Introduction
The advent of the big data era and the development of computer technology have led to artificial intelligence no longer being confined to traditional technical fields. It has begun to be applied in the arts, including literature, fine arts, and music, creating many breathtaking works of art, and entering a new phase of development. With the widespread adoption of technology and the open sourcing of algorithms, ordinary people can also use open AI platforms for creation, such as Stable Diffusion, Midjourney, mimic, and Novel AI. Why is AI drawing so favored? Its allure lies in the fact that there is virtually no barrier to entry. As long as users can accurately describe their needs in words, it can use its “intelligence” to create images. From abstract logo designs to exquisite movie and game visuals, to realistic photographic works, it can easily handle them all. Whether it’s the imaginative art style or the immersive realism, it can produce the artwork you desire. Furthermore, AI technology is constantly innovating at a rapid pace, enabling it to respond more quickly to people’s needs. For instance, while human learning and growth take a long time to accumulate, the release of Midjourney’s two flagship models was separated by less than five months.
However, the development of technology always comes hand in hand with risks. As artificial intelligence in the creative process and generative AI, representing the most advanced scientific and technological achievements, are applied across various fields, issues related to copyright risks associated with it are increasingly attracting social attention. The fundamental concept behind AI drawing involves using computer algorithms to decode, learn from, and train on a vast array of sample artworks data. Like the creative journey of humans, AI also needs to learn and accumulate enough material before creating. Once a large volume of materials is uploaded to the database, AI preliminarily categorizes and organizes these materials. When it receives a creative command from the user, AI matches the keywords in the command with the materials in the database. By analyzing works of the same theme, it extracts elements, structures, styles, etc., from these works. The neural network algorithm, through repeated creation and optimization processes, eventually deduces the correct creative patterns and generates related algorithm models. Users only need to input relevant keywords, and AI can output a large volume of similar works. This process requires systematic learning from human creativity and subsequent imitation and prediction based on acquired knowledge, involving the learning and use of others’ works, often occurring without the original creators’ knowledge or consent. This phenomenon gives rise to artistic works that bear resemblance to, or even exceed, the creators’ artistic prowess, thereby giving rise to significant apprehensions regarding copyright violation. At the same time, the absence of relevant legislation and the lack of academic research lay hidden risks for privacy and digital rights in AI’s further development, prompting people to consider whether AI art is about creating art or appropriating it. If the balance between copyright owners’ interests and AI development cannot be maintained, it will ultimately hinder the development of culture and the arts and technological progress. It has become a pressing global issue to address how AI creation can guarantee the legitimacy of its copyrights.
The Copyright Infringement Risks of AI Drawing
Artificial intelligence-powered text-to-image art generators have ignited a contentious discussion on the function of professional visual artists in society. A new breed of amateur artists, led by Reid Hoffman, the creator of the job platform LinkedIn, are utilising AI thanks to this technology. He has created and sold a number of AI-generated artworks on the NFT marketplace Magic Eden, one of which sold for the equivalent of $24,000, by utilising the DALL-E platform and the commercial rights policy of its inventor, OpenAI (Benzine, 2022).
Renowned Poland digital artist Greg Rutkowski is known for his illustrations for video games, including: Sony’s Horizon Forbidden West, Ubisoft’s Anno, Dungeons & Dragons, and Magic: The Gathering. And he’s become a sudden hit in the new world of text-to-image AI generation (Heikkilä, 2022).
His distinct style has become so well-liked that many fans have employed AI to mimic his works, making his art one of the most often used prompts in Stability AI’s recently released open-source AI art generator, Stable Diffusion. Along with other well-known AI models for creating images, this program is accessible as open-source on GitHub and lets users direct the creation process by entering words, phrases, or sentences. Users can customize the generated text to their desired length, style, tone, and format, resulting in visually striking results. However, these open-source tools are often built using images randomly scraped from the websites without the creator’s approval, leading to tricky ethical and copyright issues.
For instance, entering ” A wizard battles a vicious dragon with a sword and a shining magic sphere. Greg Rutkowski style,” the program will generate works that look extremely similar to Rutkowski’s painting style. Initially, Rutkowski was surprised but considered it an opportunity to market his work and find new fans. Then he attempted to look for his name online to check if any of his works had been published, only to find that they were not attributed to him. According to a report by MIT Technology Review, on two well-known open-access AI image generators, Midjourney and Stable Diffusion, where users submitted Rutkowski’s name 93,000 times, it is one of the most often occurring prompts. This is significantly more than the requests for pictures in the vein of Picasso or Michelangelo, whose names were prompted no more than 2,000 times apiece.
In an interview with the BBC, Greg Rutkowski said: ” In the first month of learning about it, I became aware of how it would definitely impact my career and make it impossible for me to identify and locate my own works on the internet. My name will be linked to the outcomes, but not my reputation. I won’t be the one to make it. For those who are discovering my works, it will therefore cause more confusion. ” “All that we’ve been working on for so many years has been taken from us so easily with AI. It’s really hard to tell whether this will change the whole industry to the point where human artists will be obsolete. I think my work and future are under a huge question mark.” he added (Hutchinson & John, 2023).
In order to avoid potential copyright entanglements, Greg Rutkowski and a number of other artists formally requested that their artworks be removed from the training datasets. This was done to prevent the AI from producing works that closely reflect their individual artistic styles. However, to their disappointment, their suggestion was ignored by Stability AI’s LoRA open-source model. The model claimed to be fully capable of creating artworks distinctly different from Greg Rutkowski’s style.
