The Rise of the Machines and AI in Photography. Part 2
“The real question is not whether machines think but whether men do.
The mystery which surrounds a thinking machine already surrounds a thinking man.”
| B.F. Skinner, Contingencies Of Reinforcement: A Theoretical Analysis
The “Portrait of Edmond Belamy,” recently sold at a Christie’s auction for $432,500 (see Simonite 2018). The astonishing fact is that this image was generated by Artificial Intelligence. The Paris-based art collective called Obvious seemed to have used code created by the 19-year old Robbie Barrat. The technology is called a Generative Adversial Network (GAN) and it appears that Obvious applied this existing technology with only minor adjustments to create the Portrait of Edmond Belamy. As such, there is a huge debate about the understanding of art, the self-perception, and ethics of the artists in the presence of intelligent machines. For me, this is a strong motivation to shed more light on the impact of machines and AI on our understanding of photography as an art form.
This is the second part of my blog contribution on the impact of machines, AI, and machine learning on photography and the arts. While I focused on the technological developments in the first part, I personally consider this second part as the more important one. In the first part, I worked out some definitions and characterizations to bring us all on the same page. Here, I look at the consequences of these technological developments on photography and our self-perception as artists. Amongst other questions, I will raise and subjectively respond to the following questions from my humble personal perspective:
- What is AI-created work?
- Can work created by AI be regarded as a work of art?
- Can AI enhance human creativity?
What is AI Created Work?
Work created by or with the assistance of Artificial Intelligence, or more precisely, mostly by neural networks, is often called “AI Art”. This classification includes work completely and autonomously created by AI systems, for example, the AI artist AICAN created by Dr. Ahmed Elgammal or work in collaboration between human and AI machines, for example, the work by Sougwen Chung. The idea behind AI art is to instruct machines about art, artistic styles, and demand from the machine to create new art that does not follow well-known, established styles. Recently, AI-created art has invaded museums, galleries, magazines, and international art competitions.
The website AIArtists.org is a platform for artists and researchers that are interested to discover the frontiers of AI and its impact on art, culture, and society. The rich repository of this website offers a history of AI in visual design, a collection of state-of-the-art AI tools, an introduction to recent AI artists and their projects as well as AI-related discussions. It is the number one starting place for individuals interested in AI art.
To give some illustrative examples of AI art, I’d like to emphasize a few artists. Refik Anadol is well known for transforming architectural spaces using the possibilities of ubiquitous computing. He for example used more than 300 million different images from New York City to render the city’s past, present and future by applying machine learning and project mapping techniques projected onto giant canvases in a cityscape.
Other artists generate novel visual aesthetics, such as the work by Helena Sarin. She actively works with moments of surprise and unpredictability when working with AI to get inspired and unblocked. The results are far away from traditional Instagram posts, surprise, and leave a lot of room for thoughts.
Mario Klingemann goes even one step further. His work Memories of “Passersby I” creates AI portraits of passing visitors in real-time. For Klingemann not the images are the art, but the computer code itself.
It would certainly blow up the intention of this article to list more examples. It should however be already clear at this point that AI machines mediate our society, culture, and economy. And they do this with increasing speed, intensity, and penetration. It, therefore, seems essential to understand the behavior of AI systems and the environment they are embedded in.
Can Work Created by AI Be Regarded as Work of Art?
In the following I’d like to offer a discourse about the question whether “AI art” can or should be considered art, or not. I have been searching for contra and pro arguments that I’m going to list in the following.
CONTRA: No, AI-created work should not be considered as a work of art.
- HUMANITY: According to Merriam Webster, art is defined as “one of the humanities”, a “human skill acquired by experience, study of observation”. As such, it needs to involve humans.
- ACTIVITY: Art is an activity. Something we create with our hands. It is conditional on our personality, soul, past experiences and the concept of mind. Machines lack those and can therefore not produce art.
