＜prologue＞
I started a blog called “The Baby Boomer Generation’s Miscellaneous Blog”（Dankai-sedai no garakutatyou:団塊世代の我楽多（がらくた）帳） in July 2018, about a year before I fully retired. More than six years have passed since then, and the number of articles has increased considerably.
So, in order to make them accessible to people who don’t understand Japanese, I decided to translate my past articles into English and publish them.
It may sound a bit exaggerated, but I would like to make this my life’s work.
It should be noted that haiku and waka （Japanese short fixed form poems） are quite difficult to translate into English, so some parts are written in Japanese.
If you are interested in haiku or waka and would like to know more, please read introductory or specialized books on haiku or waka written in English.
I also write many articles about the Japanese language. I would be happy if these inspire more people to want to learn Japanese.
my blog’s URL:団塊世代の我楽多（がらくた）帳 | 団塊世代が雑学や面白い話を発信しています
1. The “Algorithmic Painting” That Sold at High Price Looks Like an Out-of-Focus Photograph
In October 2018, news broke that a painting created by artificial intelligence (AI) was sold at Christie’s, a prestigious New York auction house, for the high price of $432,500 (approximately 48 million yen).
The painting, a portrait of a man dressed in black by Edmond de Belamy, was created by OBVIOUS, a Paris-based team that researches and creates AI-powered paintings.
The work was apparently created using a deep learning algorithm called GAN (genetic artificial network) based on data from 15,000 portraits painted between the 14th and 20th centuries.
However, the finished product resembles an out-of-focus photograph, with the crucial face blurred.
The term “algorithm” refers to a computational or processing procedure used to solve a specific problem or issue. A flowchart is a diagram of this, while a program describes the specific steps for computer processing.
2. Issues with Art Using AI Technology
A well-known example of research using AI to challenge human intelligence is the development of AI for the games of Go and Shogi, which have become powerful enough to defeat professional players. Shogi AI is said to have memorized 700,000 games, a feat that would take a human 2,000 years to master. In August 2019, Microsoft developed an AI capable of playing Mahjong at the level of a top human player (a rank of “10th dan,” which only about 30 people possess).
However, when it comes to using AI to create paintings, music, novels, and other artistic works, issues such as copyright remain.
(1) Who owns what rights to content such as paintings, music, and novels created by AI?
(2) If a work created by AI happens to be similar to a human creation, can the human creator claim copyright infringement against the provider of the AI work?
(3) If an AI is fed all of a specific artist’s paintings, music, and novels into a single piece of data, and then has it analyze that “big data” (and predict future trends) to create a new work, would that constitute copyright infringement?
(4) If an AI is simply instructed to create a work and the AI creates it based on its own judgment, does copyright arise?
There seem to be many issues like these.
3. “Algorithmic painting” targeted at a single artist is promis
The painting “Edmond de Belamy” in the example above is a jumble of past portraits, and it’s almost impossible to tell whose works it was influenced by.
However, it’s not hard to imagine that if a deep learning algorithm were trained to study, say, Rembrandt’s paintings and paint “Rembrandt-style” paintings, it could be mass-produced with the same level of detail as a counterfeit artist.
Similarly, there are concerns that counterfeit (imitation) paintings of Renoir, Gauguin, Picasso, and other artists could be mass-produced.
In reality, “Rembrandt-style paintings” have already been produced using AI, and these are exact replicas of Rembrandt’s paintings with not a single trace of out-of-focusness.
This is the result of a project called “The Next Remburandt,” carried out by a joint team including Microsoft in the United States and Delft University of Technology in the Netherlands.
In Japan, research is apparently underway to have AI learn the major works of manga artist Tezuka Osamu, and create “AI-generated Tezuka Osamu works” that closely resemble his works in terms of story, characters, and drawing style (touch). One example is “Phaedon,” which was published in the manga magazine “Morning” on sale from February 27, 2020. The AI read Tezuka’s manga, including “Black Jack,” and created a draft scenario, which was then refined by a human scriptwriter.
4. “Algorithmic painting,” which transforms “landscape photographs” into “master-style paintings,” is also promising.
A German research team is also developing an algorithm that can transform a landscape photo into a work resembling the work of masters like Van Gogh, Renoir, and Munch.
5. Resolving Legal Issues with “AI-Created Art” is a Future Challenge
Regarding AI, self-driving cars are nearing practical application. However, in the event of a traffic accident, various variations are possible, depending on the circumstances of the accident, including accidents between two self-driving cars and between an autonomous car and a human-driven car. There are many issues that need to be addressed, such as who is responsible for each.
As more and more “AI-created art” is expected to be unveiled in the future, won’t it be necessary to urgently resolve legal issues such as copyright?