Rutkowski contacted the group in charge of creating the LoRA model after observing this as a blatant act. At first, they said they would take down the model at Rutkowski’s request. They also seemed to be warning Rutkowski against it, saying something along the lines of “artists cannot hinder the progress of societal trends.” This comment implies that even if they eliminated the pertinent training data in response to Greg Rutkowski’s criticism, a large number of models would still be able to learn and imitate his artistic style because open-source models are available. (Lanz, 2023)
Beyond Greg Rutkowski, many artists share similar experiences, surprised by the popularity of their works in the realm of AI-generated art. Three artists, Sarah Andersen, Kelly McKernan, and Karla Ortiz found their works in the Stable Diffusion dataset and collectively filed a lawsuit against the AI company Midjourney, Stability AI, and the artist platform DeviantArt (Foster, 2023). When people start using AI-generated graphics that are based on copyrighted material for commercial purposes, they are worried that they will lose money.
Illustrator Karla Ortiz, who resides in San Francisco, is committed to bringing attention to copyright concerns and AI art. Ortiz stated, “There’s a growing coalition within the artist industry to figure out how to address or mitigate this situation.” The early phases of this coalition’s mobilization may involve advocating for new laws or rules.
“Open-source AI is a significant creativity, we acknowledge the pending issues and diverse legal opinions. As AI becomes more ubiquitous, we anticipate that the differing views on how to balance individual rights with fundamental AI/machine learning research will be resolved over time,” said Mason from Stability.AI. “Our goal is to strike a balance between community support and innovation.””
Existing countermeasures for infringement
The artists Holly Herndon and Mat Dryhurst are creating tools to assist artists in choosing not to use training datasets. They introduced the Have I Been Trained webpage, allowing artists to search among 5.8 billion pictures in the datasets used to train Stable Diffusion and Midjourney to see if their works are included (Wiener, 2023). A few online art communities have already made a stand and specifically forbidden AI-generated images, like Newgrounds.
Adobe, Nikon, and the New York Times are among the companies involved in the Content Authenticity Initiative, an industry effort that is creating an open standard to create a watermark-like digital content authentication system. It might aid in the battle against misinformation and guarantee that those who create digital content receive due credit. (Heikkilä, 2022).
Beyond copyright concerns, the increasing realism of AI-generated content has raised potential worries about the illegal use of these images and videos. In one previous instance, following an announcement by former U.S. President Trump about his impending arrest, numerous AI-generated images depicting “Trump being arrested” widely circulated on social media, misleading some internet users. Furthermore, worries about the future of deepfakes were raised by a recent story on India’s News18 news website about a scenario in which someone utilised artificial intelligence to produce an image of former German Chancellor Merkel and U.S. President Obama sharing ice cream at the beach. According to the website, “This technology is growing more and more frightening as it develops. The current state of AI’s ability to produce almost flawless celebrity voices and images threatens to undermine our confidence in our own senses.” Coincidentally, some parents also expressed their concerns to Global Times reporters about the adverse effects of violent and pornographic images generated by AI on children.
Conclusion
With the rapid advancement of artificial intelligence, the barrier to artistic creation has been significantly lowered, enabling ordinary individuals to use AI to produce the artwork they desire, thus providing infinite inspiration and opportunities for human creativity. However, since machine learning requires harvesting a vast number of existing artworks from the internet for training, users utilizing AI for creation, and considering the technology is still in the nascent phases, and there are still gaps in the applicable legal frameworks, rules, standards, and ethical principles. This might make copyright violations worse, raising concerns about data security, copyright violations, the propagation of deepfake content, and the distribution of unlawful information, among other dangers.
Because humans put their feelings and ideas into their works, human creations have an innate vitality. The distinct personalities of artists imprint their works with permanence and their original expressions drive the progress of art. Therefore, the development of humans and artificial intelligence should complement each other. Faced with the dilemmas brought about by AI creation, I believe society can address these challenges by improving the collective management system of copyright, refining the mechanism for work fees, and incorporating blockchain and AI recognition technologies for regulation. This ensures that while protecting the interests of copyright owners, it also considers the development and innovation of technology, achieving progress for the entire society.
Reference
Benzine, V. (2022, September 20). ‘A.i. should exclude living artists from its database,’ says one painter whose works were used to fuel image generators. Artnet News. https://news.artnet.com/art-world/a-i-should-exclude-living-artists-from-its-database-says-one-painter-whose-works-were-used-to-fuel-image-generators-2178352
Heikkilä, M. (2022, September 16). This artist is dominating AI-generated art. And he’s not happy about it. Technology Review. https://www.technologyreview.com/2022/09/16/1059598/this-artist-is-dominating-ai-generated-art-and-hes-not-happy-about-it/
Hutchinson, C., & John, P. (2023, July 19). AI: Digital artist’s work copied more times than Picasso. BBC. https://www.bbc.com/news/uk-wales-66099850
Lanz, J. A. (2023, July 29). Greg Rutkowski was removed from Stable Diffusion, but AI artists brought him back. Decrypt. https://decrypt.co/150575/greg-rutkowski-removed-from-stable-diffusion-but-brought-back-by-ai-artists
Foster, W. (2023, December 8). Stable Diffusion copyright lawsuit update. Medium. https://medium.com/@foster.whitney.a/copyright-lawsuit-against-midjourney-stable-diffusion-and-other-ai-art-generators-gains-momentum-8a19bbc51c7b
Wiener, A. (2023, November 13). Holly Herndon’s Infinite Art. New Yorker (New York, N.Y.: 1925). https://www.newyorker.com/magazine/2023/11/20/holly-herndons-infinite-art
Be the first to comment