Zhu Tianhua Shanghai Academy of Social Sciences, Institute of Literature
Published in “Art Studies” 2023 Issue 3
Editor’s Note:
As artificial intelligence technology has made significant breakthroughs, AIGC and its related issues have gradually become the core focus in the integration of art and technology, sparking extensive debates. This issue features three articles: “The Logical Dilemma of VR Films”, “When Will AIGC Be Considered Art? – From Morphological Similarity to Artistic Conditions”, and “Non-Death as the Only Certainty: The Phenomenology of Death in Video Games and Posthuman Narratives”. These articles explore the hot topics in the field of “Artificial Intelligence and Art” from three perspectives: the application of AI technology in artistic practices, the historical emergence of AI art, and the philosophical ontology of emerging art forms. They employ thought experiments combining practical experience, analysis of historical materials, and philosophical analysis to address certain hot debates occurring in this field, thereby outlining the research framework for such issues within the art domain.
Since 1950, the “indistinguishability” proposed in the Turing test has been regarded as the core standard for assessing the “intelligence level” of artificial intelligence. In the 1960s, Noll from Bell Labs consciously elevated “morphological similarity” as the standard for value judgment of AI-generated content (AIGC) in order to legally defend “computer art” in the field of artistic creation. However, the Stuttgart School’s computer art did not introduce the “morphological similarity” standard but instead directly linked computer art with art movements and artistic trends through specific discursive techniques, thereby carrying out a series of related practical activities based on this connection. This demonstrates that the strong demand for morphological similarity in the current AIGC field originates from historical contingencies and a cognitive bias guided by certain utilitarian purposes. From Goodman’s aesthetic theory and the denial of morphological similarity in conceptual art, we find that what truly connects AIGC and other computer-generated products with artistic works is the contextual conditions of “when it is considered art”.Art conditions.
In 1950, Alan Turing designed a method for empirical testing of artificial intelligence—by allowing computers to answer questions and having multiple unaware subjects distinguish whether the answers were generated by a human. The more subjects who failed to recognize whether the answers were from a “real person” or “computer”, the higher the AI’s “intelligence level” was deemed. This method, known as the Turing test, has been established as the standard for assessing the “intelligence level” of artificial intelligence and has been used ever since, significantly influencing the overall development goals of the AI technology field. Although the original version of the “Turing test” was limited to detecting the similarity between AI-generated text and natural language, through a kind of cognitive inertia, the “similarity” standard derived from this experiment has since been applied to value judgments of AI-generated images, audio, video, and other forms of works.
Current AI technology has enabled computer-generated products to achieve a high level of morphological similarity. For instance, a cross-cultural study in 2020 indicated that untrained subjects found it difficult to determine whether a product was generated by an AI model based solely on the images or text themselves. Since 2022, with the emergence of models such as DALL·E 2, Stable Diffusion, and Midjourney, the categories of “morphology” that AI can imitate and the accuracy of “similarity” have greatly improved. This situation has sparked various imaginings about AI-generated content (AIGC), such as the automatic production of artworks through AI. However, on the other hand, as AIGC becomes increasingly difficult to distinguish from artworks created by humans, it prompts people to contemplate whether AIGC can truly be recognized as “artworks”. To resolve this confusion, two aspects must be addressed:
First is understanding “how did we get here”—i.e., the historical reasons for the dominant position of morphological similarity in the value judgment of AIGC art. This necessitates tracing the historical origins of AIGC, exploring the circumstances of computer art’s history to investigate the reasons for introducing morphological similarity and examining its necessity.
Secondly, it is essential to inquire “what should it be for”—that is, to explore the necessary conditions under which computer-generated works, including AIGC, can be recognized as “artworks”. In this regard, the opposing viewpoints raised by conceptual artists and the analyses by the aesthetician Nelson Goodman can provide crucial references.
1. The Introduction of Morphological Similarity
The goal of producing works that are morphologically highly similar to human art has a long history. In 1966, Noll conducted an experiment with 100 participants, primarily employees from Bell Labs. He used an IBM7094 computer and a microfilm plotter to generate a series of images that were compositionally similar to Mondrian’s “Composition with Lines”. The experimental subjects were shown these images alongside replicas of “Composition with Lines”, and only 28% of them could correctly identify which image was generated by the computer, with 59% expressing a preference for the computer-generated images. Considering that the order in which the two works were presented could give subjects certain cues, he balanced the order of questioning and included analyses of other factors such as whether the subjects had received art training in subsequent studies. The conclusion drawn from the experiment was that each subject had their own varying degrees of preference for random patterns, but this was unrelated to their art training.
