Copyright belongs to content creators, not generative AI or AI company
Generative AI has revolutionized the way we interact with technology, from language processing to image recognition. One of the most impressive AI models today is ChatGPT, a large language model trained by OpenAI. However, with its immense power and capability comes the question of who owns the copyright of the generative work, especially when the AI model was trained using someone else’s work. This also happens on other generative AI models like DALL-E, an AI system that can create realistic images and art from a description in natural language. In this blog post, we will discuss the copyright issues behind generative AI’s creation and argue that the copyright of the generative work should belong to the content creators whose work have been used for training the generative AI based on the similarities between the generative work and the original work.
Training data is complicated, but we should still consider the copyright
Take ChatGPT as an example, ChatGPT’s generation process involves training the AI on a large dataset of pre-existing text. The training involves a large corpus of text from the internet, including books, articles, and websites composed by content creators. While the training data was not entirely consist of copyrighted material, there are still concerns regarding the use of copyrighted works in the model’s training data. In this case, it is crucial to determine who owns the copyright of the generative work. Like human content creators, the AI may generate new content, it is still essential to acknowledge that the starting point is based on the work of others.
AI model? AI company? Content Creators? Who owns the copyright?
In my opinion, the copyright of the generative work should belong to the content creators whose work have been used for training the generative AI based on the similarities between the generative work and the original work.
The work created by content creators is a result of their hard work, creativity, and talent. Therefore, they should be the ones who own the copyright to their work. According to the current copyright legislation, generative work is essentially a derived work of the training data, and the owner of the original work should have the right to control its use and distribution. Moreover, if the generative AI were to be used for commercial purposes, the owner of the original work should be entitled to a fair share of the profits.
This issue has become increasingly relevant as more content creators are finding their work being used as training data for generative AI without their permission or compensation. In an article “How Creators Are Squeezed by Big Tech”, content creators shared their experiences of having their work used to train AI models without their knowledge or consent. Some of them even found their work being used in ways that they did not agree with, such as in hate speech or propaganda.
The article also highlights how content creators are being squeezed by big tech companies, with their work being used without proper attribution or compensation. This is a major concern as it undermines the efforts of content creators and their ability to earn a livelihood from their work.
While some argue that the use of copyrighted works in generative AI is covered by the fair use doctrine, I believe that the potential harm caused by unauthorized use of someone else’s work in generative AI outweighs any potential benefits. To ensure that content creators are not squeezed by big tech, we need to establish clear guidelines for the use of copyrighted material in generative AI and provide fair compensation to the owners of the original works.
How to measure similarity between training artwork and generative artwork?
The way generative AI creates content is, to some extent, similar to the way human creators create content. Overall, everybody is stepping on the shoulder of our ancestors while exploring the world. From the childhood nighttime stories to the research papers on the way to a college degree, the content creators are immersed with original ideas and thoughts from others. If the human creators can argue for copyrights, why can’t generative AI, or the AI company, argue for copyrights? It turns out the boundary is all about the similarities between the original work and the generative work.
Admittedly, it is a really hard thing to accomplish: setting up a clear standard measuring the differences between generative work and the work created by its original content creator. It is even a hard thing to tell which work the generative AI is fed with. Legislation on this topic is still under development. Overtime, I believe that we’ll see more cases sitting between this boundary. Until then, I’ll still argue that the copyright belongs to the original content creators.
How do people in the art world think of generative AI
Every time I talked to my friend Emma, who studies architecture, we would inevitably end up discussing the intersection of art and technology. She expressed her desire to switch to a design field related to tech, like user experience design, rather than architecture. During our last conversation, she showed me an artwork-generative AI that was popular among her architecture friends.
