painting.

So if the user of the AI ​​tool is not the author, then who is? The programmer who wrote this AI painting program? Sounds a bit outrageous. Or the AI ​​itself? The idea is bold, but not impossible! In reality, there are already actions to help AI tools to obtain their legal rights. In 2019, a British research team submitted two patent applications with an AI algorithm as the inventor to the European patent office. The application got rejected by the European Patent Office, which concluded that “the inventor designated in the application has to be a human being, not a machine”. Another try was made in 2017 when the European Parliament urged the European Commission to propose an “electronic personality” for AI. The proposal was rejected for it contradicted the national law of EU member countries. Similar to Europe, in the US, common-law and policies and procedures of the U.S. Copyright Office and the U.S. Patent and Trademark Office reject the idea of non-human authorship or inventorship (Pearlman, 2017). Lawmakers around the world agree that only humans can be the author and copyright holder.

Many legal scholars worry that the personhood of AI could easily be abused for criminal purposes. Because AI, as an entity, cannot take any responsibility when it enjoys its legal rights. Some companies may use AI as their legal person for criminal activities like money laundering or tax fraud (Burri, 2018). In this way, the personhood of AI can create endless problems for the legal system. But many legal experts are also aware that IP laws were developed some 50 years ago, when there were no such human-brain-like tools like AI today, nor AI has ever been more widely used than it is today. So in this new context, we’d better not rush to rule out AI as the author, but re-examine the issue. But many legal experts also know that intellectual property law was enacted some 50 years ago, before there were brain-like tools like artificial intelligence or the widespread use of artificial intelligence as it is today. So in this new context, we’d better not rush to rule out AI as an author, but re-examine the issue to see if there is other possibilities.

Second, another issue we need to think about is the originality of AI art. In the US, to be eligible for copyright protection, a work must be “original” and “fixed in a tangible medium of expression; and moderately creative” (Packard, 2010). The work, using Vaidhyanathan’s words, should contain “markers of creative decision making by the author or artists”. A work must pass a threshold of originality to be copyrightable. But, you may ask, what is originality? Or, in a post-modernism view, is there any absolute originality?

Let me explain this issue with an example. The following painting is called The Next Rembrandt. Rembrandt didn’t paint it. The painting was created by Microsoft’s Deep Neural Network algorithm, which analyzed all 346 Rembrandt paintings to recognize the patterns and styles in Rembrandt’s works and paints this artwork based on Rembrandt’s use of geometry, composition and painting materials. Some people doubt the originality and creativity of such a work. After all, the techniques and styles are so unique to Rembrandt, how does such a work exhibit originality by putting together the Rembrandt elements? Even though AI-painting is not a simple salad-making process, but you may still argue the elements and styles belong to Rembrandt, rather than Microsoft.

This is the most controversial issue of AI painting at the moment: if you only write a few keywords and click a few buttons, and you automatically generate a Yayoi Kusama- or Jackson Pollock-style image, or a Wes Andersonian short film, can you confidently say that you are the Author?? Or, can you use them for commercial purposes? Currently, you can find the works of almost any well-known artist online, but that also means people developing the AI painting tools can collect them. The artworks are heavily used to train the painting algorithm so that the AI ​​can easily learn the style of each artist to generate new works in batches.

But everyone knows how important style is to an artist. It is the crystallization of their painstaking work, their icon, and contains all their passion and spirit. In a practical sense, an artist’s style is also his/her source of income. Just like Disney hates a cartoon duck to be the mascot of a university athletic team, why can the iconic styles of artists be freely used in AI paintings? According to the current copyright law in the United States, the copyright protection of artists expires 70 years after their death. Is it really okay and ethical to use contemporary artists’ works that are still under the protection period as the training data for AI painting algorithms?

Currently, if you are free users of the most popular AI painting tools like Disco Diffusion, Midjourney, DALL·E 2 and Stable Diffusion, your work are most likely to be under MIT License or Creative Commons (CC). Midjourney, for example, makes it clear that if you are not a Paid Member, it grants you a license to the Assets under the Creative Commons Noncommercial 4.0 Attribution International License. This license allows reusers to distribute, remix, adapt, and build upon the material (your AI painting) for noncommercial purposes. That is to say, your work enters the public domain if you are a free user of those AI painting tools. However, if you are a paid user, Midjourney says that “you may bypass some of these public sharing defaults.” Ok no problem, paid users can keep their AI paintings private without having others reuse or remix them. But how about the benefits and legal rights of the original artists whose artworks are used for training the AI algorithms?

Midjourney indicates that it “uses an artificial intelligence system trained on public datasets to produce its Assets. Such Assets may be unintentionally similar to copyright protected material or trademarks you hold…. If you believe your copyright or trademark is being infringed by the Service, please write to takedown@midjourney.com and we will process your request”.

Voila! That’s it. Artists also have to learn a bit about copyright law and take time to track down people who might infringe on their intellectual property. Artists have to report to Midjourney themselves, wait for a reply from customer service, and go through a long process of rights protection. . . Emmm. . . What a hell mode. . .

But some people are questioning, isn’t art progressing like this all the time? Aren’t all arts in history borrowed from each other? If we were to strictly protect every style, there would probably be only one figure in any art genre in history. Or, there won’t be any art genre at all. Impressionism is owned by Manet, and everyone else can only paint their paintings as required by the French Academy. Then in fact we don’t have Impressionism anymore, we only have Manetism. Who is Monet, Degas, Gauguin, or Van Gogh? just forget about them…

But wait, wait, wasn’t Manet also influenced by the use of flat and abstract spaces in the ukiyo-e style? He is not absolutely original! He learned from Spanish and Japanese art, and his works reflect those styles. If we strictly protect styles, then we don’t even have Manet at all.

Some people went a step further and said, remember how we treated digital art 10 years ago? Or, how the Salon (of the French Academy) treated Manet? What if Jason Allen was the Duchamp of our time? Okay, I know I’m going a little far, Allen and Duchamp seem to be on opposite ends of the art world, but they’re the same kind of pioneers that both make us think about the fundamental questions: what exactly is art? Who defines art? Who owns art? These grand questions cannot be discussed thoroughly in such a short article, and there will never be perfect answers.

Back to the discussion of intellectual property for AI painting. In fact, in addition to the question of who is the author and what is originality, we still have many considerations that need to be implemented in real life. For example, should AI paintings be bylined with AI? There is another important concept in intellectual property, fair use, how can we fairly use AI paintings? Also, should we create a separate arena for AI to avoid competing with human artists? Or do we encourage human artists to collaborate with AI to produce various new artistic styles (like post-human style or post-AI art)?

We all know that protecting the artworks and styles of existing artists is because we want to encourage people to continue to create, but if the protection is too strict, existing resources will be firmly held in the hands of a small number of people, and art creation will become rigid and loses its vitality. There is a lot of space between protection and innovation, and this space is undoubtedly dynamic and post-modern. Much of our exploration in this space is simply thinking about a fundamental question: “What does it mean to be human?” What makes us human compared to AI? We used to turn to religion and philosophy for answers to this question, but now, maybe it’s time to turn to science and art.

References

Packard, A. (2010). Digital media law. John Wiley & Sons.

Pearlman, R. (2017). Recognizing artificial intelligence (AI) as authors and investors under US intellectual property law. Rich. JL & Tech., 24, i.

Vaidhyanathan, S. (2017). Intellectual property: A very short introduction. Oxford University Press.

Figure of The Next Rembrandt: https://medium.com/@DutchDigital/the-next-rembrandt-bringing-the-old-master-back-to-life-35dfb1653597