Art and AI is a new series of articles that will focus on the intersection of artificial intelligence and the art world.
Artificial Intelligence (AI) - is the branch of computer sciences that emphasizes the development of intelligent machines, thinking and working like humans. For example, speech recognition, problem-solving, learning and planning.
Deep Learning - Deep learning (also known as deep structured learning or hierarchical learning) is part of a broader family of machine learning methods based on artificial neural networks. Learning can be supervised, semi-supervised or unsupervised.
Neural Networks - a beautiful biologically-inspired programming paradigm which enables a computer to learn from observational data
I first heard the term deep learning several years ago at the NVIDIA Graphics Technology Conference. Their amazing CEO Jensen Huang, made mention of “breakthroughs in algorithms (math) that led to new methods of analyzing massive amounts of data”. Since then I’ve been reading and researching the subject and watching deep learning affect nearly every aspect of computer learning (artificial intelligence).
One area where deep learning has had an impact is in the are of creating and analyzing art. You’ve probably noticed a lot of attention paid to apps that turn your photo into a stylized painting or deep learning structures that can create original artworks. Another interesting experiment is by programmer Mario Lingman who developed a set of neural networks that use a database of thousands of 17th to 19th-century portraits to create new, original portraits in real-time. His exhibition titled “Memories of Passersby I”. Some of the works created by the AI were sold at auction by Sotheby’s as “the first self-contained, generative work of AI ever to appear on the market”.
What’s odd about the portraits created by the AI in this video is that they look pretty strange compared to the paintings used in the database. There is a lot of the uncanny valley effect. I think it’s because viewers turn off their critical sense due to the novelty: a machine trying to create antique portraits.
A more interesting idea seems to me to be that these bad portraits are actually mistakes. My question then would be why the programmer who created this AI doesn’t use the mistakes as ways the AI can learn, like a human artist. Mistakes and growing from then are what makes us human. The AI doesn’t make mistakes they only respond to the data that they have been given. So, in effect, the real creativity is coming from the person who programmed the AI and not the AI itself.
However, in an article in the American Scientist by AHMED ELGAMMAL he makes the case that the programmers who create art-making AI’s are focused on “novelty” as a form of creativity. He says we shouldn’t focus on the individual art pieces but on the process of collaboration between the AI and the artist who built the neural networks (the AI brain). Ahmed also sees this as a form of “conceptual art” that dates back to the 1960s.
At present, artists have no need to worry about neural networks replacing them. I tend to side with Ahmed in that we are in the novelty phase of art-making machines. There are those who will find the work interesting and those who will find it boring. What matters is how it affects those artists/programmers who might go on to create new ways of artists and machines collaborating to create startling new works. But that is in the future at this point.