Recherche FFJ Research Statement Koichiro Eto

Koichiro Eto

A Study of Cognitive Differences between Europe and Japan on Artificial Intelligence-Generated Creations



Artists are always searching for the most advanced expressions. If there are new technologies, they may be interested in them and use them to express themselves. In the expression of such artists, the nature and limitations of those technologies may come to light.

One of the most advanced media of our time is Artificial Intelligence (AI). The cutting-edge approach to artificial intelligence by artists seems to reflect people's perception of artificial intelligence.

People have always had complex feelings about the intelligent beings they have created. By examining in detail the evaluations of the new artistic expressions created by artificial intelligence, it will be possible to consider how to accept artificial intelligence in the future. In this research, we will interview artists who create art with artificial intelligence, and investigate their ideological background. In addition, through psychological experiments, we will compare and examine how people accept the expressions produced by artificial intelligence.

The proposer has been working as a media artist. I have always been interested in artistic expression using new technologies. My university years (1991-1994) coincided with the spread of the Internet. I was the first artist in Japan to present a work using the Web.

My work "WebHopper," which I created in 1996, visualized users browsing the web on a map of the world. The sensorium project, including this work, won the Grand Prix (Golden Nica) at the 1997 Prix Ars Electronica held in Linz, Austria.

Later, I became interested in collective intelligence on the Internet. We created Modulobe in 2005, a platform for creating complex 3D models that move like living creatures by combining simple shapes like matchsticks [1]. Users can use this application to create and submit a variety of models. As users influence each other, a variety of creations are created. In this way, we studied how collective intelligence is created on the Internet.

In 2011, we launched "NicoNicoGakkaiβ" as a place to promote user-participatory research [2]. In the world of the Internet, new media such as YouTube and Nico Nico Douga have emerged. Among them, there were people who posted videos of their research results. They are not professional researchers, but they are citizen researchers (we call them wild researchers).


Deep learning, a type of artificial intelligence, has been generating new innovations since around 2012. An image recognition contest held at the ILSVRC in 2012 achieved a dramatic performance improvement of nearly 10% from the previous error rate [3]. In February 2015, the accuracy of object recognition by machines surpassed that of humans. The artificial intelligence is now able to judge objects more accurately than humans. The invention of this technology soon spilled over into the industrial world, and object recognition startups became the new trend.

In March 2016, AlphaGo, developed by Google DeepMind, defeated South Korea's professional Go player Lee Sedol 4-1 [4]. Because Go is more complex than chess or shogi, it was thought that machines would not be able to beat humans for the time being. However, the development of deep learning has been remarkable, and it has beaten the human champions much faster than expected.

A key concept in understanding deep learning is "disentanglement" [5]. For example, human knowledge, such as image recognition or a game of Go, cannot be represented by some simple mathematical formula. It is a complex interaction of parameters that cannot be neatly described. Deep learning can take such a complex set of intertwined parameters and untangle the entanglement by accumulating deep layers. In this way, deep learning has made it possible for machines to replicate the way humans recognize complex problems that could not be described before.

Deep learning can recognize images, but it can also generate images by using their features in reverse: deep generative models called VAE [6] and GAN [7] allow artificial intelligence to generate images. In particular, GANs have made it possible to generate high-definition images that are indistinguishable to humans.

Artists are using the most advanced technologies to search for cutting-edge expressions. If artificial intelligence is the most advanced technology of our time, then artists will use it in the creation of their works. Thus, art works using artificial intelligence has become the latest craze in contemporary art. Especially in Europe, artists using AI art are appearing one after another. "AI Artists" [8] is a list of artists who are creating art using artificial intelligence, and many of the artists on the list live in Europe. Is this a coincidence? I don't think so. I believe that there is a Western art tradition that considers the relationship between media and artistic expression in the background.

Will artificial intelligence surpass humans?

There has been a lot of discussion about artificial intelligence. Among them, the term "singularity" [9] is often heard. This word refers to the time when the artificial intelligence will surpass that of humans intelligence. It is said that after the singularity, artificial intelligence will evolve to such an extent that humans will not be able to understand their intelligence. Therefore, it is expected that after this singularity, humans will not be able to fully understand the intelligence of machines.

There is also the argument that artificial intelligence will take away human jobs [10]. If artificial intelligence is beginning to replace human capabilities, as it has in the world of Go, then sooner or later it will replace our own jobs as well. This is the thinking behind the artificial intelligence threat theory.

