Rhizomatiks Beyond Perception

In the present day, anyone can generate AI images. This exhibition asks, “What is the value of these generated images?” The artwork shown in this exhibition was generated based solely on images independently created to train a proprietary new AI model.  The artwork shown in this exhibition was generated with a proprietary new AI model trained based solely on images that we have created up until today. As such, buyers will be able to generate images ad infinitum, adjusting the inputs to the AI model. By developing this own AI model and making the model itself available for purchase, Rhizomatiks hopes to inspire new perspectives and discussions on the relationship between AI and art. We invite you to visit and explore this innovative exhibition.

For deeper analysis, we will now pass the baton to Kazunao Abe, a professor at Tokyo Polytechnic University who has known Rhizomatiks from the start.

Rhizomatiks Beyond Perception: A Meditation on AI and Generative Art

Text: Kazunao Abe

Since the early 2010s, Rhizomatiks has garnered widespread acclaim as a driving force on the vanguard of the media art scene, both in Japan and on the international stage. Moreover, the nature of their unique practice has required them to navigate many challenges from a risk management perspective unknown to conventional contemporary artists. In a landscape where cutting-edge media technology is commonly levied as spectacle for entertainment, advertising, and promotional purposes, Rhizomatiks has laudably maintained a critical cultural and technological distance from them, focusing on the felicity of the work itself—surely no easy feat. The activities of this creative collective comprising artists, engineers, and designers can perhaps be likened to a cadre of expeditionary researchers ^(*1)surveying the uncharted terrain of media art and media performance, a still unclearly defined, and for this reason all the more exciting, new frontier.

Rhizomatiks’ mediations play out in two distinct directions. The first could be characterized as cutting-edge research with a thoroughly computational and data-oriented approach. By contrast, the second entails the development of proprietary hardware engineering mechanisms, as well as the research and practical implementation of sensing/control technologies. These two directions, when coupled with the dual axes of art ↔︎ performance（dance performance + electronic music performance^(*2)）, form the four-quadrant matrix that is Rhizomatiks. The majority of Rhizomatiks’ prolific output can be charted somewhere along this matrix, each project positioned on a fresh new bias.

By making their gallery debut at KOTARO NUKAGA, Rhizomatiks opens a new chapter, dipping their toes into the art marketplace for the first time in the gallery exhibition format. However, make no mistake, the new work contained in this exhibition addresses a larger purpose by presenting an innovative approach to this pivotal crossroads in human history. Over the span of a few short years, generative AI technologies have advanced at a rate that has far exceeded expectations. Not too long ago, NFT art seemingly emerged overnight as a hot topic on the art scene, piggybacking on the tokens that can act as surrogates of digital assets. In the words of Tina Rivers Ryan: “Because it forever points to a single asset, the NFT implicitly privileges the ideal of a stable, unitary artwork over the messy reality of digital projects that are dispersed, interactive, contingent, iterative, or ephemeral.” ^(*3) Yet it’s important that NFT art should also be imbued with this element of ephemerality found in generative art. This goes beyond simply storing the latest outputs of representational culture on a given user’s favored personal mobile device. More essentially, digital art has engendered another dimension that bestows cultural and economic value upon generative/regenerative data flow, even without representational images.

Rhizomatiks Beyond Perception ambitiously attempts to question the nature of art projects that use generative AI, extending this line of questioning toward their very existence. Rather than problematizing the results or repercussions of representational images created by AI, the exhibition seeks to demonstrate the nature of AI models themselves as data mediation. In other words, the exhibition constitutes an “attempt to unveil, visualize, and sell the training data and very AI models whose inner workings have heretofore been kept hidden from the public in a proverbial black box.” ^(*4)

Rhizomatiks is largely steered by Daito Manabe, the organization’s de facto leader and an internationally acclaimed media artist in his own right. However, we must not overlook the reciprocal synergies that arise from Rhizomatiks’ collectivistic approach to creation, their decentralized distribution of specialized R&D roles, and open discussion at the conceptual planning stage. To wit, Manabe is joined by a diverse team, two representative figures being Motoi Ishibashi and Yuya Hanai.Ishibashi has a comprehensive background in high-level engineering and computational research. Since the early 2000s, he has brought his vision to bear on a wide spectrum of art forms, paving new ground in the development and control of various devices and autonomous mechanics. Since joining Rhizomatiks in 2014, Yuya Hanai has had a hand in the development of numerous proprietary visual systems utilizing cameras and projectors, such as seamless MR, dynamic VR, and interactive lasers. In recent years, Hanai has garnered attention for his ethical approach to generative AI, developing image generating AI tools such as Mitsua Diffusion, Mitsua Likes, and Elan MitsuaMT, trained exclusively on openly licensed images in the public domain and images whose use has been permitted by the copyright holders, thereby eschewing those foundation models trained indiscriminately on information available on the internet.

