Algorithmic art
The increasingly blurry line between art and AI
My interest in digital art prompted me to visit the exhibition “Uncanny Valley: Being Human in the Age of AI” held at the De Young Museum in San Francisco, CA from Feb 22, 2020 — June 26, 2020. The artworks explored the concept of what it meant to be human and the increasingly blurry line between art and artificial intelligence (AI) algorithms that were used to generate art. The concept of the “uncanny valley” was first introduced in the 1970s by Mashiro Mori, a professor at the Tokyo Institute of Technology to describe his observation that robots are increasingly becoming more human-like, but only up to a point. For me, the exhibition raised some fundamental questions about the role of the artist and the value of AI-generated art.
- Can art only be created using traditional mediums like acrylics, oil paints, color pencils, watercolors, and clay or can AI modeling and algorithms be also used to create art?
- What is the true measure of artistic abilities? Does technology generated art display the same level of artistic creativity as traditional art?
- Who has copyright over art if it is generated by a computer? Should artists be paid to create generated art?
- Who is the artistic persona behind generated art — the computer or the human being?
- How can the intersection of arts and sciences be used to create diverse artistic expressions?
- Can bias be incorporated in artistic representation, especially when art is generated using code?
- What are the limitations of algorithmic art?
The most phenomenal piece and in my mind the showstopper was the piece, They Took The Faces From the Accused and the Dead — an art installation by Trevor Paglen, that was organized by an algorithm. In this piece, the artist presented a collection of human facial images similar to what computer vision uses for facial recognition. He highlighted his concern about datasets that are used to train AI systems and the “biases, errors and ideological positions” that are inherent in those datasets. This is an inherent problem with AI and facial recognition technologies, which is often misused by the government, corporate employers, and law enforcement agencies to further their own agendas. See my article here.
Another work that caught my attention was the Mythiccbeing by artist Martine Syms. While it is based on the concept of threat modeling in cybersecurity to map out a system’s vulnerabilities, the artist uses it to highlight her own insecurities and negative thoughts.
In The City of Broken Windows, artist Rita Steyer uses a combination of video, text, and sound to portray a society in which technology functions as a tool of capitalism for the privileged few, and exploits the underprivileged, which exacerbates social inequalities.
Another interesting piece was the sculpture of the Amazon worker cage by Simon Denny, which demonstrates how a sleek design hides the humanitarian cost of abysmal working conditions and environmental disruption.
While viewing these AI-generated artworks, I felt that the artistic vision was more important than the actual art piece. To fully appreciate them, it helped to have a written explanation about the source of inspiration and the context in which the artwork was created. If art is evaluated based on the emotions that it evokes in the audience, these art pieces were definitely able to convey a sense of disillusionment, disarray, and destruction.
What is algorithmic art? What is an algorithmic artist?
A great hands-on activity to see how machine learning works to create art is through Google QuickDraw. You can use the software to draw the requested image within a time frame. While you are drawing, the software attempts to identify the image until it ends up guessing it right or gets stumped. The program ingests the drawings provided by millions of users and applies algorithms to correlate images to words. However, it also displays how limiting and biased the dataset may be to user perception of how objects are supposed to look.
Recently, the Digital Humanities Research Laboratory at Rutgers University created AICAN (an artificial intelligence creative adversarial network) that can create algorithmic art by learning existing styles and customizing images, all on its own with little or no human involvement. What is interesting is that the quality of the artwork is so superior that human beings cannot identify any differences on purely technical grounds. We may even be able to replace the artist. Meet Ai-Da (named after the first female computer programmer Ada Lovelace) and the “world’s first ultra-realistic robot artist”, who uses a camera in the eye and a robotic arm to draw using algorithmic logic.
Is it then possible to outsource the artist with an AI robotic system? After all, we have transferred many of our manual factory jobs to robots. Maybe, a happy medium can be reached whereby human beings can collaborate with AI to create richer, diverse, and even more intense artistic expressions. It is important to remember that art (whether it be human or computer-generated) usually develops in a specific social context and is inspired by politics, history, places, and people. Art must represent the artist’s vision and his interaction with the world. At the Chelsea Gallery, the exhibition on AI art called “Faceless Portraits Transcending Time,” generated tremendous interest because it was a “collaboration between an artificial intelligence named AICAN and its creator, Dr. Ahmed Elgammal.” The renowned auction
Christie’s recently sold its first AI art — a blurred face for $432,500 created using machine learning. Thousands of portraits were fed into an algorithm to teach it the basic aesthetics of portraiture and the result was the “Portrait of Edmond Belamy”.
While it seems that manual labor in creating art may be replaced by a robot, human involvement still exists. Only, the role of the artist has changed and they are now required to write code to generate art.
The intersection of humanities and sciences to create art
Incorporating ideas from various disciplines to create art is not a new phenomenon. Walter Issacson in the article, The Science Behind Mona Lisa’s Smile illustrates how Leonardo da Vinci incorporated arts, science, optics, and illusion to create some of his most famous paintings including the Mona Lisa. He “dissected human faces” to study muscular movement over bones to authentically delineate the facial expressions on his models. He used optics to manipulate the details of the shading around the smile of Mona Lisa. He used Chemistry to create pigments and glazes that would reflect light and converted a one-dimensional portrait to be the first manifestation of “augmented reality”. (See video) Nowadays, you can use filters that are available online on social media and other applications, to create customized effects and manipulate photos and art. You can use Snapchat’s Starry Night filter to create your own starry night photo.
With the numerous resources and disciplines that are now available at the artist’s disposal, AI can be used to enrich art. However, algorithmic art is not without its limitations or biases. While it is fairly easy to ingest datasets and apply machine learning algorithms to generate AI art, creating original artwork is still a challenge. Ultimately, AI art is created through the human curation of data, and inevitably certain ideas will be under- or over-represented.
Resources to create your one AI art
- AI Art generators: https://aiartists.org/ai-generated-art-tools
- Deep Dream generator: https://deepdreamgenerator.com/
- Repaint your picture in the style of your favorite artist: https://deepart.io/
- See your photos turned into artwork: https://www.instapainting.com/ai-painter
- Open source tools to create AI art: https://www.wired.com/story/we-made-artificial-intelligence-art-so-can-you/
- Train a computer to recognize your own images, sounds, & poses
7. Introduction to generated art: https://www.freecodecamp.org/news/an-introduction-to-generative-art-what-it-is-and-how-you-make-it-b0b363b50a70/