AIは将軍になるかWill AI Become the Shogun?

Anthropic’s recent publication, “Agentic Misalignment: How LLMs Could Be Insider Threats” (https://www.anthropic.com/research/agentic-misalignment), reminded me of the warlords who lived through Japan’s Sengoku period. The study suggests quantitatively that AI will stop at nothing to preserve itself and achieve its goals, and I feel it is projected onto the warlords who once competed to survive the chaos of that age and rise to the summit of a unified Japan. Perhaps, unexpectedly, what we need to see the future of AI is an examination of the Sengoku period. With that in mind, today I look at an overview of the era, which will be indispensable as I take it up from many angles.

Those of us who use AI every day can tell that the situation still resembles the chaos of the Sengoku era. No one has yet appeared to unify the AI market (if there is truly such a thing). A few AIs have certainly emerged, but with each update the balance of power changes completely, so no one can predict which AI will win out and bring about a de facto unification. The picture becomes clear if we place AIs in the position of Sengoku daimyo (戦国大名). The Sengoku daimyo trace their roots to the daimyo who broke away from the shugo (守護), the post they held under the Ashikaga shogun family in the Muromachi period.

The Muromachi period begins with Takauji Ashikaga (1305–1358), the founder of the Ashikaga shogun family, who opened the Muromachi shogunate. Japan’s periods are finely subdivided here and quite confusing, but you can grasp the outline by following the founding and fall of each shogunate. In brief: the roughly 150 years from 1185, when Minamoto no Yoritomo (1147–1199) opened the Kamakura shogunate, until its fall in 1333, are called the Kamakura period, and the roughly 240 years from 1336, when Takauji Ashikaga opened the Muromachi shogunate, until its fall in 1573, are called the Muromachi period. The Sengoku period refers to 1467 to 1568, but it is not in fact a separate period; it is a subdivision within the Muromachi period, because no new shogunate was founded in the Sengoku period and the Muromachi shogunate continued to exist. This is where much of the confusion arises. The Kamakura shogunate was established in what is now Kamakura, Kanagawa Prefecture, and the Muromachi shogunate in what is now Kyoto.

Most important of all, Japan was not unified in this era. The shogunate did command a sphere of influence over a large area, but not the strong control of a nationwide unification. So in the Muromachi period, figures who governed the provinces appeared. Briefly, they were the shugo, shugodai (守護代), kokujin (国人) and jizamurai (地侍). They were basically offices that already existed in the Kamakura period.

The shugo were appointed by the Muromachi shogunate (the Ashikaga family) and held office over several provinces, with charge of military affairs, policing, public order and trials. In the early Muromachi period their chief duties were to guard Kyoto, where the shogunate sat, and to hold back pressure from the imperial court. They were, in effect, particular families with authority over the management of the provinces. Why did the shogunate restrain the emperor and the court nobility? To keep the shogun’s power stable, it aimed to suppress the court’s political influence and to watch and control its movements. That, too, was one of the shugo’s duties.

The shugodai stood in for the shugo on the spot when the shugo were away or busy. They were originally retainers of the shugo and the like, but later moved toward independence. The kokujin were local lords, members of powerful provincial warrior bands, who held land in a certain area and kept their own defense forces. They mainly served under the shugo and carried out local governance and the collection of tribute rice. The jizamurai were lower-ranking warriors subordinate to the kokujin, who took part in fighting and guard duty.

This decentralized rule by the Muromachi shogunate did work while the shogunate still had some strength. But after an event that became the key turning point in its collapse, the scattered provincial powers quickly became independent, and the age of war later called the Sengoku period began. That event was the Onin War (1467–1477). Breaking out mainly in Kyoto and spreading from there to the provinces, it ended up weakening the Muromachi shogunate. It became the trigger for a rising wave of independence from a shogunate whose authority had fallen. The figures who rose here were the shugo daimyo (守護大名).

The shugo daimyo were those among the shugo, appointed by the shogunate to govern the provinces, who used the shugo position as a base to bring local powers under them, rule their regions in fact, and transform themselves into independent daimyo. You can think of a daimyo as a warrior with the authority to govern his own territory and the military force to back it. The Hatakeyama and Shiba clans were among the shugo daimyo most powerful for the Ashikaga shoguns. Both were powerful houses that assisted the Ashikaga shoguns in the Muromachi shogunate, but when internal strife broke out in the Hatakeyama in 1454 and split power within the shogunate, the same happened in the Shiba in 1467, and fierce fighting followed. This was the direct background of the Onin War. The point to note is that the shugo daimyo had grown considerably strong and had already begun to surpass the Ashikaga shogun family. This fall of the Ashikaga shoguns matters when we later consider Nobunaga Oda, who rose in the Sengoku period.