Through rigorous behavioral investigation, Noll provided a public proof of the similarity between computer-generated images and human-created works, but this “proof” had the appearance of psychological research from the outset. This complemented the work of his colleague at Bell Labs, experimental psychologist Bela Julesz. As early as 1960, Julesz had begun using computer-generated patterns to study the effects of different visual stimuli on binocular depth perception. Noll himself declared: “I’ve always been interested in the humanistic and perceptual aspects of technology. Therefore, I naturally wanted to use my computer-generated patterns as stimuli to investigate people’s aesthetic preferences. I learned about Mondrian’s paintings from a book… (they) were very suitable for being produced by computer programs.” Under the guise of psychological research, Noll was able to publicly showcase his computer-generated images, which were highly similar to Mondrian’s works.
As early as 1965, similar works by Julesz and Noll had already been exhibited at the Howard Wise Gallery. Intriguingly, at that time, they did not disclose their identities as Bell Labs employees nor did they mention their psychological research. In a 2016 memoir, Noll revealed the reasons behind this. According to archival materials from AT&T, the parent company of Bell Labs, AT&T expressed strong opposition to the public holding of the exhibition and even attempted to prevent it. At the time, Howard C. Craig, who was in charge of research and development, believed that a series of works, including the use of computer graphics, belonged to a “marginal field and personal interest”. He opposed categorizing computer art as part of “computer applications” and also opposed packaging this research as psychological research.
In response, senior management at Bell Labs protested Craig’s opinion through internal company meetings while also allowing Julesz and Noll to continue their research and creation using Bell Labs’ computer equipment under their personal names. In 1966, Bell Labs convened over 200 scholars for a symposium titled “The Human Use of Computing Machines”, publicly showcasing the explorations of many members, including Julesz and Noll, in computer graphics, animation, music, etc., and revealing the conflict with Craig.
In 1968, as a mark of resolving this internal struggle, AT&T released a documentary short titled “The Incredible Machine”. The segment of the film involving Noll and Julesz compared these computer-generated patterns to the painting techniques used by pointillist painters, stating that painters “only use a dozen patterns in their lifetime, while today’s researchers can generate so many with a computer in a day”; it also claimed that researching “how to separate relevant parts from unrelated visual information is crucial for transmitting three-dimensional colored visual information over ordinary phone lines”. The computer art practices of Julesz and Noll, through experimental psychology’s research on visual perception, were ultimately defended as preparatory work for the development of future communication technologies.
As noted by Steven Shapin and Simon Schaffer, the establishment of facts in science involves three types of “technologies”: first, the “material technology” of building “experimental instruments and operations”; second, the specific writing styles (such as the style of academic papers) required to convey the “produced phenomena” to those who did not directly witness them, which they refer to as “written technology”; and finally, the “social technology” that integrates the conventions that scientists should use when discussing and thinking about knowledge claims, thus enabling them to influence society. The pioneers of computer art at Bell Labs, as well as members of the Stuttgart School discussed later, and even participants in today’s AIGC, all involve the construction of basic theories that encompass these three types of “technologies”: in terms of material technology, the technology of using computers to produce images, which concerns how to manipulate computers and output devices; in terms of written technology, constructing narratives about material technology according to certain disciplinary paradigms and discursive systems, such as in Noll’s case, mainly establishing experimental environments and recording subject responses in experimental psychology, thus forming experimental reports; and in terms of social technology, it involves the ways individuals and groups reach consensus, such as in the earlier example, involving Noll persuading Julesz to accept the name of computer art, or the negotiations between the supervisors at Bell Labs and representatives from AT&T.
