There is a lot of talk these days about how new technologies are changing different areas of our lives. Our contribution addresses AI contextualized in the visual art world.
The Portrait of Edmond de Belamy is auctioned at Christie’s, New York, with a record $432,000 sale. This work is the first creation by artificial intelligence. He is a Puritan-style French gentleman in a black suit and white collar. A machine is able to create faces and bodies and to make them completely similar to real ones. The synthesis work is entrusted to an algorithm, which is asked to analyze a database containing 15 thousand paintings.
GAN (Generative Adversial Network), the technology behind the opera, is the same one already used in interior design, astronomy and 3D model making videos. It has turned the world of imagery upside down by creating completely new but lifelike images. People who do not exist, deepfakes made through photos and videos. One person’s face is superimposed, imperceptibly, on another person’s body; voice and lip can be reproduced. It is the technology behind videos of politicians that have gone viral and made a dent in the world of pornography.
GAN operates and creates on its own initiative; having done so, it evaluates its own work.
How is AI applied to the art world?
Creative artificial intelligence involves the use of software capable of performing functions in the creative arts world. Service or utility AI consists of a range of services to aid in the dissemination or enjoyment of art. It refers to cataloging activities and the use of augmented reality-based tools that facilitate the sale and enjoyment of artworks. This can be read as an approach to art; it opens up the possibility for a personalized experience.
AI is being used in production with a view to machine learning. This second category of use, with controversial appeal, is a topic of discussion. From a technical point of view, the algorithm behind Edmond de Belamy’s portrait, for example, features a Generator and a Discriminator. Generator is responsible for creating new images, and the latter identifies which of these images are the result of human inventiveness and which are not.
The network watches, understands and learns from the artistic work of humans and it can be taught how to do it. The game consists of cheating cards. Discriminator can no longer recognize which portraits are machine-generated. Thus the result is achieved: the image is created.
The paradox of a machine producing art
Since artist and machine collaborate, the fruit of their labor is a co-creation. This provides a description of an equal relationship. The ethical issue originates from the nature of this relationship by questioning what its future meaning might be. Is computational creativity something contained and, therefore, controlled by human inspiration or does it flow externally? The network has access to an endless amount of data and potentially to all knowledge. Culture is explorable by computer without regard to temporal, linguistic or mental limits.
Even more emblematic, then, is the role of error. Related to this, not infrequently, are concepts such as originality and novelty. The network studies through criteria; anything outside these criteria is thrown out. System is capable of error so it means that it cannot be equated with a mere calculator, in which the incorrect is not contemplated.
The creative concept can creep behind the margin of error.