Back to blog

AI and art: is there a pattern to creativity?

by AprilSix Proof

The past couple of weeks I have started my working day scouring the Internet for breakthroughs in a wide range of scientific sectors. Normally, any one of these sectors will be teeming with activity while others are relatively quiet. There is one exception: every day seems to bring stunning new advances in AI and its applications. From biotechnology to dispute mediation and self-driving buses: across disciplines, the development of AI is a Kuhnian scientific revolution.

If there is a limit to AI’s ambitious applications, it’s creativity, which continues to defy AI’s immense grasp. But even this assertion comes with caveats. An AI-created painting entitled ‘Portrait of Edmond de Belamy’ has been sold for $432,000. AI systems have been writing pop songs since 2016. In AlphaGo’s historic Go match victory over Lee Sedol, the tide-changing 37th move was immediately described as “very strange” and “a mistake” by the two commentators. “It’s not a human move. I’ve never seen a human play that move,” Fan Hui, three-time European champion, later added. “So beautiful… so beautiful.”

These examples show that AI is already capable of some form of creativity, perhaps due to an important paradigm shift that took place decades ago. Originally, AI systems ‘learned’ new information through brute force operations, supported by an increase in computing power and massive data sets. This power-based approach later evolved to a knowledge-based one, where AI systems adopt a diverse range of problem-solving schemata and learning techniques.

Still, many of AI’s successes come from a specific branch of machine learning called deep reinforcement learning, a technique combining pattern recognition with a reward system, like scoring points in a game. This approach, first realised in 2013, is what propelled DeepMind to its current status as a global thought leader. For all its brilliance, it is still memorisation-based learning. Anything novel the AI system divines can easily be seen as extrapolation of its inputs, not true creativity.

It’s easy to jump to the conclusion that AI will never possess anything approaching human creativity. Jason Toy, CEO at machine learning consultancy Somatic said, “I don’t think it’s going to be possible for quite a while to take what humans think of as beautiful and creative and put that into an algorithm.” A statement from IBM’s John Smith, manager of Multimedia and Vision, is similarly pessimistic: “Deep learning isn’t the answer to creativity, as we still have to define what creativity means. We do know some of the attributes have to do with finding something novel, unexpected, and useful.”

Still, AI definitely has a role to play in the creative process. Collaboration between artists and AI is likely to lead to a different perspective on humanity and a new avenue of artistic thought. As for autonomous AI art, it is not hard to imagine a successful AI pop band pop up (forgive the pun) in future, given the genre’s penchant for repetitive patterns. Some visual artists, too, have found success in taking a pattern and showcasing its aesthetic quality.


Piet Mondrian’s Composition A is considered a masterpiece; it is a simple pattern of rectangles in primary colours (Wiki commons)


We are still at the dawn of the AI age. John McCarthy coined the discipline’s name a mere six decades ago, and we are only starting to see its application in society now. To already put limits on what AI will be able to do in future seems folly. Still, true creativity may require an entirely different approach. A lot of great art is great not because it follows patterns, but precisely because it does not.

Written over 400 years ago, the poem Virtue by George Herbert follows a very careful structure in each stanza, an iambic tetrameter, before breaking its rhythm, rhyme, and tone in the final line. Over three stanzas, we get used to this disruption, a pattern in itself, only for Herbert to pull the rug from underneath us again, ending a dark (morbid even) poem on a thoroughly different note. Whether AI will ever master such pattern subversion remains to be seen but its prospect, just like the poem’s last line, teems with excitement for the future.

Sweet day, so cool, so calm, so bright,
The bridal of the earth and sky;
The dew shall weep thy fall to-night,
For thou must die.

Sweet rose, whose hue angry and brave
Bids the rash gazer wipe his eye;
Thy root is ever in its grave,
And thou must die.

Sweet spring, full of sweet days and roses,
A box where sweets compacted lie;
My music shows ye have your closes,
And all must die.

Only a sweet and virtuous soul,
Like season’d timber, never gives;
But though the whole world turn to coal,
Then chiefly lives.

Blog post by Louis Van Der Linden