It is evident that among specific practitioners such as Noll and Julesz, their approaches to written technology still differ. Noll did not initially adopt the route of psychological research: compared to Julesz’s avoidance of discussing “art” and his use of a large number of randomly generated point-line patterns in experiments, Noll directly cited existing artworks and even registered copyright for the images he generated with a computer as artistic works. For Noll, adopting the written technology of psychology and narrating his intentions alongside painting through experiments and investigations into human perception and evaluation behaviors was already a compromise. To convert a creation that was indeed a personal interest, born out of a love for art, into an activity of knowledge production, Noll needed to produce generative products that were difficult to distinguish from artworks in form. For him, only those images of artworks that were “very suitable for computer program generation” could enter the realm of this “written technology” treatment.
This means that the important standard for value judgment in AIGC today—morphological similarity—was not initially introduced for reasons at the material technology level, but rather because the designers adopted the written technology of experimental psychology. According to the requirements of experimental psychology, only by first building a computer system capable of generating products indistinguishable from human works can one generate “scientific understanding”. Noll chose to adopt written technology, establishing a connection between computer art and cognitive science, as Boden pointed out, using computers to “create” art is intended to help people understand human creativity through computational concepts. As for whether computers can perform seemingly creative tasks now or in the future, and whether computers can recognize creativity—such as appreciating poetry written by human poets—these questions can only be affirmatively necessary in the sense that computers help explain the significance of human creativity. She also pointed out that whether computers possess “true creativity”—distinguished from merely appearing creative at a superficial level, where their originality is entirely imparted by human programmers—belongs more to the realm of moral attitudes, and even a negative answer to this would be inconsequential.
From the moment Julesz and Noll defended computer art through psychological experiment reports, through the systematic development of cognitive science, the morphological similarity between computer-generated products and human artistic creations was ultimately endowed with a function of epistemic promise—based on the morphological similarity between computer-generated products and artworks, people can use computer systems as models of human psychological functions, gaining knowledge about human aesthetic and creative processes.
At that time, the experimental goals led by Noll and Julesz, which leaned towards psychological research, conflicted with the “corporate tradition” represented by Craig—colleagues at Bell Labs believed that through this psychological written technology, the company could gain important knowledge about how humans perceive images and construct aesthetic judgments; however, Craig believed that only research related to telecommunications, such as telephones, constituted the “real business” of the experiments. Ultimately, the two sides reached a compromise through a kind of social technology coordination, pre-setting a purpose related to the telecommunications industry for these psychological experiments, thus forming the scenario depicted in the earlier mentioned documentary short “The Incredible Machine”—using computer-generated images to study human visual perception, and researching human visual perception as a foundation for the future development of visual telephony and other communication technologies. Through these two steps, Noll’s initially unintentional attempt to use a computer to create art out of personal interest was endowed with new meaning by written and social technologies. As the embryonic form of AIGC, the material technology of computer art rapidly developed with strong support from Bell Labs. However, this way of compromising with reality also suppressed the value generated in other aspects—those new theories that emerged alongside computer art, and their potential impact on art theory at that time and even in the future, became seemingly insignificant, excluded from the mainstream attention of society and academia.
2. The Stuttgart School and Non-essential Morphological Similarity
As one of the important “origins” of computer art, the creators at Bell Labs established the historical status of “morphological similarity” in the value judgment system of computer art. However, across the Atlantic at the same time, the Stuttgart School centered at the University of Stuttgart provided a strategy for legitimizing computer art that does not rely on morphological similarity. This indicates that the introduction of morphological similarity is not a logical necessity but merely a historical contingency.
The Stuttgart School, centered around philosophy professor Max Bense and his university, is typically depicted as “a loose and informal collection of artists, writers, theorists, architects, and composers, whose only commonality is their reliance on certain aspects of information aesthetics.” It initially formed as a literary group with core members such as Helmut Heissenbüttel and Reinhard Döhl. The focus on visual presentation effects, known as “concrete poetry” (konkrete Poesie), is one of the group’s most notable contributions, and in their further contemplation of poetry, the idea of using mathematical methods to depict artistic creation patterns began to emerge.
This idea became a reality as Georg Nees and others formally entered Bense’s tutelage to pursue doctoral degrees. Since Nees had previously worked as an engineer at Siemens and was familiar with computer programming, the theoretical propositions of the Stuttgart School gained the conditions for practical implementation on actual computer systems. In this context, the use of computer-generated images became an extension and expansion of “concrete poetry” and subsequent artistic propositions. It still holds epistemological significance, but this significance is more about providing evidence for Bense’s “information-theoretical aesthetics” rather than understanding human psychological perception and creativity.
