Automating Art: Gilbert Simondon and the Possibility of Independently Creative Machines

Abstract

The modern concept of creativity as an attribute of human beings has, since its very beginnings in the 18 Century, routinely been defined in opposition to that of the programme. In Edward Young’s famous letter “Conjectures on Original Composition” from 1759, the original creativity of the genius is radically distinguished from anything that can be accomplished through the execution of a plan. “An original,” he writes, “may be said to be of a vegetable nature; it rises spontaneously from the vital root of genius; it grows, it is not made.” The conviction that true creativity is a spontaneous and intuitive process, which cannot be attained through following a set of predetermined rules and procedures still endures in the popular imagination today. It is the principle underlying most of the common objections to the idea that machines might one day be considered independently creative. For if a computer can only follow a sequence of operations that it has been programmed to perform, so the argument goes, then creative invention must – by definition – be beyond their capabilities. Ada Lovelace is often credited with being the first to formulate this objection as long ago as 1843. As she famously wrote, with respect to Charles Babbage’s theoretical computer the Analytical Engine, “[the machine] has no pretensions whatever to originate anything. It can do only whatever we know how to order it to perform.” Those arguing against this prevailing doxa and in favour of the possibility of fully autonomous, independently creative machines typically advance a cognitivist view of creative behaviour that is more functional and less romantic than that invoked by Edward Young. The standard claim is that once we have expunged all mystificatory remnants of innate talent, divine inspiration, or individual genius, what we are left with is a psychological process that can be described in formal computational terms and – theoretically at least – implemented in a real machine. This debate recurs frequently, most recently as a result of a series of high-profile machine learning projects aimed at recognising and reproducing human creative processes in the arts. Notable examples of this trend include Google’s major Magenta project launched in 2016 to “explore the role of machine learning in the processes of creating art and music,” and the research team at the University of Tübingen who trained a neural network to recreate photographic images in the signature style of canonical artists such as Van Gogh or Edvard Munch. Possibly the most well-known single example was the exhibition in 2016 of a “new” Rembrandt painting at the Cannes Lions International Festival of Creativity. The painting was the striking result of an ambitious digital venture entitled “The Next Rembrandt,” devised by the Dutch