Creativity is one of the greatest mysteries of the human experience. Artists and scientists alike have long wondered, where does it come from and how does it work? Creativity is also one of the few things we can point to that makes us unique as humans. Which is why it's so thrilling, and terrifying, to think that artificial intelligence could become creative – if it hasn't already.
The myths of creativity
The ancient theory was that creativity came from a higher power than humanity: that it was bestowed upon our puny human consciousness from some kind of infinite spirit, or universal force (known to most religions as their flavor of God) Dr. Larry Dossey explains this transcendentalist position was held by scientists such as Erwin Schrödinger, David Bohm, Ervin Laszlo, and artists like Piet Mondrian and Paul Klee.
Writer Elizabeth Gilbert touches on this theory in her 2009 TED Talk, "Your elusive creative genius." She explains that in ancient Greece and Rome "genius" did not live inside of individual people, but was a spirit from the gods that lived in the walls of an artist's studio. The Renaissance flipped this idea on its head and centered people as creative geniuses (your Michelangelos, da Vincis, and other great men of history). She thinks that this is part of what puts pressure on modern artists that drives them to self-destruction, and questions whether we should go back to some form of this ancient understanding of it as an external force when we investigate the source of creativity.
It's a popular theory today that creativity comes out of problem solving and a need to defy known rules or conventions. This definition implies that creativity is somehow diminished when seen as a leisure activity, or something that is only used to produce art. In this theory, creativity is credited with all great breakthroughs in science, medicine, and technology.
"How does a grapefruit-sized heap of meat crackling with electricity conceive of mathematical theorems, create beautiful art, discover the laws of nature, invent kitesurfing, and design buildings that look like sea shells?" – Arne Dietrich, "Where does 'creativity' happen in your brain?"
Arne Dietrich's work on the cognitive neuroscience of creativity defines creativity as "The ability to produce work that is both novel (i.e., original, unexpected) and appropriate (i.e., useful, adaptive concerning task constraints)...Creativity is the epitome of cognitive flexibility. The ability to break conventional or obvious patterns of thinking, adopt new and/or higher order rules, and think conceptually and abstractly is at the heart of any theory of creativity"
Dietrich and many, many other teams of neuroscientists have tried to pin down the location of creativity in the brain, but haven't found a single spot that lights up when having creative ideas, or that you could apply electrodes to and stimulate to produce creative genius. Creativity uses almost every part of the brain.
All of this to say that we still don't fully understand creativity at all, but that is part of why we think it makes us special. Some are afraid that creativity may be the only thing standing between us and robots taking our jobs.
Artificial intelligence developing creativity is often one of the first signs in science fiction that we are truly fucked.
Painting robots, fiction and facts
In this iconic scene from I, Robot, Will Smith's detective Del Spooner is interrogating a robot that he believes has murdered its creator. He asserts that robots don't have emotions, and are therefore incapable of producing great works of art. The robot asks, "Can you?"
It's a great question, because it makes us wonder what standard we should hold artificial intelligence to when attempting to define its capacity for creativity. There are plenty of bad (human) artists in the world, why should we expect robot artists to be (evil) geniuses?
Most people have heard of the Turing Test, but not the Lovelace Test. It's named after Ada Lovelace, the original programmer of Charles Babbage's "Analytical Engine," the theoretical model of a functional computer which was conceived 100 years before the Turing machine.
Lovelace famously said, "The Analytical Engine has no pretensions whatever to originate anything. It can do [only] whatever we know how to order it to perform," meaning that it is not possible for computers to have a creative output.
The first Lovelace test was outlined in a 2001 paper entitled Creativity, the Turing Test, and the (Better) Lovelace Test, and was revised in 2014 as the Lovelace 2.0 Test.
However, before these tests were defined, in 1994, professor of cognitive science Margaret A. Boden, wrote in the introduction to the book, Artificial Intelligence and Creativity: An Interdisciplinary Approach, a series of questions she called "Lovelace-questions."
The Lovelace-questions are as follows:
- "whether computational concepts can help us understand how human creativity is possible"
- "whether computers (now or in the future) could ever do things which at least appear to be creative"
- "whether a computer could ever appear to recognize creativity"
- "whether computers themselves could ever really be creative (as opposed to merely producing apparently creative performance whose originality is wholly due to the human programmer)."
And all of the answers to these questions appear to be yes.
In her later (1998) paper, "Creativity and artificial intelligence," she explains artificial intelligence can create new ideas in three ways: "by producing novel combinations of familiar ideas; by exploring the potential of conceptual spaces; and by making transformations that enable to the generation of previously impossible ideas... The ultimate vindication of AI-creativity would be a program that generated novel ideas which initially perplexed or even repelled us, but which was able to persuade us that they were indeed valuable. We are a very long way from that."