However, I myself am skeptical of this kind of argument. To begin with, at this point in time, the computing speed of computers is already far beyond that of humans. The same goes for memory. However, there are few people who think that computers are taking away human jobs. The difference is whether or not we can use the new tools when they come out. The same is true for artificial intelligence. It is not a question of whether it is a threat or not, but rather the need to determine exactly how it will affect human work.

Nevertheless, it is true that there are people who consider artificial intelligence to be a threat. How should researchers of artificial intelligence approach the natural perception of such people?

In thinking about this, the art of artificial intelligence provides a good sample. As mentioned earlier, many of the recent attempts at art of artificial intelligence are using deep learning. The good thing about art is that it is visible. To understand deep learning itself, we need to understand the complex theoretical background, but the images produced by it can simply be seen and felt by our own eyes. How do we feel it? Is it really possible for a picture created by artificial intelligence to surpass a picture created by a human? What kind of emotions would be projected onto a picture created by artificial intelligence? By examining such things, it should be possible to verify how the intelligence created by artificial intelligence will be accepted by people in the future.

At the same time, it is also possible that the way in which such creations of artificial intelligence are accepted in Japan and Europe will differ. In Japan, comics in which robots play a possitive role (such as Astro Boy and Doraemon) are widespread [11]. In addition, robot animation such as Gundam is also widely seen. Perhaps due to this background, Japanese people are not afraid of using robots in their daily lives [12].

On the other hand, in Europe, there is what is called Frankenstein syndrome [13]. In Frankenstein written by Mary Shelley in 1818, there is a scene where the "monster" takes revenge on its creator one after another [14]. Ridley Scott's Blade Runner also depicts a story where the creation kills its creator. The repeated depiction of such stories may exist as a background to people's fear of artificial intelligence.

Research Scope and Methodology

There are two aspects to this research.

1. Content study

Why do artists engage in art with artificial intelligence? There may be various reasons for different artists. When a new field is about to be opened up, there is historical value in examining the concerns of its pioneers and predecessors. Through interviews, we will explore the background of each artist.

  • What kind of activities have you been involved in?
  • Why did they decide to use AI?
  • What has influenced them? Works, people, events.
  • What are their future plans and visions?
  • What do they think about the singularity and artificial intelligence threat theory?

In addition to the content of the interviews with the artists, the video and audio itself also has historical value. Therefore, the interviews should be recorded in high-definition video and audio so that they have value as content in their own right. If possible, these interviews should be made available on the Internet as content.

It is also important to choose the right artist to be interviewed. Research current AI artworks and create a map of current AI art. In some cases, I also create my own works to understand the current situation.

2. Psychological experiments

In order to investigate the differences in the acceptability of artificial intelligence between Japan and Europe, a psychological experiment will be conducted. As an experimental method, we will first prepare two types of creations, one created by artificial intelligence and the other by humans (the number of works will be discussed later). Next, we will gather subjects in Japan and Europe. We will show the creations to these subjects and have them evaluate them using the SD method [16]. As for the evaluation axes, first, in addition to the usual axes of painting, we will set axes that are considered to be influenced by artificial intelligence, such as human and mechanical. As for the creations to be presented to the subjects, we assume paintings first, but also consider the possibility of music. We will set up three conditions for the experiment.

  1. No prior information will be given.
  2. Give the correct prior information. In this case, each picture is presented as either a picture made by a human or an artificial intelligence.
  3. Give false prior information. In this case, for each picture, the human-created picture is presented as the artificial intelligence-created picture. Or vice versa.

It is assumed that there will be differences in the evaluation of each picture due to differences in the prior information. We will examine these differences.

Next, the differences in the evaluation of each picture will be compared between Japan and Europe.

Expected Results

The discussion of creations by artificial intelligence may repeat the discussion of the early days of photography [17]. With the invention of photography, people were able to use cameras to record what they saw. This raised the question of the significance of the painter's existence.

With the invention of photography, there was a demand for the expression of the artist's inner world. This led to the birth of modern painting. Clement Greenberg states that the condition for modern art is "to pursue that which is inherent in the nature of the medium of art". He theorized that art becomes purer when it eliminates elements borrowed from other art forms.

If it becomes possible for artificial intelligence to replace the work of painters, the meaning of their expression will be questioned. If this happens, will the inner life of the painter become more important than ever? Or, as Greenberg puts it, will the meaning of "the nature of the medium" be further questioned? These changes need to be seen in the context of the times, and are the subject of this study.

