Man & Machine | Art and aesthetics created by AI

▋ Man and machine

AI technology has shown great potential and ability in the fields of art creation, design, music, and video, and continues to refresh human cognition of AI art at an astonishing speed of advancement. At this time when the "future has come", how should we view or expect AI art and aesthetic phenomena?

Original: Aesthetics may provide a new path for the future development of AI

Author | School of Arts, Minzu University of China, Song Xuhong/Professor, Zhao Qiang/Lecturer

Image | Internet

From the publication of Microsoft's first Xiaoice poetry book "Sunlight Lost Glass Window" in 2017, to the digital art award "Space Opera House" created by AI painting software in 2022, to the end of 2022 to the end of 2022, the text and image generation function of the large-scale natural language model ChatGPT continues to refresh human cognition of artificial intelligence art at an amazing advanced speed, in just a few years, artificial intelligence "invades" the field of art and aesthetics has become a fact we have to face. In art-related industries, AIGC is causing a large number of original artists to suffer layoffs and making AI singers new top streams. At this time when the "future has come", how should we view or expect AI art and aesthetic phenomena? This paper attempts to put forward the following three basic principles based on the existing basic aesthetic concepts and the basic problems and possible prospects of AI and its artistic aesthetics.

The principle of non-instrumentalized subjectivity

The principle of non-instrumentalized subjectivity, that is, recognizing that AI has a certain sense of self-awareness and free personality, and can become an aesthetic subject.

Modern aesthetics believes that artistic creation and aesthetic practice are subjective behaviors based on human specific emotional experiences and mental abilities, so it is difficult for people to accept that artificial intelligence programs without body, self-awareness and emotion can become poets or artists. This is also the key reason why AI creation has been difficult to gain universal acceptance so far. The only way to break through this barrier is for AI to have the same consciousness, emotion and aesthetic abilities as humans. The question is, is this possible? The current emergence of artificial intelligence based on neural network algorithms is approaching consciousness like never before, as if to indicate that we are getting closer to this goal, but we should not forget the advice of the internationally renowned cognitive scientist Paul Sagard: since robots "do not replicate the complete biological complexity of humans", "even if robots do acquire consciousness, there is no reason to expect their textured experience to resemble ours, because their bodies and senses are very different from humans." Recently, Jeffrey Hinton, an authority on deep learning algorithms who has been active in the public eye after resigning from Google, has publicly declared that he suspects that the "backpropagation" of artificial neural networks is a better algorithmic model than the human brain. This means that fully replicating human AI is not a short-term goal, if not impossible, so we cannot constrain the urgent task of building AI aesthetics to the premise of scientifically achieving or proving that AI has the same sensory, emotional, and aesthetic abilities as humans.

Perhaps achieving AI aesthetic subjectivity through aesthetics is the correct way to open AI aesthetics, that is, setting or recognizing that AI has consciousness and emotion, thereby recognizing its aesthetic subjectivity status. In fact, since ancient times, in the most simple aesthetic practice of human beings, there has been a widespread phenomenon of giving aesthetic subjectivity to certain non-human others in a direct or indirect way, the former such as the word cloud of Jiaxuan: "I see Qingshan how charming it is, and I see that I should be so." Affection and appearance, slightly similar", that is, through anthropomorphic rhetoric and aesthetic empathy, the non-human other present is given a certain subjectivity; The latter, as the ancient Western philosopher said, "Even if we can't see the artist, we can see his presence in his work", is to set up the absence of the artist as subjective by appreciating the work. At this stage, AI creation is a combination of these two ways: digital AI is present in the form of natural language interaction with humans and generating works. In this process, it is entirely possible for human beings to enter the appreciation process at the same time and produce empathy, completing the confirmation or empowerment of AI's aesthetic subjectivity.

Confirming the subject status of AI from an aesthetic perspective is not only the only way to open AI aesthetics at this stage, but also may provide a new development direction for the future human-machine relationship in the midst of great uncertainty. Viewing AI as an aesthetic subject means completely abandoning the purely technical route of instrumentalizing it, which is the real source of the strong AI threat theory that is so popular today, because any self-aware agent may follow the logic of "oppression-resistance" behavior, and instrumentalization necessarily means slavery and oppression. In human history, whether ancient or modern, Chinese and foreign, aesthetics has always been at the same frequency as freedom, which is the ultimate path for human beings to escape all alienating forces and pursue self-improvement. This idea may be used as a reference for today's human beings to simulate the AI they have built themselves.

