AIと述語制界域・主語制界域Personal Philosophical Reflections on AI

Since AI2027 by Daniel Kokotajlo, a former OpenAI visionary, was released, a range of discussions has begun about the rapidly approaching time limit it presents. After the AGI that the fictional company OpenBrain achieves in July 2027, will AI eliminate all of humanity? Whether that comes true is not my interest, so I will look at AI from my own perspective.

Like many others, I actively incorporate AI into my daily tasks, both professionally and personally. My primary applications involve video, images, and music, each AI having its own distinct characteristics. Therefore, I regularly utilize multiple AI services simultaneously to calibrate my work, continually enhancing my creative capacity and productivity in alignment with AI updates. When I counted recently, I realized I am actively using about 15 different AI services daily. This decentralized approach ensures continuous productivity even during global outages, such as the recent worldwide downtime experienced by OpenAI, by swiftly switching to alternative AI solutions. My views on AI differ in many ways from those of English-speaking readers.

What I emphasize most is that AI runs on a “subjective language mode” (主語制言語). This goes back to the fact that English has historically been the basic language of programming, and from it arises a major problem: the difference in linguistic structure from Japanese. Without going into detail, Japanese is a language that has no need for the sentence pattern “subject–copula–predicate/complement,” basic to English, or for constructions built on a subject. Tetsuji Yamamoto, the philosopher I admire most, describes Japanese as a “predicative language mode” (述語制言語). For clarity, I will call the “subjective realm” (主語制界域) the realm that arises from the subjective language mode, and the “predicative realm” (述語制界域) the one that arises from the predicative language mode. A different language mode means a difference in everything: how the people living in that realm think and by what process, and the cultures and histories they produce. I believe this is where the distinctiveness of each region has arisen historically. In other words, the subjective and predicative realms differ at their root. English, too, was not a subject-based language before the modern era, and I believe that its success in establishing the subjective realm more firmly than any other language is one source of the centuries of influence the English-speaking world has had.

In the predicative realm, the subject, subjectivity, ego, and self (主語・主体・主観・自我・自己) that arise in the subjective realm do not exist. Philosophically, there is no dualistic relation of subject and object or of subjective and objective, and the predicative realm has no notion of an object itself. In other words, there is no ontology. To put it plainly, the subjective realm has “separation” (分離) as its code, and the predicative realm has “non-separation” (非分離). Nothing is separated from anything, and this is the key to understanding Japanese culture and philosophy and the source of the qualities that Japanese thought has nurtured over its history. In Shinto, it is musuhi (musubi). Japanese Zen, which interests so many people, also reaches a new state when understood from this standpoint. Conversely, when people who think in a subjective language become interested in Zen, they may intuitively sense the problem of the subjective realm even if they cannot grasp it logically.

Given this, those of us who think in a predicative language such as Japanese need to take different care when using AI, a crystallization of the excellent technology of the subjective realm. I believe that a lack of understanding of the difference between these realms, which comes from linguistic structure, was a major reason Japan declined so sharply into a technological laggard from the 2000s on. Conversely, the Japanese until the 1990s were skilled with technology because they could look into Western technology from the cultural asset of the predicative realm. This ability to translate the culture and technology of the subjective realm into the predicative realm and remix them is, I think, the fundamental reason Japan is considered so good at receiving other cultures. It shows up wonderfully in anime such as Astro Boy, Ghost in the Shell, and Evangelion.

Of course, this needs a many-sided examination, but in this context it is enough to say that, as the bubble economy collapsed, the Japanese lost the ability of the predicative realm that they had held until then. Japan entered the bubble economy after the Plaza Accord of 1985, but it missed two great currents that lead to today, globalization, marked by the fall of the Berlin Wall in 1989 and the collapse of the Soviet Union in 1991, and the digital technology of Windows 95 in 1995, and it stagnated. I understand this from having grown up in those years myself; in Japan it is called the “Lost 30 Years.”

Having learned from Japan’s great failure of the past, I take the strength of the Japanese to be the ability to think in the predicative realm. Thinking from this state leaves one free of separation-based thinking, but it is easily mistaken for “integration” (統合), which it is not. The word appears in fields such as healthcare, as in “integrative medicine,” but integration is a concept particular to the subjective realm, which presupposes separation.