Though this AI model was still in its beta phase, it had already demonstrated more power than our beloved DALL-E from OpenAI. As we were hanging out, this AI model reminded us of how advanced technology can threaten human beings, similar to the dystopian world portrayed in the TV series “Black Mirror.” Naturally, an idea popped into my head about how artwork generative AI could impersonate a real human:
“Once upon a time, in the digital age, an accomplished artist retired from the art world. He lived a contented life and appreciated the fact that his unique style of painting was well acknowledged. One day, his peaceful existence was shattered when he stumbled upon a couple of paintings that looked similar to his artistic style. He had a clear memory that those paintings were not coming out of his hands.
To his utter shock, the paintings were all over the internet, and one of them was even sold for $10,000! The retired artist couldn’t believe his eyes. He couldn’t figure out how someone had replicated his style so impeccably. He felt like he had been robbed of his identity.
As he dug deeper, he found out that a cunning individual had been using a generative AI to mimic his style and create artworks extensively. The cunning “artist” impersonated the retired artist and earned a lot of money, while the real artist was oblivious to the whole scenario.”
The “double-sidedness” of generative AI
It was ironic that while the imposter was living his dream, the real artist, who spent years honing his craft, was left with nothing. Even though this is only a story Emma and I made up in the air, it seems totally convincing that it originated from the real-world. How capable of generative AI is right now is far beyond our expectation a couple of years ago. If such an occurrence had already happened, it urges us to think about the double-sidedness of generative AI without any lag. Generative AI is a double-edged sword. While it could make art more accessible and create new opportunities for artists. Still, it could also pose a significant threat to the artistic community’s livelihoods, reduce the value of human creativity, and raise questions about the role of technology in shaping the future of art.
AI companies are artists too
With all of the discussion above, there’s another intriguing observation, namely that AI companies and algorithm developers can be viewed as a different type of artist — one whose artwork is the creation of state-of-the-art AI models. Thinking of it in this way can help us appreciate the years of effort that these developers have spent refining their models. In this way, these generative AI models are truly their masterpieces.
However, it’s important to remember that the AI models and the music or artworks are ultimately different types of creations. These generative AI models are just tools created by human beings. Just like any other tool, whether it’s good or bad depends on how people use it. Using generative AI models wisely can lead to better outcomes, while ignoring the unique qualities of human beings can have negative consequences.
The tool itself is just a tool, and it’s the human creator who adds their own creativity and perspective to the final product. If we see it through this way, it’s fair to say that any creation made using this tool should be considered an individual creation, and the creator of that work should own the copyright, rather than the artists whose work was used to train the generative AI model.
That being said, each individual’s work builds on the progress of those who came before. It is important to respect the contributions of artists whose works are used as training data. As we continue to advance technologically, it is crucial to consider the ethical implications of AI and its impact on society. The use of AI in art must be approached with caution and critical thinking to ensure that it benefits artists and the art world as a whole.
Reference
- “How Creators Are Squeezed by Big Tech” https://themarkup.org/newsletter/hello-world/how-creators-are-squeezed-by-big-tech
- “What is CHATGPT and why does it matter? here’s everything you need to know.” https://www.zdnet.com/article/what-is-chatgpt-and-why-does-it-matter-heres-everything-you-need-to-know/
- “Copyright Basics FAQ” https://fairuse.stanford.edu/overview/faqs/copyright-basics/
- Chatgpt — legal challenges, legal opportunities https://www.fieldfisher.com/en/insights/chatgpt-legal-challenges-legal-opportunities
- Is chat GPT free for commercial use? Chat GPT Pro. https://opchatgptai.com/is-chat-gpt-free-for-commercial-use/
- Who ultimately owns content generated by CHATGPT and other AI platforms? https://www.forbes.com/sites/joemckendrick/2022/12/21/who-ultimately-owns-content-generated-by-chatgpt-and-other-ai-platforms/?sh=7205359e5423
- OpenAI’s DALL-E 2: A dream tool and existential threat to visual artists. https://newatlas.com/computers/dall-e-2-ai-art/
- Algorithms Can Now Mimic Any Artist. Some Artists Hate It. https://www.wired.com/story/artists-rage-against-machines-that-mimic-their-work/