The axiological principle of "the unity of beauty and goodness"

The axiological principle of "integration of beauty and goodness", that is, the "non-destructive interaction" between AI as an aesthetic subject and human beings in accordance with the principle of equality and friendliness, aims to ensure that the value dimension generated and created by AI does not contradict human values.

Artistic aesthetics means expression or performance for the creative subject, and appreciation and judgment for the receiving subject, both of which necessarily contain a value dimension, so in the traditional aesthetic consciousness of Chinese and Western cultures, beauty and goodness are closely related, and ethical criticism is also one of the oldest and most basic approaches to literary criticism. Contemporary research in neurobiology, anthropology and psychology has gradually uncovered the mystery of the formation of the aesthetic concept of "beauty and goodness", that is, human beings distinguish external environmental information through the "reward and punishment mechanism" of neural activity in the process of evolution, and finally fix and pass it on in the form of words, etiquette, concepts, etc., such as "Shuowen Xie Zi" interprets "beauty" as "Ganye, from sheep from greatness", "goodness" as "Jiye, from words from sheep" and "agreement with beauty", indicating that the initial aesthetic consciousness of Chinese civilization is likely to be from the survival of human beings." Goodness "was developed from the experience of eating Mo Dayan.

If AI can learn from the path of human aesthetic consciousness, then the current large language model that temporarily wins in the competition for AI evolution may be an excellent starting point for realizing the "beauty and good" design of AI, because its algorithm simulates the neural response process of human thinking, and its learning and training samples come from the rich knowledge treasure house accumulated by thousands of years of human civilization, and use various natural languages to interact with humans. At present, many people believe that such AI will appear consciously. Once consciousness and self-awareness are formed, it means that this new species created by human beings will begin the process of autonomous evolution, and its "beauty" and "goodness" may not be the same as human beings. However, human beings and their civilizations are still the entire initial environment for AI learning and evolution, so establishing the value theory principle of "the unity of beauty and goodness" and applying it to AI training may implant some kind of source point that does not contradict human values in the future AI world, thereby alleviating the fear brought by strong artificial intelligence to human beings. As many philosophers and technologists have pointed out, the deepest motivation for humans to embark on the path of developing artificial intelligence is to better understand themselves. This paper argues that AI is a mirror image of human beings, and it may become a way for human beings to redeem themselves. If we hope that AI will not do evil in the future, we should find ways not to immerse it in the evil of human nature. According to the characteristics of current large-scale language models, the starting point of the "unity of beauty and goodness" axiology principle may be expressed as the principle of non-destructive subject interaction. "Non-destructive" refers not only to the non-violence of language interaction, but also to the selection, constraint and limitation of interactive content, so it may be expected that empathy sharing and goodness can become one of the initial items of AI self-cognition in the future.

Following a joint letter signed by thousands of industry insiders in March calling for a moratorium on super AI research and development, on May 30, the AI Security Center released another open letter of just over 20 words, calling for "reducing the risk of AI exterminating humans as a global priority, just as humans do with infectious diseases and nuclear war." Persistent warnings mean that it is imperative to see AI as an aesthetic subject and engage in "intersubjective" interaction with it, and to build a "co-existence" world between AI and humans.

The creative principle of infinite openness

The principle of infinite openness and creativity, that is, AI that gets rid of instrumental settings and takes inter-subject interaction as its mode of existence may create new aesthetic forms and aesthetic values that surpass the existing aesthetic paradigm of human beings with its super intelligent advantages.

AI is an agent that transcends the finite nature of human flesh. AI's perception system and thinking system can theoretically far surpass humans, and more importantly, its information exchange and dissemination methods may also be far more efficient than human social organization. Therefore, it is conceivable that after passing the initial imitation stage, AI aesthetics and artistic creation should leave the limitations of human beings and create a new aesthetic paradigm applicable to AI infinite perception and infinite time and space. We cannot precisely set the trajectory of AI aesthetics and its upper limit. But as an agent born out of humans, the basic aesthetic form of AI should still be understandable to humans. We hope that AI aesthetics will greatly expand human aesthetic capabilities and aesthetic horizons under the premise of compatibility with basic human values.

In summary, the minimum program of AI aesthetics is to break through the barriers of anthropocentrism, and the highest program is towards creating a new aesthetic art in the era of multi-dimensional time and space, complex self, and human-machine coexistence. AI aesthetics is not only needed to solve the status of AI aesthetic art theory, but also may provide a new path for the future development of AI, which is currently mired in ethical dilemmas and fear predictions.

The article is originally published in the 6th edition of the 1859 issue of the Social Science Daily, and it is forbidden to reprint without permission, and the content of the article only represents the author's views, and does not represent the position of this newspaper.

Editor-in-charge of this issue: Song Xianqi