In This Age of Overload, Is This Humanity’s Last Breath?

──When Physical Actions Collide with Artificial Intelligence in jactynogg zontaanaco
Text/Cheng Hsien-Yu

A collaboration between Japanese performing arts group contact Gonzo and artist yang02,entitled jactynogg zontaanaco, is among the works presented at the Reeling Hyperreality-themed Taipei Arts Festival. It explores the integration of machine learning technology and artificial intelligence (AI) in the interpretation of images and incorporates a live announcer in the style of boxing or wrestling matches. Audience members observe how machines read improvised interactive movements of humans and how they are interpreted using algorithms.
This work highlights the current limitations of AI. Moreover, from the perspective of Anatomy of an AI System research, it implicitly criticizes the huge hidden infrastructure that supports this technology.

Contrast is at the core of jactynogg zontaanaco, such as that between improvised physical contact among members of contact Gonzo and machine learning devices equipped with cameras. An automated guided vehicle (AGV) and a Segway modified to move on its own appear to be wandering bystanders. As they capture images of the performance, text descriptions of the actions are generated. This application of image annotation technology is clearly positioned as “an attempt to use language to replace or interpret physical expressions as they are presented on stage.” The creators emphasize the methods by which humans challenge AI, such as conflicts between the results of machine recognition algorithms and improvisational performances, which include wrestling moves and slaps, as well as nonsensical physical confrontations. A human announcer also provides live commentary. When “the improvised physical performance on stage is transformed into language, it is relatively abstract.” In this process of translating dynamic non-verbal human behavior into language, the contrast between the two is the core power of the performance.

There are many deliberate mismatches hidden within jactynogg zontaanaco. They seem to be flaws in the performance but are, perhaps, its core artistic proposition, triggering the flow of
ideas among audience members, artists, and machines and producing a humorous and thought-provoking atmosphere. Scholars, such as Melanie Mitchell and Ernest Davis, have
pointed out that despite the significant progress that has been made in machine learning to AI, AI lacks basic common sense. This work translates dynamic, non-linguistic human
interactions into static language, which deeply echoes the current capabilities and limitations of AI, including understanding of the subtle variations, contexts, intentions, and nuanced complexities of human behavior in an improvised physical performance. Although AI is adept at image recognition and generation of descriptive text, it is currently not capable of
capturing the meaning or emotion behind certain physical interactions among performers, such as trust, tension, and playfulness. The machine learning technology that is used to
describe the visual elements of the performance misses the mark in terms of understanding human intelligence and interactions on a deeper level.

The machines that are on stage are not only abstract processing units, but also part of a technological ecosystem. Anatomy of an AI System reminds us that typical AI systems are often exploitative products of global infrastructure. With recent trends from language models to open-source AI models, human performers have become data sources for AI, echoing how human activities are continuously extracted and monetized by AI on a planetary scale. The inability of AI to grasp the full meaning of human interactions is not only a cognitive limitation, but also a consequence of its inherent construction, which is based on big data and decontextualized statistical models. As such, AI is more about data collection and statistics than real “ability.”

The improvisational nature of contact Gonzo’s performance embodies the unmediated, present “sense of reality.” Real physical presence and interactions are contrasted with AI’s difficulty in capturing them, highlighting the challenge of translating raw reality into a digital linguistic format. Online authenticity is to some extent “complex, highly subjective, and socially constructed.” With the advent of social media, people strive to present their “real selves.” However, they face the challenges of optimism bias and impression management. Carefully planned personal content is usually just simple self-presentation, which can hardly be authentic when communicated to different audiences simultaneously. At present, AI output is purely descriptive and lacks insight into human intentions, becoming a metaphor for the superficiality of automated online interactions.

jactynogg zontaanaco creates an atmosphere for audience members to think critically and presents the limitations of AI based on a humorous Japanese-style wrestling approach, highlighting the mismatch between human actions and descriptions based on algorithms. The actors perform, slapping one another (Ōfuku Binta (or double slap)) or taking a sip of water and then throwing their water bottles at one another, which are wonderful interpretations of vitality and helplessness. Perhaps the future will be filled with humanoid robots. At that time, real human interactions will be extremely precious and real human flatulence may have great value and profound meaning.

photo by Yurika Kono