Of course, back in 1998 it may have seemed like we were a long way from a lot of technology we take for granted in the 21st century. 1998 was the year Google was founded, most people were accessing Usenet via Netscape or Internet Explorer on their Windows 98 PCs.
Boden has since acknowledged, in a 2015 View for the MIT Technology Review, that even though artificial intelligence is capable of creating great abstract works of art, the lack of cultural contextual knowledge that it has is its greatest weakness. I.e., AI can make good art now, but it can't yet convince us that it's good.
Matthew Putman, CEO of Nanotronics, has expressed a similar opinion in his essay, "Artificial Objectivity," that robots can't be great artists because they can't appraise their own work: "For art, objectively good is distinguished from subjectivity only by universal human values."
Luckily, the responsibility for creation and appraisal of art have never fallen to the single human artist anyway... well, maybe one:
"You know it's ART, when the check clears" - Andy Warhol
The Lovelace and Warhol definitions of art rely on the human critic being convinced - in the Lovelace test they must be convinced that the programming is not entirely responsible for the output and that the output is satisfactory, and for the Warhol test they have to be willing to buy it.
So... bring on the robot art fairs!
AARON is a painting robot that, according to the MIT Technology Review, had been collaborating with artist Harold Cohen since 1973. As early as the 1980s, Cohen was quoted saying he was "the only artist who would ever be able to have a posthumous exhibition of new works created entirely after his own death." Cohen passed in April of 2016, and, sadly, AARON has not continued to make new work, but this creative collaboration between man and machine was one of the first instances of robot paintings shown in galleries (although shows featuring work by both Cohen and AARON are still listed under "One-person exhibitions" in Cohen's biography).
The Review argues that AARON was a true artist working under Cohen in the lineage of Renaissance painters, whose works would often be executed by teams of apprentices and copyists, but still credited to the master. However, Cohen himself has been quoted saying that a robot would have to develop a sense of self in order to become creative in the same sense that humans are creative.
One of the things that was missing in AARON's painting process was visual feedback - a capability that many of today's artificially intelligent painting robots have.
Pindar Van Arman's CloudPainter is equipped with "a custom 3D printed paint head, two robotics arms, deep learning, artificial intelligence, and computational creativity to compose its own original artwork." VICE Video's coverage of his robots show CloudPainter using style transfer, combining photographs and painting styles to create new portraits.
For those not yet convinced that these robots, having mastered the craft of painting, are truly artists, because they have no internal experience of their art, or have only succeeded in competition against other artificial artists, here are two more examples:
Simon Colton's painting robot called The Painting Fool can read emotions from photographs and use several different painting styles to convey that emotion in its portraiture. It can also read, and has used keywords from the news to create a collage about war. It even has a setting in which, when overwhelmed by too many negative keywords, it will refuse to paint.
A General Adversarial Network (or GAN) trained by the Art & AI Laboratory at Rutgers University, trained on over 80,000 paintings, created abstract paintings that were indistinguishable from humans'. When placed alongside images of real Abstract Expressionist paintings and work from Art Basel 2016, critics had to answer how the paintings made them feel, inspired, or if they found the paintings complex or novel (and remember, novelty is one of the keys of creativity!) The results: "Not only could the human evaluators not tell which images were AI-created, in many cases they rated the AI’s artwork higher than the humans’."
How fucked are we?
Popular opinion seems to be that yes, robots with artificial intelligence are creative, but only as collaborators. So we return once again to the question of what creativity is – must it come from a singular, creative genius?
Psychology Today says, "This Makes Us Human": the ability to blend knowledge, taking old stories and new ideas, as the root of creativity. Augustin Fuentes, author of the book The Creative Spark: How Imagination Made Humans Exceptional, equates wisdom and creativity, and uses evolutionary psychology to back up his theory that creative collaboration is what made human society possible today. Else how could "fangless, clawless, hornless, naked upright primates" accomplish so much?
And perhaps creative collaboration is what will help evolve artificial intelligence as well. We shouldn't worry about AI painters demanding attribution (much less taking over the world) until we have I, Robot levels of autonomous artificial intelligence running amok.
For now, our fragile human egos are safe, if we believe that artificial intelligence is a collaborative force; but that actually makes them just like us already. If we believe that the cult of the genius has passed, and that collaborative, not individual, creativity is the true measure of human intelligence, then AI has already more than sufficiently filled that role.
Or if we return to the theory that creativity comes from some kind of divine intervention, then from the perspective of AI, we are the gods.