In Bense’s view, the creation and appreciation of art are two aspects of the “artistic communication” process, where artworks carry “aesthetic information”, representing the artist’s viewpoints and “encoding” them; art viewers then acquire knowledge of the artworks through “decoding”, thus indirectly achieving communication with the artist (Figure 1). Such an information-theoretical aesthetics framework struggles to achieve its overarching goal of “explaining” human art. For instance, while it claims that value is a kind of “encoding function”, it fails to provide a specific functional expression or operational method, nor does it establish any conceptual connections between “operations” and “information” as art theory constructs the relationship between “art” and “aesthetics”. Consequently, its main point is simply that value should change along with the changes of aesthetic information—what is commonly referred to as aesthetic objects—which is already known. For example, Hegel believed that the beauty of art is superior to that of nature, meaning that aesthetic objects in the natural world have lower value than those provided by art. Information-theoretical aesthetics uses a seemingly complex and “scientific” discourse to obscure genuine aesthetic issues behind borrowed technical terms and symbolic formulas, and to a considerable extent, it overly simplifies previously established general views on art.
Figure 1 Bense’s Conceptualization of the Aesthetic “Communication System”
However, comparing the computer art of the Stuttgart School with that of Noll and others at Bell Labs, it is evident that “information-theoretical aesthetics” is actually a type of “written technology” tailored for computer art. It establishes an equivalence between technical matters and cultural matters, fully utilizing the ambiguity that exists between the everyday and technical meanings of terms (for example, the everyday meaning of communication as “interaction, communication” and the technical meaning of “communication”). Thus, it gains a space for semantic sliding: “communication” and “communication” no longer stand in opposition as in the struggle between Bell Labs and AT&T, but rather become interchangeable.
At the same time, compared to the indirect strategies adopted by Noll and others, the written technology that legitimizes computer art in the Stuttgart School maintains a terminological connection to traditional aesthetic thought, which also aligns with their chosen social technology. As an avant-garde literary group, the Stuttgart School provided assistance and shelter for poets expelled from churches and universities and actively participated in various avant-garde arts that had already emerged, which itself sufficiently marks the position of its members on the spectrum of artistic viewpoints. In 1968, the exhibition “Cybernetic Serendipity”, co-organized by Bense and Reichardt, became a landmark event often referenced in discussions about computer art; Bense’s establishment of the Studiengalerie TH Stuttgart (1958-1978) provided a venue for a generation of computer art creators to showcase their work. Based on these actively engaged works in contemporary art, Stuttgart became another center leading the development of computer art.
By aligning themselves with avant-garde art through social technology, the Stuttgart School closely connected computer art with the art world. This allowed them to avoid relying on morphological similarity to provide artistic status or achieve additional epistemological tasks, and instead maintain a more open and direct connection with the avant-garde art of the time. Therefore, the Stuttgart School associated their creations with the art recognized by history, allowing them to transcend the constraints of morphological similarity and relate to artistic thoughts that could connect with Kandinsky, Paul Klee, and the entire Bauhaus school.
Although due to technical limitations, the Stuttgart School also relied on forms that are “suitable for computer program generation”, there are no fundamental differences in material technology from Bell Labs. However, due to the differences in written technology and social technology, the two took different paths. This indicates that in the development history of AIGC, morphological similarity is actually a “local” practical option, rather than the only route for computers to engage in artistic creation. In summary, the widespread use of the “morphological similarity” standard in current AIGC value judgments arises from specific historical conditions and is not a logical necessity for computer engagement in artistic creation; thus, while it is “popular”, it does not possess theoretical legitimacy.
3. From Morphological Similarity to Artistic Conditions
In the process of developing artificial intelligence technology, Turing discovered the contradiction between the enhancement direction of engineering and the philosophical value judgments on “intelligence”. To separate the two, he proposed the “Turing test”, a thought experiment based on the “imitation game”.
