Ai Arts

The Future of Chinese Painting - The Convergence of Art and Artificial Intelligence

Artificial Intelligence in art, like the Daozi AI painting system from Tsinghua University’s Future Lab, can now learn the style and content of Chinese paintings very well using Generative Adversarial Networks (GANs). It also continuously optimizes its understanding of the unique blank spaces, ink and brush work, and lines in Chinese painting. As datasets expand and computing power improves in the future, AI generated paintings will increasingly close in on “form likeness”. Eventually, they will be able to “paint” Chinese paintings that are indistinguishable from, or essentially are, real ones. However, the “spirit likeness” that Chinese painting seeks, emphasizing human subjective feelings and understanding, is still beyond the reach of current AI technology.

The starting point of Chinese painting, as revealed in Cao Zhi’s “Ode to the Goddess of the Luo River”, has moved from the utilitarian nature of primitive mural and pottery painting towards the artist’s self-consciousness of beauty. Chinese painting has evolved from a tool and symbol to express human emotions and personality. The exquisite techniques and lines of Chinese painting are infused with the soul of beautiful work. From my own experience with images generated by current AI technology, I believe they lack this beauty consciousness and an inherent “soul”. Can this gap be filled with inspiration from the history of Chinese painting?

In the era of AI, when we re-examine painting from the perspective of a technology like AI, I believe that ancient Chinese painting can be divided into three levels: imitative painting, expressive painting, and creative painting.

Analyzing the Three Levels of Painting

Imitative Painting

The first level is imitative painting, which is the lowest level. This involves improving one’s painting skills and creating works by continuously observing nature and copying others’ masterpieces. Just like AI, it can produce very natural and renowned works of art, such as the mimicking of flowers and birds in flower-and-bird paintings, or the imitation of characteristics of mountains, stones, and trees in landscape paintings. Combining modern photographic technology with style transfer algorithms, AI can achieve realistic imitation and even surpass humans. Imitation of others’ works - their lines, techniques, styles, and content - can also be replaced to a certain extent by AI technology.

Expressive Painting

The second level is expressive painting, where painting is a medium for expressing one’s will and emotions. This medium can be easily replaced by AI technology. Current AI technology can learn the style of others’ paintings and generate corresponding paintings based on definite human instructions. This process does not require the author to know how to paint, and the final result can entirely depend on the author’s expressive needs. For example, compound painters in literati painting (literati painters, amateur painters) express their feelings by integrating “poetry, calligraphy, painting, and seal carving”. Calligraphy and painting are mutually integrated and referenced. As Zhao Mengfu said, “Stone is like flying white, wood is like Zhou, and bamboo should be written with eight methods. If there is someone who can understand this, it must be known that calligraphy and painting are the same.” This connection between calligraphy and painting is difficult for humans to capture, but AI’s machine learning can extract the common features between the two through training and calculation of a large amount of calligraphy and painting data. AI algorithms enable the objective parts of the “poem-reciting-calligraphy-painting-seal carving” process in literati painting to be realized to a high degree by AI technology.

Creative Painting

The third level is creative painting, where new styles, techniques, concepts, patterns, imaginations, systems, artistic views, cultures, and aesthetics are created. This creativity is where AI is currently most lacking, and it is also where humans have the most creativity, abstraction, and subjectivity compared to machines. Acceptance of these creations essentially requires aesthetic ability under a certain cultural context, and the discourse power of aesthetics is still in the hands of humans. Perhaps AI does not have its own ideas, but the humans who write its programs do. Therefore, future Chinese painters might program their own AI tools to assist their painting, or they could first generate some paintings according to their own ideas and then select and detail them according to their aesthetics.

Looking at the Future Development of Chinese Painting in the AI Era

From these three levels, I believe that co-creation by humans and machines will become one of the mainstream methods in the future. Humans will be responsible for the expressive and creative parts, and AI will be responsible for the imitative parts. Thismethod combines the advantages of humans and AI, improving efficiency, effect, expressiveness, and creativity of painting. However, compared to traditional painting methods, this approach has several issues we must pay attention to:

1. Loss of the traditional painting process: The process of painting itself is a critical part of the art, where the artist projects their inner world onto the real painting through the brush and hand. This process is a black box in AI, preventing artists from precisely mastering the brush and ink, lines, and color usage in traditional Chinese painting. They can only hope that the black box will generate the desired effect according to the specific instructions given.
2. Whether the AI program is written by the artist: Although we have entered the information age, many people in the art field still have a limited understanding of programming. If they use programs developed by others (who may not understand painting) for painting creation, the program may not fully execute what the artists want to express.
3. Lack of creativity: Current machine learning algorithms are built on the basis of feeding large amounts of data, and the final generated paintings will have similarities to the images in the fed data, because the rules it learns are entirely from the fed data. Can we construct a creative AI algorithm?

Given these, I would like to propose a few reflections and suggestions for the future development of AI technology in the field of Chinese painting:

1. Embed the general principles of painting summarized by humans into AI algorithms. As Xie He proposed in the Six Principles of Chinese Painting during the Southern Qi Dynasty, “There are six principles of painting: one is spirit resonance, life movement; two is bone method, using the brush; three is according to the object, form; four is according to the type, apply color; five is planning and positioning; six is transmission, copying.” This method is historically proven to be comprehensive and universal. Future AI-generated Chinese paintings can expand these six principles more universally and include them as the basis for judging the quality of generated works. These evaluation methods can be quantified in GANs and other algorithms. Spirit resonance, life movement is the most abstract and difficult one. The other five are more specific, and I believe that with enough training of rich datasets annotated with human aesthetics, AI can “master” the subtleties hidden in these six principles.
2. Establish a programming environment suitable for artists to use AI algorithms. Although there are some creative programming software like PROCESSING that have significantly lowered the threshold for artists to program, a large proportion of artists still lack programming skills. AI technology is a challenging task for those without a foundation in mathematics and programming. How can the potential of AI technology be freely explored and used by more artists? I think we can construct a good open-source tool combining AI and art, educate artists about AI technology, and establish a virtuous cycle of ecology. The real creative power of art will be well stimulated in this ecosystem of artist clusters.
3. Continue to explore creative AI algorithms. In my opinion, what current machine learning algorithms do is still mimic data features and have no creativity to speak of. If this trend continues, AI will always be “artificially intelligent”, and AI art will always need to rely on humans. There is no possibility of a qualitative change due to quantitative growth, and the growth of beauty cannot be achieved by imitation alone. As David Deutsch said in “The Beginning of Infinity”, “The process of running a program does not create any knowledge, only the process of the programmer developing it creates knowledge.” The intelligence in the so-called AI comes entirely from humans, and humans still do not know how creativity works.

Conclusion

Although current AI technology lacks creativity, it is still an excellent medium for painting creation. The use of AI technology can allow human artists’ expressiveness and creativity to be more fully developed and reflected in the final painting. The combination of AI and painting has a long way to go. I believe that there will be an era in the future history of Chinese painting where a group of people are exploring the application of AI technology in Chinese painting.