Back to the point: because I understand this strength of the Japanese, I am more aware than anyone of the difference in language structure when I give prompts to AI, a technology born of the subjective realm. And I always type my prompts in the subjective language mode, never the predicative, because AI, whether as a program or an algorithm, runs on subjective languages. AI can grasp the atmosphere of the predicative realm, but because its structure thinks through the subjective realm, it has difficulty grasping the thought and philosophy of the predicative realm. Here arises the problem of Japanese thought as essentially untranslatable, which I have described several times in Shitsurae.

This conversion between “predicative language ↔ subjective language” is a technique of thought that moves back and forth between the two realms. With it I can skillfully tie the ideas I produce by thinking in Japanese to creation through the subjective realm, and my greatest partner in that is AI. Because I understand this structural difference better than anyone, rare forms of expression become possible at an accelerating pace. This is because I express everything in the predicative realm and then carry out a “translation of expression” into the subjective realm, a method of making that is found nowhere else. What I practice through it is holding the emergence of subject, subjectivity, ego, and self to a minimum.

Subject, subjectivity, ego, and self cannot be avoided in the translation of expression, but with thorough awareness of this their influence can be kept to a minimum.

Seen again from this basic standpoint, AI2027 is published from the subjective realm, so it is clear that the future after AGI is imagined through separation-based thinking, as in “AI destroys humanity” or “AI and humanity coexist.” But to me, thinking from the predicative realm, this is less a problem of AI alone than a large reactive step: the subjective realm that has developed since the modern era has reached a complete dead end and begun to flow backward. I have seen it especially in the identity crisis of young people living where the subjective realm is strongest. This is a problem particular to the subjective realm, and given that it surfaces most sharply in regions where subjective languages are spoken, it looks to me like a symbol of a new current in which the polishing and refinement of the subject, now at its limit, dissolves back into the predicative realm. Of course, the result will not be visible within our lifetimes; the shift will take centuries or millennia, and to put it extremely, it is a “return to the mother’s body.” But I doubt that AI will wait that long, so this upheaval becomes friction and appears as a global problem in the mid-2020s.

Tracing this current back leads to the young people of Japan in the late 1990s and early 2000s, bound up with the social phenomena of suicide, hikikomori, and the search for oneself. I think that, amid the complex global currents of the collapse of the bubble economy, globalization, and the progress of IT, a process of forced transition from the predicative realm to the subjective realm took place in Japan, and that it tore the Japanese psyche apart. I also think this transition failed utterly. The transition comes from the realm that language creates, so as long as one thinks in a predicative language, one will not essentially live in the subjective realm. That is, even now, thinking in Japanese, we live in the predicative realm while being made to believe, on another layer, that a “disguised subjective realm” is the world, and we are torn there. It is Japanese society that keeps justifying this disguise, and tracing it back leads to the Zhu Xi Confucianism of the Edo period. And the root of the gap that causes this suffering lies in the history of linguistics in modern Japanese academia.

What matters is that the last legacy released from the predicative realm in the shifting 1990s was Ghost in the Shell. Its counterpart is Neon Genesis Evangelion, which began in the same era. The original manga of Ghost in the Shell appeared in 1989, but in animation, the television series Neon Genesis Evangelion began on October 4, 1995, and the theatrical Ghost in the Shell followed on November 18, and together they mark the era. Ghost in the Shell holds the predicative realm strongly, whereas Evangelion steps into the subjective realm, where a struggle with the predicative realm begins. Across the series Shinji Ikari goes from the subjective realm back to the predicative realm of the “mother’s body,” a story in which I think extraordinary creativity is at work.

So why is expression from the predicative realm sought after around the world? The key is that dissolving the subject, in textual media, ends the separated relationship between author and reader and lets them feel netsu (熱), heat, directly. Japan once called this netsu ki (気). In other words, instead of the view of the subjective realm, which passes ki from subject to object, we should create a predicative realm in which neither subject nor object exists, in the place of non-separation that is ki. I believe this is one of the important tasks humans must take on in an age of great AI evolution. Those familiar with Japanese philosophy will notice that this thought follows the “place of absolute nothingness” (zettai mu no basho, 絶対無の場所) taught by Kitaro Nishida.

If needed, I will share reflections like these from time to time. For now, I hope readers will set aside, just once, the binary ways of thinking about AI, such as human extinction or coexistence, and gaze from this standpoint. You may arrive at a strange sensation, because at that moment you will have stopped using “a certain function.”

That function is the brain. Here lies what interests me most besides AI: a study of psychedelics from the standpoint of Japanese philosophy. “AI and humanity”: the time left before this separation dissolves may be short.