With the rapid development of artificial intelligence all over the world, different organizations, mechanisms, and platforms use artificial intelligence in different areas. However, artificial intelligence technology not only brings significant development opportunities for economic development and social progress but also brings profound challenges to ethics and the social rule of law. The uses of artificial intelligence technologies raise several ethical problems, and artistic creation is one of them. How to treat the participation of artificial intelligence in artistic creation? Can artificial intelligence replace human beings for artistic creation? Due to the improvements in AI technologies, numerous organizations and companies have started attempts at artistic creation for their artificial intelligence. The use of artificial intelligence in culture and arts raises interesting ethical questions. This article will focus on the ethical dilemma of AI’s participation in artistic creation and what is the future of “AI arts.”
Artificial Intelligence’s Artistic Creation
The Next Rembrandt
In 2016, 351 years after his death, a painting in Rembrandt’s style, The Next Rembrandt, using an artificial intelligence design, was 3D printed. To bring the painting to life, the 3D printer recreated the texture of the brushstrokes and the painting overlay on the canvas to achieve a stunning effect that fooled the eyes of all art experts. To perform technical and artistic prowess, the project team built a unique database of 346 Rembrandt paintings for pixel analysis, upgraded with deep learning algorithms. In this way, every detail of Rembrandt’s artistic style can be captured, laying the algorithmic foundation for creating this unprecedented masterpiece. But who is the creator? The company, the engineer, the algorithm, or Rembrandt himself?
Furthermore, in late October 2018, Christie’s sold its first-ever artificial intelligence work called Edmond de Belamy to an anonymous bidder at auction for a staggering $432,000 (Hicks, 2019). The Obvious, a Paris-based AI and art collective, took 15,000 portraits from the 14th to 20th centuries and turned them into data fed into a machine learning program based on an open resource Network called Generative Adversarial Network (“GAN”). The result is the artificial intelligence portrait sold by Christie’s with abstract characteristics, Edmond de Belamy.
Edmond de Belamy
“AI stands on three key pillars: algorithms, hardware, and data. You collect large amounts of data, and then using the methods of machine learning, algorithms learn to find interdependencies among these pieces of data and then reproduce this logic on every new piece of data they meet.” (Megorskaya,2022) Same as how artificial intelligence worked for both paintings. Nevertheless, compared to The Next Rembrandt, Edmond de Belamy as a work sold at auction and earn benefits, this work which was generated by artificial intelligence, was met with scepticism. The art community doubts that it is not an original work but a derivative work; Some software developers questioned the Obvious’ unauthorized use of code written by others.
In 2019, the Chinese technology company Huawei announced that artificial intelligence algorithms had been able to complete the final two movements of Symphony No. 8, the final unfinished work Schubert began writing in 1822, 197 years ago. So, what happens when AI becomes capable of creating its own works of art? If human authors are replaced by machines and algorithms, how should copyright belong? Should algorithms be recognized as creators, with the same rights as artists?
Some answers to these questions seem to be open to interpretation by-laws from different countries, such as the copyright belonging problem. In China, Stipulation 2 of the Regulations on the Implementation of the Copyright Law stipulates that “works referred to in the Copyright Law refer to intellectual achievements that are original and can be reproduced in a tangible form in the fields of literature, art, and science.” Therefore, a work entitled to copyright must be the “original” intellectual achievement of “human beings”. In other words, if the output mainly relies on present procedures and instructions, and does not reflect the judgment, choice, or other creative intellectual activities of human beings (especially users of AI programs), it will be difficult to get protection under copyright law. Similarly, U.S. copyright law is designed to protect the “original intellectual output of the author,” and the U.S. Copyright Office explicitly refuses to register content that is not created by humans or is generated automatically by machines without human creative input or intervention.
Copyright problem is just one of the AI ethical problems which can luckily be defined by laws. Nevertheless, there are several problems that still cannot be solved right now. Artificial Intelligence needs original data to do an algorithm, but we cannot notarize what is their original data. For instance, number of social media platforms published products about AI painting and AI story-producing for users. All the companies always claim that their original database is from the public domain, although most writers and painters do not trust them.
Laws and stipulations can somehow limit those objective questions for artificial intelligence work, then how about the subjective problems? Compared with science and engineering which are based on formulas mostly, artistic creation itself is a very emotional and subjective field. Artistic work not only gathers patterns of knowledge and principles that are already known but also shows the personal emotions and feelings of the creators. Nonetheless, most of the artistic works are published to earn benefits from the public by companies and capitalists. In order to reach the purpose of getting money, those works cater to most people’s aesthetics and values. Therefore, artificial intelligence was used here for calculating and even creating part of the works so that the companies will easily attain “success”. Unfortunately, the aesthetics and values of groups can be guided and changed through public opinion and education, especially now when cultural communications and exchanges are based on the global Internet.
Case Study of NetEase LOFTER
On March 6th, 2023, the AI painting feature sparked controversy at LOFTER, where the official response was that the AI painting product is the avatar generator. The model training is from open resource data rather than the platform’s user work, and it should not be used for commercial purposes. LOFTER is a content community platform owned by NetEase, which gathers tens of millions of original writers and painters. However, the response from LOFTER did not extinguish users and creators’ anger. Instead, a number of users, including RC – the author of the Chinese popular comic book Dali Temple Log, issued statements saying they would stop streaming, cancel their accounts or change platforms. On March 7th, the incident continued to simmer and became a hot topic on Weibo. In the middle of the day, LOFTER officially responded again, trying to make a compromise sound to users. According to the statement, by adjusting the entrance of the AI painting function to the “centre of the profile picture box,” AI-generated pictures can only be used as platform profile pictures and cannot be downloaded or published. Nevertheless, at that time, LOFTER still did not cancel the AI painting function. On March 16th, NetEase LOFTER published “an apology letter to the vast majority of creators” on its official Weibo account, apologizing to creators for the AI painting function that was previously launched. The controversial feature was taken offline on March 8th, and LOFTER has promised to never use users’ work data for AI training.