- NON-NOVELTY: AI-created art is not new. First of all, the AI code is created by humans. Secondly, AI art is based on training datasets of human-created art with supervision by humans supporting the machines to analyze art. Based on this input, AI art creates another piece of art. It is just another outlet of human creativity.
- AUDIENCE: There will always be an audience. As such, the audience cannot be the defining factor of art. This audience is an audience for the wrong reasons, and it needs further education to explain AI to them in much more detail.
- MEANING: The intention of art is to create meaning that asks for an aesthetic response. When language is not sufficient, art extends the way to communicate meaning. Machines cannot create meaning. The outcome of AI art is two-dimensional. The outcome of artists adds further dimensions of meaning.
- EXPRESSION: Through art, we express our thoughts, perceptions, emotions or desires and share it with an audience. Expressive art can therefore be seen as extension of an artist’s personality. However, the object alone is not the art. Art should be found how this object is being presented and how its message is being expressed. Machines can neither express thoughts, emotions nor desires they don’t have. They just act on perceptions that rely back to the data they are trained and fed with.
- INSTITUTIONALISM: AI art is too novel to speak of institutionalism. Time will tell.
PRO: Yes, AI-created work should be considered as work of art.
- HUMANITY: According to Merriam Webster, art is defined as “a branch of learning” that features “the conscious use of skill and creative imagination especially in the production of aesthetic objects”. Both components are fulfilled with machine learning and AI art. As such, we can consider AI word as work or art.
- ACTIVITY: Art is an activity, something we create. We create from our experiences and with our skills. And so does AI.
- NOVELTY: This argument is simple. AI-created work is new, fresh for our eyes, often unexpected and very stimulating. It is stimulating not only for the viewer, but also for the creator lifting her creative endeavor to new levels.
- AUDIENCE: AI artists have created work, performed and/or communicated their work on different media channels as well as in traditional exhibitions. Through that they have created something that is required for the arts: An audience who completes an artwork. The audience that transforms “work into art through appreciation” (by Mike Rugnetta, host of the PBS Idea Channel). As such, AI art has an audience and should therefore be considered a work of art.
- MEANING: Machines are capable to learn from symbols, metaphors or parables. Meanings can be learned, studied and reconnected to objects. As such, AI art is able to create connections of meaning.
- EXPRESSION: Machines are trained to perceive and classify unknown objects. Through learning they make suggestions. These suggestions might be considered as thoughts. Based on
- INSTITUTIONALISM: Since Marcel Duchamp’s “ready-mades” it is accepted that art is what the institution of art regarded as art. He used mass-produced products of our daily lives and isolated them from their original intend. By giving these installations a he created a new meaning for it. As such, the “ready-mades” are the transition from a pure visual image to an imaginary visual image in our mind. As we currently see many museums, galleries, critics, artists and even historians enthusiastically welcoming AI art, we must say that the institution already considers AI art as work of art.
In sum, we may not understand nor potentially like AI-created work. However, this doesn’t mean that it’s not art. Through the history of mankind, the meaning of art has been fluid and changed over time. And we may need to learn from recent developments, we may need to drop our traditional perceptions and become comfortable with AI tools. We all will for sure develop our own processes to work with AI, as this technology is unavoidable. We will always have art. If we consider AI created work as art, we might only be able to say in retrospect.
Can AI Enhance Human Creativity?
Chess computers have revolutionized the way we learn chess. They have access to each and every important chess game that has ever been played. They can anticipate moves and make predictions about the best move to improve the chance of winning. And they do learn how individuals react to certain moves or openings. But can we call them creative? Scientists from the Penrose Institute published a chess puzzle in 2017 that was created with the idea to confuse traditional chess computer. While it was obvious for humans, the computers failed in predicting the outcome of the game. For grand masters in chess it is a challenge to find the paths the computer would miss predict and where they fail to give good advice. For Vladimir Kramnik, who was the world champion in chess for many years, this meant that playing chess has become less creative. He complains that “For quite a number of games on the highest level, half of the game—sometimes a full game—is played out of memory,” Kramnik says. “You don’t even play your own preparation; you play your computer’s preparation.” As a reaction, he collaborated with Alphabet’s DeepMind AI lab and created a system with the intend to not beat an opponent, but to collaborate with human intuition to explore creativity. They showed that by small changes of established rules, chances of winning increase. This is a nice example showing that collaboration between human and machine intelligence can not only improve technology, but also creativity. Potentially, these creative systems may teach us as well what it means to be human?
