“I’ve seen floating boulders, cube-shaped hills and rock formations that worm their way across the landscape like snakes many kilometres long. I’ve found plants made of platinum, eagle-sized butterflies with pterosaur wings and sausage-bodied dogs with too many legs.” Douglas Heaven, New Scientist
Can art be created by computers?
Is there such a thing as a chaotic pattern?
What does it look like when a new universe is generated by algorithm?
If you’ve ever thought of mathematics as dull, you might be uninspired by these questions. You may consider algorithms to be, by their nature, repetitive, predictable, and lustreless. Or, you may have heard of the renowned beauty of the computer game No Man’s Sky, one of the first games to create procedurally generated worlds on a grand scale– a whole 18 quintillion of them. This means that instead of the game’s environment being hand-crafted by a team of designers, the design constraints for landscape, flora and fauna were instead loaded into a computer and thus generated without a human eye upon them – a kind of pick-and-mix for planets.
Using chance to pick these permutations means players can truly be discovering new worlds for the first time as they play. Some report the breathtaking feeling of stumbling upon a biological oddity or a remarkable vista and the need to photograph and document it for others. This is space exploration via the awe-inspiring mightiness of mathematics, and it simply would not be possible otherwise – it has been estimated that the universe in No Man’s Sky would take 5 billion years for one person to explore (never mind create). The developers decided during the project to make the change from 32-bit numbers to 64-bit, which meant the number of planets changed from 4.2 billion (232) to roughly 18446744073709551616 (264).
There is a long history of using algorithms to make music (Marcus du Sautoy is fascinating on this subject). Creativity is clearly more than just chaotic imagination – it takes order, rhythm and rules, too. But the question of whether machines can create art is one that humans may often find hard to digest: consider, for example, the computer which can turn an image into a Van Gogh using a ‘convolutional neural network’, or the work of this doodling robot. Is the fact that they are programmed to perform a limited number of actions reason to discount their creativity? When we make or draw a fractal, are our imperfect ways of following the algorithm producing iterations that are more or less beautiful than a machine’s would be?
Because permutations can be used by humans, too – and to beautiful effect. This time, instead of using an algorithm, a French poet named Raymond Queneau wrote down an innocuous book of ten sonnets – but, unusually, each line was on a flap and each part could be combined with any other. How many unique poems do you think this made?(I’ll give you a clue – a sonnet has fourteen lines).
Generating art using a repeated procedure is, of course, not something new – but the scale of the digital means we have the access, as well as the ability to programme AI to start to make more ‘human’ decisions, meaning that projects like Google’s Magenta are pushing artificial creativity to its limits. Its designers say their greatest challenge is stimulating curiosity:
“It’s not enough just to sample images or sequences from some learned distribution. Art is dynamic! Artists and musicians draw our attention to one thing at the expense of another. They change their story over time—is any Beatles album exactly like another?—and there’s always some element of surprise at play. How do we capture effects like attention and surprise in a machine learning model?”