To be honest, on other platforms, such AI tools might not be as controversial. In fact, some other APPs, such as Douyin (Chinese TikTok) and Meitu (Mobile Photoshop App), have already published tools for AI generating and getting positive feedback from users. Without a doubt, the technology is neutral, but the platform is specific. Meitu as mobile Photoshop is a tool that attributes strong software. Similar to Douyin, most of their users are ordinary users who are lack of creativity ability. However, LOFTER is positioned as an original community where creators share content. LOFTER is a UGC content community, to be more precise, this is a platform based on doujin and fan-fiction culture. Compared with commercial platforms, the motivation of LOFTER’s creators is often just to please themselves or others who have the same interests. Most Creators in LOFTER do not earn money from their works. In short, the creation of creators in NetEase LOFTER is further away from commercialization and purer. If an AI painting can easily be exposed and featured on a user’s home page, it will hurt the interests of the original artist. More worrying for creators is the possibility that LOFTER will use original works on the platform to analyze data and train AI models. This means that AI-generated works will bear the imprint of the creator, who may not know it.
(screenshot of LOFTER)
LOFTER is not the first platform where users have boycotted AI-generating. In 2022, the same thing happened with a comprehensive CG visual arts platform, ArtStation. A picture titled “NOT TO AI GENERATED IMAGES” was uploaded by CG artists, which was the same picture used in the LOFTER incident. One of the points made by this group of creators is AI plagiarism, and they contend that AI art is theft. Because image generation works similarly, so-called AI paintings may actually be piecing together pieces from different existing paintings. The research paper “Extracting Training Data from Diffusion Models”, published by Google, DeepMind, and other research institutions, has proved that AI models indeed remember some original images for training, and artificial intelligence can even recreate the entire original image in some cases.
However, after these two events, there are still many people who hold different views and post their opinion on a different websites. They claim that artificial intelligence is the inventible trend of future development, and so is the participation of artificial intelligence in artistic creation. And they argue that most of the work produced by human users is not as good as AI, because AI is based on rigorous algorithms that derive optimal solutions from huge databases. In the future, although AI users will gradually eliminate the traditional normal human users who only tap keyboards by hand, they will not completely eliminate the high-quality, cross-professional, and high-consumption elites who are willing to recharge a large number of platforms.
What is next?
Understood as the ability to create new and original content through imagination or invention, creativity plays a central role in open, inclusive, and diverse societies. To this end, the impact of artificial intelligence on human creativity should be taken seriously and thought about.
Some say that the essence of AI art is visual output, while others argue that it is literal input. Some people will say that even if it is text input, it should first form an artistic picture in the mind before the text output can be converted into the AI text input. However, the real essence is the commonality of words and art — imagination. Midjourney founder David Holz once said that AI art is “an engine of imagination”. Without a doubt, artificial intelligence provides ordinary users without creative ability with the possibility to complete their imagination.
While AI is a powerful creative tool, it raises some important questions about the future of art, the rights and remuneration of artists, and the integrity of the creative value chain. Many artists are very opposed to having their work taken to artificial training intelligence and also refuse to have their work in the source database. It’s popular to talk about what jobs can be easily replaced by AI, and the same is true in the arts. Rather than intuitively feeling that technology is inferior to artificial intelligence if artificial intelligence replaces artists by grabbing their works, it is understandable that artists will oppose AI.
In general, new frameworks need to be developed to separate piracy from originality and creativity. And recognize the value of human creative work in their interactions with AI. We need these frameworks to avoid the deliberate exploitation of human work and creativity and to ensure that artists are properly remunerated and recognized, the integrity of cultural value chains, and the ability of the cultural sector to provide decent work. “Civilization is meaningless without human beings” (The Wandering Earth, 2023), art is not the accumulation of skills, but more need to experience the story and a strong desire to output expression in order to have a real move. There is no doubt that AI will surpass human beings in all kinds of skills by imitation in the foreseeable future, but in terms of artistic attainments, there will be wonderful works, not amazing works.
Hicks, O. (2019). ART-ificial Intelligence: The Curious Case of Edmond De Belamy. The Isis Magzine. https://isismagazine.org.uk/2019/03/art-ificial-intelligence-the-curious-case-of-edmond-de-belamy/
Next Rembrandt. (2016). https://www.nextrembrandt.com
Landau, S. S. (2022). Copyright Office Review Board Reaffirms that Human Authorship is a Prerequisite for Copyright Protection. Cowman Liebowitz Latman. https://www.cll.com/CopyrightDevelopmentsBlog/copyright-office-review-board-reaffirms-that-human-authorship-is
Carlini, N. et al. (2023). Extracting Training Data from Diffusion Models. https://doi.org/10.48550/arXiv.2301.13188
Vincent, J. (2022). ‘An engine for the imagination’: the rise of AI image generators — An interview with Midjourney founder David Holz. https://www.theverge.com/2022/8/2/23287173/ai-image-generation-art-midjourney-multiverse-interview-david-holz
Megorskaya, O. (2022). ‘Training Data: The Overlooked Problem Of Modern AI’.
Copyright Law of the People’s Republic of China. (2021)
Guo, F. (2023). The Wandering Earth II [Movie].