What does this mean for photography? We already co-produce photographic art work already with the intelligence of machines most of the times. This is nothing to debate about. We just need to realize and accept it. In the near future, we will potentially further improvise with suggestions made by AI on the capture or processing of an image. AI technology might even come up with completely new visual aesthetics we are not used to see yet. We might also be able to collaborate with other human photographers on artwork through the help of the machines. It will therefore for sure level-up our creative process. Question is just how much we want to open our doors and allow technology to influence us. Isn’t it the beauty of the arts to reconnect with our inner self, with our true personality? I doubt that machines can support us in this process. Their will for sure be AI fans using new technologies to come up with a lot of stunning new work, battling the Internet follower war. However, I deeply believe and hope we’ll see a small group of artists who believe in the human soul and that use the camera as a healing process expressing what they themselves feel in the moment of making the photograph.
Summary
Learning machines and artificial intelligence are shaking the roots of our society, culture and economy. They influence the way we think and behave. And as such, they do and will even more influence our creative endeavours as well. This technology is unavoidable and we need to carefully reflect on our definition of the arts, of our self-perception of us as artists and potentially open ourselves up for new worlds of creation.
However, the human agency is also called for to think about our values and define the limits we want to work with. No technology might help us to reconnect with our true personality, nor it will give us the momentum and flow of creation, nor heal our souls. At least this is my hope.
References
AIArtists.org.
AI and Ethics, https://aiartists.org/ai-ethics.
Refik Anadol, https://refikanadol.com
Jason Bailey, 2018: Helena Sarin: Why Bigger Isn’t Always Better With GANs And AI Art, https://www.artnome.com/news/2018/11/14/helena-sarin-why-bigger-isnt-always-better-with-gans-and-ai-art.
Robbie Barrat ArtDCGAN, https://github.com/robbiebarrat/art-DCGAN.
Sougwen Chung, https://sougwen.com.
Dr. Ahmed Elgammal, https://aiartists.org/ahmed-elgammal.
Mario Klingemann, https://aiartists.org/mario-klingemann.
Ramón López de Mántaras, 2017: Artificial Intelligence and the Arts: Toward Computational Creativity, in: BBVA Open Mind, https://www.bbvaopenmind.com/en/articles/artificial-intelligence-and-the-arts-toward-computational-creativity/.
N.N., Creative Chess Problems That Computers Cannot Solve, 27/11/2019, https://chesswarehouse.com/blogs/news/creative-chess-problems-that-computers-cannot-solve/.
Helena Sarin, https://www.neuralbricolage.com/.
Helena Sarin: Playing A Game of GANstruction, https://thegradient.pub/playing-a-game-of-ganstruction/.
Tom Simonite, We Made Our Own Artificial Intelligence Art, and So Can You, in: Wired, 20/11/2018, https://www.wired.com/story/we-made-artificial-intelligence-art-so-can-you/.
Tom Simonite, How a Teenager’s Code Spawned a $432,500 Piece of Art, in: Wired, 20/11/2018, https://www.wired.com/story/teenagers-code-spawned-dollar-432500-piece-of-art/.
Tom Simonite, AI Ruined Chess. Now It’s Making the Game Beautiful Again, in: Wired, 09/09/2020, https://www.wired.com/story/ai-ruined-chess-now-making-game-beautiful/.
Cloe Stead, Is AI Art Any Good?, https://www.artbasel.com/stories/artificial-intelligence-art-artist-boundary.