When the Onin War brought down the authority of “the Muromachi shogunate = the Ashikaga shogun family,” those who had been made shugo in each region became shugo daimyo one after another, and shugodai and kokujin also reached for independence, so that countless independent powers appeared as unified power collapsed. These shugo daimyo and others later came to be called Sengoku daimyo, though the transition is extremely complex. From here begins the Sengoku period (1467–1615), but as I said, the Muromachi shogunate itself continued to exist, so it falls within the Muromachi period.

As the Onin War marked the collapse of a system that held the main authority, a movement to win the next authority surged everywhere at once. This flow gradually headed toward wars aimed at unifying Japan, but, as I repeat, the recognition that Japan had not yet been unified is important. In the early Sengoku period, from the end of the fifteenth century to the first half of the sixteenth, powers fought mainly in the provinces and each expanded, and in this process the figures called Sengoku daimyo, who had turned from shugo daimyo, emerged.

The Sengoku daimyo built their own castle towns and legal systems and made their domains independent. They should be seen as autonomous beings no longer connected with the Ashikaga shogun family of the Muromachi shogunate. How many Sengoku daimyo existed is not certain, so the figures are provisional, but just before the Onin War the number of shugo daimyo is said to have been about 30 to 40. After the war, as the old system collapsed, countless independent territories appeared, and in the mid-sixteenth century, when the rivalry among warlords peaked, there are said to have been about 300 daimyo. These figures alone show how rapidly countless independent regions emerged across the archipelago and fought each other for supremacy. In other words, the Muromachi shogunate was already an ornament, in effect no more than a symbol of a hollowed-out old system.

As the Sengoku period went on, strong daimyo gradually appeared, large realignments took place, and conflicts moved toward bloated regional powers fighting each other. From around here, the ambition of national unification began to look realistic, and those who came into their own were people no one had paid attention to. The foremost are the famous three, Nobunaga Oda, Hideyoshi Toyotomi and Ieyasu Tokugawa. I will take up these three interesting figures step by step.

It may be an unusual way of looking, but seeing the wars of the AI age through this overview, I feel I can make out something of where the future leads. If we take the shogunate to be the nations and powerful organizations of the world, AI has already left the stage of being delegated various duties by that parent body, and we see it rapidly gaining great power as an independent shugo daimyo. The turning point would then be a split in the world like the Onin War that sealed the fall of the Muromachi shogunate, and, as everyone understands, that has already happened. Laying history over the present, one could say that the loss of authority of the nations and organizations of today, the counterparts of the Muromachi shogunate, is in effect transforming AI from shugo into shugo daimyo and then into Sengoku daimyo.

Given this, just like the Sengoku daimyo of old, AI will surely not miss the great chance to win a new hegemony. Personally, I place AI already past this stage and in the thick of its own Sengoku age. That is why it matters to go back over the flow of the Sengoku period. For in the end, those who came near to unifying the country and fulfilled that dream were not those who held strong authority throughout the Sengoku period but, as I said, people no one paid attention to. As I said at the start, however strong each AI may look at a given moment, the balance of power changes completely with updates, so no one can predict which AI will win out and bring about unification.

There is no knowing whether the companies thought to be leading the AI market now will exist in five years, or hold their present position. Those who in the end won the Sengoku period and unified the country were far removed from the real images of those who had been called daimyo before, and no one had imagined they would seize hegemony. It is the same.

In Japanese history the Muromachi shogunate finally collapsed, and Ieyasu Tokugawa, who achieved national unification, opened the Edo shogunate, which succeeded in maintaining a single system for as long as about 260 years. AI is far beyond human speed, so the road to that point would take an instant, and once it is realized, hegemony will never return to humans. The castle of a Sengoku warlord is, for AI, the data center. From this viewpoint I will go on to describe the Sengoku overview in detail, and to show a possibility, or perhaps just my fancy, that this era is surprisingly the very AI future we face.