Using “morphological similarity” as a standard has indeed achieved some results in the AIGC field, but these successful experiences merely demonstrate the effectiveness of this strategy, not its necessity: after achieving sufficient similarity, beyond the results of “imitating” human art, non-imitative artificial intelligence still exists. In this regard, scholars have already questioned the dominant “imitation” in artificial intelligence technology from aspects such as narcissism and self-image. Furthermore, as people increasingly utilize AIGC to write business letters and copy, synthesize voices and videos, and even use AIGC as drafts for artworks, the shadow of AIGC increasingly permeates human creation, thus reversing the machine’s imitation of humans.
In fact, the aforementioned tracing also prompts us to note that the 1960s, when computer art was born, was a period when artists strongly questioned morphological similarity. The most representative of this is conceptual art, which criticized Greenberg’s modernist critique project—morphology. It established a new essentialism of “art existing as a concept”. Conceptual artist Kosuth pointed out that establishing artistic identity based on morphological similarity is itself a historical contingency. In his 1969 publication “Art After Philosophy”, he discussed that using various forms to provide various visual experiences is merely a “minimal creative measure”; while numerous visually similar objects or images may seem relevant (or connected) due to similarities in visual/experiential “reading”, one cannot claim an artistic or conceptual relationship based merely on that. Therefore, what is truly important is the “art condition”, which is the premise for people to view artworks and the precondition for understanding and considering it as an artwork, so as to “see it as an artwork”.
Years later, Nake similarly connected computer art with conceptual art in his memoir. He wrote: “(Computer art) may seem to have left conceptual art behind, but at the same time, it gives the ideas of conceptual art a new strength. The entire pattern, structure, and categorical attributes of the image are described and handed over to the machine to solve the details. Individual works become part of a generative pattern.” “Individual realizations on the wall no longer carry the essence of art.” These all suggest a perspective different from the examination of “morphology”. In fact, even Greenberg, who was sharply criticized by conceptual art, had to admit the impotence of “morphology” for abstract art, stating, “I find that, apart from historical reasons, I have no other explanation for the current superiority of abstract art. Thus, what I have written has turned into a historical defense of abstract art.”
The morphological similarity constructed through the history of modernist art also encounters its limits in the development and changes of art. Although Greenberg still hoped for the emergence of a more inclusive standard in the future, when he resorted to the notion of a “not the only valid and not eternal” historical standard, it largely foreshadowed that static, established morphological analysis is insufficient. It should be replaced by an analysis of the changing “art conditions”, that is, only within the specific historical context of art can one make appropriate understandings of why those works in that era can be considered “art”.
Kosuth advocated that grasping the “art conditions” should be holistic. “Any and all physical attributes (quality) of contemporary works, if considered separately or specifically, have no relation to the concept of art. The concept of art (as Judd said, although he did not mean it) must be considered in its entirety. Considering a part of a concept inevitably involves aspects unrelated to its art conditions—or like reading a part of a definition.” Thus, the “art condition” of art is a conceptual state. The language forms used by artists are often “private” codes or languages, which is an inevitable result of art being unbound by morphology; hence, it can be seen that one must be familiar with contemporary art to appreciate and understand it. … Only in the fields of painting and sculpture do artists speak the same language.” Kosuth thus reversed the paradigmatic position of painting and sculpture in traditional aesthetics, relativizing them as exceptions within art. He particularly pointed out that “objects” are conceptually unrelated to the conditions of art, which not only removes the fundamental premise (the morphologically determined objects) upon which morphological similarity relies but also firmly binds conceptual art with the rejection of “objects”.
Analyzing Goodman’s thought experiment on “perfect forgeries” and his call for “when is it art” summarizes and develops the tendencies emerging in these artistic practices and criticisms. Goodman asks readers to imagine two images X and Y that appear completely identical, where X is the original work of a certain painter and Y is its forgery. From a visual perception standpoint, X and Y are indistinguishable. Goodman argues from the point that we can distinguish “X as the original and Y as the forgery”, indicating that we can discern such distinctions through analysis using scientific instruments or tracing historical archival information, which ultimately influences how we view X and Y, thus one cannot automatically deduce consistency in artistic status, value, or connotation based solely on their perceptual similarities. Therefore, Goodman asserts that the aesthetic characteristics of an image “include not only those things discovered through viewing it but also those that determine how it is viewed”. For theories that “completely disorient us and evaluate the aesthetic power of a work based on the intensity and duration of the stimuli it produces”, Goodman sharply points out that they are “absurd when judged superficially and are of no use in addressing any significant aesthetic issues; yet they have become part of our everyday nonsense”.