In terms of cultural background, it is possible that there is less aversion to creation by artificial intelligence in Japan. In Japan, there is a tendency not to actively distinguish between artificial and natural objects. For example, Japanese gardens are designed in such a way that the scenery in the back and the garden itself seem to be connected. This is called borrowed scenery, and the entire landscape is composed as one. Therefore, although the garden itself is artificial, it is accepted as something continuous with nature. This is one of the characteristics of Japanese culture [19].

Speaking of creations by artificial intelligence, there are two types of creations by artists: the artificial intelligence itself and creations by artificial intelligence. At this point, how do we distinguish between these two types? Do we consider both to be creations of the artist? Or do we consider the artist to have created the artificial intelligence, and the artificial intelligence to have created the creation? There would be a difference between the two. In the latter case, the artist has created a new artist, the artificial intelligence. It can also be understood as an artificial creation of a new nature.

As a historical background, I would like to refer to the Italian Futurists [20]. The Futurists were inspired by the factories and locomotive engines that appeared in the industrialized society that was emerging at that time, and they thought that such things would create new forms of expression. For example, they considered the noise of factories as a kind of musical material and sublimated it into a musical work. They valued the sense of speed and thought about how to reproduce it on the canvas.

What kind of development can we expect in the art of artificial intelligence in the future? In order to examine this development, I would like to examine people's perception of it today.


1. Eto, Koichiro, et al. "Modulobe: A creation and sharing platform for articulated models with complex motion." Proceedings of the 2008 International Conference on Advances in Computer Entertainment Technology. 2008.

2. Eto, Koichiro, Tom Hope, and Hideaki Takeda. "A pattern language for an open academic society with non-professional users." Proceedings of the 21st Conference on Pattern Languages of Programs. 2014.

3. Krizhevsky, Alex, Ilya Sutskever, and Geoffrey E. Hinton. "Imagenet classification with deep convolutional neural networks." Advances in neural information processing systems 25 (2012): 1097-1105.

4. Silver, David, et al. "Mastering the game of Go with deep neural networks and tree search." Nature 529.7587 (2016): 484-489.

5. Bengio, Yoshua, Aaron Courville, and Pascal Vincent. "Representation learning: A review and new perspectives." IEEE transactions on pattern analysis and machine intelligence 35.8 (2013): 1798-1828.

6. Kingma, Diederik P., and Max Welling. "Auto-encoding variational bayes." arXiv preprint arXiv:1312.6114 (2013).

7. Goodfellow, Ian J., et al. "Generative adversarial networks." arXiv preprint arXiv:1406.2661 (2014).

8. Benney, Marnie. The Top 25 AI Artists of 2020. Accessed Feb. 5, 2021.

9. Kurzweil, Ray. The singularity is near: When humans transcend biology. Penguin, 2005.

10. Brynjolfsson, Erik, and Andrew McAfee. Race against the machine: How the digital revolution is accelerating innovation, driving productivity, and irreversibly transforming employment and the economy. Brynjolfsson and McAfee, 2011.

11. Alt, Matt. Pure Invention: How Japan's Pop Culture Conquered the World. Constable, 2020.

12. Murakami, Takashi, ed. Little boy: The arts of Japan’s exploding subculture. Yale University Press, 2005.

13. D. S. Syrdal, T. Nomura, H. Hirai, and K. Dautenhahn, “Examining the Frankenstein Syndrome: An Open-Ended Cross-Cultural Survey,” in Social Robotics: Third International Conference, ICSR 201

14. Shelley, Mary Wollstonecraft. Frankenstein, Or, The Modern Prometheus: In Three Volumes. Vol. 2. Lackington, Hughes, Harding, Mavor, & Jones, 1818.

15. Scott, Ridley, et al. Blade runner. Los Angeles: Warner Home Video, 1982.

16. Osgood, Charles, et al. The measurement of meaning. No. 47. University of Illinois press, 1957.

17. Bajac, Quentin. Discoveries: The Invention of Photography. Harry N Abrams Incorporated, 2002.

18. Greenberg, Clement. "Modernist painting." Modern art and modernism: A critical anthology 5.6 (1982).

19. Kato, Shuichi. Space and time in Japanese culture. Iwanami Shoten, 2007. (Kato, Shuichi. Le temps et
l'espace dans la culture japonaise. CNRS Editions, 2009)

20. Goldberg, RoseLee, and Christiane Paul. Performance art: From futurism to the present. London: Thames & Hudson, 2001.