In broad brushstrokes, I might summarize Daito Manabe’s modus operandi in two phrases: “a propensity for the unfinished” and “I am controlled, therefore I am.” From an art historical perspective, we might draw parallels to two artists who epitomize the pursuit of the “unfinished” ^(*5): Leonardo da Vinci and Marcel Duchamp（who was amply conscious of his Italian Renaissance forebear）. Da Vinci continuously evolved and updated his works, unhesitatingly incorporating the unproven new techniques of his day into his classically trained toolkit（even if this meant many of his works would not survive as a result）. Moreover, da Vinci was notable for his constant experimentation with unexpected combinations of the theoretical implications of those techniques, as well as the significance their implementation entailed for society to come. As such, his works remained perpetually unfinished, in the word’s many senses. Meanwhile, I personally think of Duchamp not as a “readymade” artist, but rather as a caster of molds. As a young man, Duchamp apprenticed at a printer, where he learned engraving, etching, and typesetting. Throughout his life, he evinced a craftsman-esque approach to the relationship between the original and the mold（which would in turn beget form in new substance）, captivated by the intangibly minute yet infinitely diverse shades of difference（“infrathin”）generated in the interstices by each pressing from the same mold. In principle, this means that the generative process continually exists within the margins of the mold （the medium）. ^(*6) Manabe’s views on technology run nearly parallel to Duchamp. When new technology emerges, Manabe embarks on projects that explore associated thought processes. This initial departure point rarely contributes to what is articulated in the “completed” artwork, instead serving as a stepping stone for reaching the next stage of generation, or another technological approach altogether.

Ultimately, Manabe’s praxis is made possible by his platforms that can be “thoroughly controlled.” On a human level, the question of control—how one group exerts control over another—has been a timeless topic of philosophical debate. We might posit that media is an extension of the body, and modern surveillance technology is an extension of the human（i.e. subjective）sense of sight. However, there is another topology altogether: envision a platform in which the body, existence, and subjectivity are dropped into a completely autonomously controlled technological world（à la the metaverse/multiverse）. As an example, consider Rhizomatiks’ recent media performance, Syn ^(*7), presented in 2023. Whereas spectators are typically a transparent, invisible presence, in Syn, the audience’s movement through the performance stage was recorded, and then played back in reverse as post-processed imagery toward the end of the performance, compelling the audience to confront their past selves through their stereoscopic lenses.

Always at the experimental cusp, Rhizomatiks Beyond Perception represents Rhizomatiks’ latest approach toward the act of “generation” through an AI lens（including proprietary technology debuted at this exhibition）. This critic looks forward to seeing what further wonders this foray into the subjectification and monetization of AI itself will hold from Rhizomatiks in the future to come.

Citations:

*1 Indeed, their media art division was until recently named “Rhizomatiks Research.”

*2 Daito Manabe also applies the content gleaned from his copious art and science research projects to DJ/VJ events with astonishing precision. His ability to harness and improve upon even the world’s most complex multilayer systems/programs, while still delivering intrinsic danceability, results in performances that exceed imagination.

*3 Tina Rivers Ryan “TOKEN GESTURE” (ARTFORUM MAY 2021）
https://www.artforum.com/columns/token-gesture-249731/(last access:2024/06/29)

*4 As media researcher Lev Manovich explains: “At this point we know much more about human creativity and how it works than we do about ‘AI creativity.’ In philosophy, psychology, cognitive science, and other fields, many alternative theories and types of creativity have been proposed. With time, we are likely to see the same for ‘AI creativity.’ But we are not at this point yet. After neural networks are trained on trillions of text pages or billions of images harvested from the Web, they can compose new texts and images. However, the abilities of the creative nets are not defined by traditional algorithms; instead, they are distributed among trillions of connections between billions of artificial neurons. In other words, we managed to create a technology that is in some ways similar to a human brain—and that is so complex we can’t understand how it works.”
Lev Manovich “The AI Brain in the Cultural Archive: What new artifacts emerge when we look at the next revolution in media?” (MoMA Magazine Jul 21, 2023) https://www.moma.org/magazine/articles/927(last access:2024/06/29)

*5 To be clear, “unfinished” can connote an unintended consequence (i.e. “unrealized/unbuilt”) or a more essential, deliberate state of “incompleteness.” Duchamp and da Vinci of course fall into the latter camp.

*6 Although Duchamp went on to create many works that could be classified as molds (or else “measures”), his early paintings are particularly indicative. For example, see how he depicted the left hand in Portrait of Dr. Dumouchel (dated 1910, currently in the collection of the Philadelphia Museum of Art). Notice how the hand itself is complemented by the surrounding space, or rather, how the connection between hand and space inhabits its own unique domain.

*7 Syn: Unfolded Horizon of Bodily Senses by Rhizomatiks × ELEVENPLAY (TOKYONODE inaugural program at Toranomon Hills Station Tower, Oct. 6–Nov. 12, 2023). In the second and third acts, the audience perceives the space in stereo vision, yet in a surprising twist, their own true shadows were also made visible in stereo vision with a special device. https://www.tokyonode.jp/sp/syn/(last access:2024/06/29)