In AIGC, similar “forgery” situations also exist. Using the latest commercial version of the Midjourney model, users can generate portraits with a cinematic quality, even replacing real model photos in some cases. When contrasting such “realistic” portraits with actual photos, it is challenging to discern their differences (Figure 2). However, at the same time, through textual labels, people can understand that one image is AIGC while the other is a real photo. These methods provide information about the origins of the images, which also trigger people’s judgments regarding the meanings and values of the two images.
Figure 2 Left image is the actual photo of the model, right image is the Midjourney generated model with similar characteristics, left image sourced from https://www.sophiebuhai.com/products/ss18-double-pearl-earrings, right image sourced from https://twitter.com/juliewdesign_/status/1636444785819959316/
Goodman pointed out that once components beyond direct perception are included in the aesthetic characteristics, the discussion of aesthetics regarding art should shift focus from “what is art” to “when is it art”. The “when” he refers to can be understood from two aspects. In a relatively detailed aspect, it relates to the viewer’s viewing experience: knowing facts about how an image is produced, etc., will influence how people perceive the image, leading to aesthetic differences. Goodman highlights this time dimension with the phrase “subject x viewing at time t”. For instance, regarding a portrait generated by an early version of the popular open-source model Stable Diffusion, a viewer lacking relevant experience might treat it as a photographic work, evaluating its shooting techniques and speculating on the character’s emotions. However, with sufficient observation, “my current self” notices that the portraits generated using this model tend to have flaws in the hands, and through this characteristic, I can ultimately discern that this portrait belongs to AIGC and exclude it from the category of photographic works. More generally, as the number of times using the model increases and experiences with AIGC accumulate, “I” can develop a more intuitive understanding of the degree and scope of AIGC’s morphological similarity, thus becoming more cautious in viewing images in the future. At this point, although AIGC remains highly similar in form to artworks created by humans, “my” approach to “similarity” is no longer the same.
Of course, excluding AIGC from artistic forms such as photography or painting does not mean completely denying its potential to become art. This prompts us to pay attention to a more important aspect of “when it is considered art”, which refers to its artistic conditions. In Goodman’s view, aesthetic literature is “filled with desperate attempts to answer the question ‘what is art’. This question has become difficult to accept since the emergence of ready-made art, as it is often hopelessly confused with the question ‘what is good art'”. The reason for this lies in the fact that the objects or items in art do not play the same role in a fixed manner; “an object can only be an artwork at certain times and under certain circumstances”. For example, a stone picked up by the roadside is not originally an artwork, but when displayed in an art museum, it may be considered a ready-made artwork. Similarly, AIGC appearing in a technical demonstration often merely represents an intuitive display of technological advancement, which is different from those AIGC that people are willing to appreciate as artworks.
The artistic conditions are the basis on which people judge something as an artwork, determining “when”; Kosuth notes that this condition is holistic and has no disassemblable parts, even an artwork itself signifies a unique definition of art; Goodman, through his example of “signs”, turns attention to an object that may become “art” (including performances) and closely links the viewer’s viewing method with the object’s state and context, these signs ultimately point to the functions fulfilled by the object.
4. Individual and Collective Reflexive Communication as Artistic Conditions
When exhibiting computer-generated images at the Howard Wise Gallery (Figure 3), we can say that these images are “art at this moment”; whereas in other situations, such as when the same images appear as decorative patterns on promotional posters, they may not necessarily be termed “art”. As critic Michael Fried stated, “To perceive a thing, one must perceive it as part of the whole situation. Everything plays a role—not as part of the item but as part of the situation.” Starting from Kosuth’s emphasis on the holistic nature of artistic conditions, linking Goodman’s notion of the viewer’s changing viewing methods with the roles objects play, these various elements and figures effectively constitute a situation. As a whole, it is challenging to determine which elements are necessarily required or which elements can sufficiently form an artistic situation, but the situation must possess certain structural characteristics that indicate the fundamental conditions for its establishment. Since the establishment of an artistic situation means “it is art at this moment”, the conditions that enable this situation to be established are also the artistic conditions.
