The Curious Case of AI Art
AI has already started to reshape entire industries. Tech giants like IBM and Microsoft are banking on it big.
It’s easy to imagine the day when the lower rung of white collar workers is replaced by AI systems. Code doesn’t ask for benefits, and it doesn’t spend productive time scrolling its Instagram.
But we’re used to technology performing mechanical tasks faster, more accurately, and on a greater scale than humans.
What machines can’t do is think and create, right? Software may be able to run dry calculations better than us. But surely, only humans can exercise aesthetic judgement and produce great works of art.
An AI-powered bot called “Botnik” recently wrote a Harry Potter book. That’s right – an entire book, inspired by many of J.K Rowling’s “human” writing style.
And while it spewed out some hilariously ridiculous robot sentences (“He saw Harry and immediately began to eat Hermione’s family.”), the book was deemed readable by many Potter fans. That’s a pretty amazing feat.
Another AI application called FlowComposer can now generate songs automatically, with minimal human input. It recently wrote and produced a song called “Daddy’s Car” – a passable attempt at a “Beatles-like” song (but it doesn’t make the Sgt. Pepper cut by a longshot …). Listen here:
Legally speaking, it’s far from clear that the works generated by Botnik or FlowComposer-type apps qualify for copyright protection. Do they emanate from “skill and judgement”, or the “original intellectual conceptions of the author”, ? Do they have an “author” in any traditional sense?
Some court decisions – not all of them recent – give us some hints on how these issues might play out in the real world.
In Goldstein v. California, the US Supreme Court reasoned that “in order for a computer to generate any kind of artistic work [it] would require significant input from an author or user.”, “[B]ut when the user does very little and most of the output is left up to the computer, then it is less likely that the user may own the copyright in the output.”
The UK took the opposite path, where copyright in computer-generated works is owned by “the person by whom the arrangements necessary for the creation of the work are undertaken.”
Courts in South Africa and Australia have taken a more sophisticated approach, distinguishing “computer-generated” works – for which there is no “human author” from “computer-assisted” works – in which case a human author simply received assistance from a computer. Copyright would vest in the author in the latter case, whereas in the former, no copyright exists.
Initially, it may seem like AI-produced works not being protected by copyright is a victory for human creators.
More works in the public domain we can freely use. We have copyright, they don’t. Bust out the champagne. Right?
It’s highly probable that AI art will get way better than Botnik’s Potter book or Daddy’s Car in the very near future.
Will it ever get predictably better than most commercially viable, human-produced art? If it does reach that point – a kind of “art singularity” of sorts – why would publishers, labels, and studios ever deal with humans and their pain in the neck copyrights at all? A victory for the public domain in the short run could turn into a nightmare for creative labor markets in a hurry.
For the time being, we can just wait and pretend that this point in time is far away. We can safely hang back and make fun of AI bots and their adorably naïve attempts to produce works of art and entertainment worth purchasing.
But one day, the bots might become our entertainers.
At that point, for copyright to survive, a UK-like approach may be necessary in all countries. Maybe that leads to a whole new generation of amazing, AI-assisted art. Maybe it leads to a technological arms race in creativity and squeezes out artists and their outdated methods. Or maybe this is all a bad dream, and we’re underrating the complexity of human genius, which will always prevail and create the best stuff we connect with in the end.
One thing is certain. AI will make sure we answer these questions. Fast.