(Continued from Part 2)
The three images at the beginning all depict Oda Nobunaga. The portrait sculpture on the left was created for Nobunaga’s first death anniversary, the illustration in the center is from a book published in the Netherlands in the 17th century, and the image on the right was created by a generative AI based on the text in this book. While the real portrait sculpture is a statue of a Sengoku warlord in a dignified sokutai (traditional court dress), the images in the center and on the right are depicted as if they were a Western king or a Buddha statue.

Why does this gap occur? In this article, I would like to unravel this mystery and consider the “wall of creativity” common to creators about 350 years ago and us today.

After the flourishing of the Nanban trade during the Oda-Toyotomi period (around 1568-1600), the Dutch East India Company became the only Western access point under the Tokugawa Shogunate’s policy of national isolation. Arnoldus Montanus (1625-1683), a Protestant Reformed missionary from the Netherlands, wrote Gedenkwaerdige Gesantschappen der Oost-Indische Maetschappy in’t Vereenigde Nederland aen de Kaisaren van Japan, which is known as one of the earliest “Japanese histories” that comprehensively discussed Japan in Europe at that time.

The book contains descriptions of Nobunaga’s self-deification and the Honnō-ji Incident that differ significantly from Japanese historical records, as well as from Jesuit letters and scripts of plays performed at Jesuit seminaries. Here, I will use generative AI to summarize only the text related to the illustration in the center.
(1) Nobunaga built a magnificent temple in the village of Dubo, just half an hour from the capital. He placed an idol in the temple that was a living image of himself..
(2) The idol is seated on a circular metal plate elaborately carved like a seashell.
(3) The idol’s arms are folded on its stomach. A wide veil is draped around its neck and spread out on both sides. The chest is decorated with luxurious ornaments. Three strands of pearls adorn the neck, torso, and abdomen. But the most striking feature is the crown on its head.
(4) Nobunaga took the crown of the Muromachi shogun and placed it on the idol in Dubo. He did this to make people worship him as a god, and Nobunaga’s followers always obeyed his will.

The painters, engravers, and other creators who were in charge of the illustrations had no experience in Japan, just like the author Montanus. For this reason, they are said to have relied on “a few rough sketches in the Dutch trading post’s logbooks” provided by Montanus and “pictures of the Near East” that were far more accessible than those of Japan, and tried to visualize the textual information using their imagination.

Therefore, the illustration in the center was drawn in a unique way that could be interpreted as either a Western king or a Buddha statue. In addition to information bias, this may have been influenced by the image (preconceptions) of the East held by Westerners in the 17th century and the awareness of idol worship based on Christian values.

On the other hand, the image on the right is a Nobunaga statue created by modern generative AI based on the text of (2) and (3) above. AI learns the most likely patterns from vast amounts of data and generates images. At present, AI can learn patterns contained in data, but cannot fully understand cultural backgrounds and contexts like humans (although it is quite possible that AI will be able to make more advanced inferences in the future as technology advances).

For this reason, it was affected by biases in the training data (e.g., there are many images of Buddha statues) and cultural biases (e.g., the assumption that “statue” = “Buddha statue”), and associated “circular metal plate with shell-like carving,” “sitting,” and “statue” with “lotus pedestal of a Buddha statue” and “sitting Buddha statue,” resulting in a figure resembling a Buddha statue.

Both the 17th-century illustration and the modern generative AI image show the “wall of creativity” due to biases in information and preconceptions. This can be said to be a universal issue of humans and technology that remains unchanged even after about 350 years.

So how can we overcome this wall? We, living in the present age, need to be more conscious of the “biases in information” and “preconceptions” that we tend to fall into, and to consider things from diverse perspectives. Our company believes that by understanding the characteristics of generative AI and utilizing it appropriately, we can open up new creative possibilities through the collaboration of humans and AI. We hope to be able to talk about the specific details next time.

 

Image Source:
（Left）Kyoto Tourism Navi
（Center）Münchener Digitalisierungs Zentrum (MDZ)
（Right）Generated by Microsoft Designer

References:
Frederik CRYNS and Masaaki Miyata, “<Document research>Information about Japan that was disseminated in 17th century Holland: Descriptions of Japan in De Vries’s ‘Curiosities of the East and West Indies’”, Nihon kenkyū (Journal for Japanese Studies), Vol. 33, International Research Center for Japanese Studies, 2006

Montanus, et al. “Montanus’s Account of Japan”, Heigo Shuppansha, 1925, National Diet Library Digital Collection