Figure 3 The “Computer-Generated Pictures Exhibition” held at the Howard Wise Gallery (1965), image sourced from A. Michael Noll, “The Howard Wise Gallery Show Computer-Generated Pictures (1965): A 50th-Anniversary Memoir,” Leonardo 49, no. 3 (2016): 232-239.
To find these structural characteristics, we need to place the constituent elements—what Kosuth refers to as “concepts”, what Greenberg refers to as “morphology”, and what Fried refers to as “objects”—into “brackets”. They are the objects being viewed or understood, those individual elements that inquiries into “what is art” and reductionist analyses tend to discover and identify. After doing so, examining the “remaining” parts of the artistic situation, the factors of the object’s creator, producer, performer, viewer, and other “human” factors emerge. At this point, the artistic situation presents a simplified model similar to the “communication system” described by Bense:
Artist/Performer—Other Constituent Elements—Viewer
From this model, we can further discover that other constituent elements are what we typically refer to as the “material” part, while artists, performers, and viewers all belong to the category of “humans”. In the communication that unfolds from this, it is not solely the artist or performer who initiates a certain “communication” process unilaterally; rather, they and the audience engage in a reflexive “communication” as a whole. On one hand, people do not perceive art in a manner that transcends time. As Goodman pointed out, “I” or someone else’s viewing is always accurately “at time t”, which reflects the limitations imposed on the self by time and implies that distinctions that cannot be made in the past or present may one day be possible in the future. Even within the context marked by that “time t”, the prior experiences of the individual serve as the foundation for this “viewing”. In this regard, the viewer can engage in an individual reflexive communication, which is manifested in self-touch, self-reflection, self-acknowledgment, and even self-correction when facing the work.
On the other hand, people do not exist independently in the world. Whether it is an audience present simultaneously (such as viewers watching a movie together in a screening room) or people who view sequentially, they collectively engage in a collectively reflexive communication triggered by those material factors, which naturally includes artists or performers who themselves are viewers of other artworks. At this point, regardless of how the material technology used for creation is packaged for what creative purpose, it ultimately allows the originally individual experiences to become shared experiences among people participating with different identities, making those who facilitate communication become “one” in a certain sense… as if speaking with one voice. Forming this “shared experience” is precisely the original meaning of the term “communicate”.
Thus, both the computer art of Bell Labs and the Stuttgart School do not adhere to any established rules in the formation and application of written and social technologies. However, the tension inherent in the opposed, reconciled, or conflated terms “communication” (technical term) and “communication” (everyday language) emerges from the different written and social technologies of computer art, becoming the key to understanding the structural characteristics of artistic situations. As “communication”, this dissemination requires the aid of perceivable objects and performances; even the concept art that strongly opposes form ultimately requires a textual form that presents those factors in brackets sequentially to the viewer. However, in effect, it achieves the “communication” of the creator towards the public. Compared to the “communication” of perceivable factors, this “communication” occupies a more fundamental position: it is precisely because the goal is to achieve “communication” that the perceivable factors required in “communication” come into being.
In AIGC, the situation is similar. Regardless of whether art is discussed, when users employ models such as Midjourney, Stable Diffusion, they often set a context for display. For instance, during the 2022 Chongqing mountain fire rescue, CCTV news used AI-assisted generated promotional images (Figure 4 left). This promotional image was juxtaposed with other photos taken by photographers, aiming to convey the disastrous nature of the event and pay tribute to the brave “reverse walkers”. Questioning whether the application of artificial intelligence technology is “emotional” is meaningless; rather, because the audience has emotions, this emotion drives them to assign emotional significance to the images—whether AIGC or photographs.
Figure 4 CCTV’s promotional image, left image sourced from https://weibo.com/2656274875/M2P0PsB5A, labeled “non-news image”, right image sourced from https://weibo.com/2656274875/M2VirB6jk?, which is a news photography image
Through certain attributes of objects, especially those manifested in sensory levels, to convey “thoughts” is the technical perspective of “communication”. It is judged by whether the initial text can be clearly received and whether the encoding matches the decoding results. However, “communication” contains a command: to understand, to listen. What truly allows art to occur is a discourse directed towards the self, while also directed towards others, which is this tendency of (self and mutual) understanding and listening. It makes the arrangement and combination of these elements become “art”—these elements do not necessarily have to become art; rather, at a certain “time t”, through the intersection with the viewer, they just happen to become art. It is evident that the basic structure of the artistic situation is a reflexive communication on two levels that unfolds when humans face material factors, and the fundamental conditions for its establishment lie in the emergence of such “humans”: on an individual level, they seek self-understanding and self-renewal; on a collective level, they possess the will for mutual communication and make efforts for it.
In conclusion, the most general structure of artistic situations can be simplified to the interaction between other constituent elements and participants such as artists and viewers, as well as the reflexive relationships among the participant groups themselves. The various constituent elements placed in brackets (“material”) are the results, records, or traces of participants implementing material technology, which arises from the need for mutual communication among people and influences how people subsequently face themselves and their groups.
As the pioneer of computer art Cohen stated, “We grow to create art; we create art to grow.” Through revealing artistic conditions and situational analysis, we can say: if AIGC holds significance, then this significance arises from people’s direct experiences as individuals and their life experiences as group members, and is associated with their self-understanding and renewal, as well as mutual communication and connection. This association condenses originally dispersed personal perceptions into collective or even “species attributes”, endowing it with the implications of anthropological aesthetics.
Indeed, computer systems capable of “imitating” humans in “creation” transform the image of artificial intelligence from rigid robots to empathetic artists. This vision has long been pursued by technical experts and is also a prevalent anthropomorphization of machines in broader societal contexts. In the foreseeable future, AI-generated content can certainly produce images that resemble paintings, sounds similar to musical instruments, and texts akin to those written by authors, making these products increasingly difficult to distinguish from human works. However, without human participation and observation, without humans assigning meaning to them, and without that moment of “touch” encountered in facing the generative products, they cannot become “art”.
Since the invention of photography, contemporary painting has gradually diminished its pursuit of “realism” under its influence, further promoting movements like modernism and cubism, which have altered the landscape of modern painting art to some extent. Photography itself, under the premise of continuous technological development, has gradually formed its unique artistic pursuit through borrowing and reflecting on the concepts of painting art. AIGC, by pushing morphological similarity to its limits, exposes its boundaries, but this limit serves to prevent a downward slide, helping us eliminate attempts to elevate purely perceptual morphological similarity to other levels, such as equating humans with machines, and subsequently applying human-specific “modes of life” to artificial intelligence. The rapid development of AIGC may lead to a historical abandonment of morphological similarity on a larger scale, allowing myriad technical paths beyond “imitation” to reemerge as artistic expressive means, and potentially prompting academia to initiate a new round of reflection on the entire value judgment system of modern art.
This article is a phased achievement of the National Social Science Fund’s Youth Project “Anthropological Aesthetics Research on the Historical Generation Issues of Computer Art” (Project Approval Number: 21CA169).
Author of the article: Zhu Tianhua
Images in this article provided by the author
For reading convenience, references have been omitted
Yu Shui | The Logical Dilemma of VR Films
Kong Degang | Non-Death as the Only Certainty: The Phenomenology of Death in Video Games and Posthuman Narratives
Zhu Tianhua | Re-examining Technical Analysis: The Meaning Generation and Artistic Potential of Computer Images
Zhao Yi, Jia Tianxiang | The Media Perspective of “Shadow-Game Integration”: Paradoxes, Dilemmas, and Key Issues
Geng Hongming | The Generation Methods of Art Concepts in the AI Era and Their Reflections
Click the link below to read
“Art Studies” Manuscript Invitation
Call for Papers for the Column “Chinese Art Concepts and Thought Sources”
Call for Papers for the New Media Column “Art Scene”
Welcome to subscribe to the 2023 issue of “Art Studies”
Recommended Reading for Articles in “Art Studies” 2023 Issue 2
Recommended Reading for Articles in “Art Studies” 2023 Issue 1
Recommended Reading for Articles in “Art Studies” 2022 Issue 6
Recommended Reading for Articles in “Art Studies” 2022 Issue 5
Recommended Reading for Articles in “Art Studies” 2022 Issue 4
Recommended Reading for Articles in “Art Studies” 2022 Issue 3
Recommended Reading for Articles in “Art Studies” 2022 Issue 2
Recommended Reading for Articles in “Art Studies” 2022 Issue 1
Click the lower left cornerto read the original text to purchase previous issues of the magazine.
