Artificial Art: Can human creativity be computerised?
Art communities have long tried to distance themselves from more commercial driven industries. So you can bet that when the narrative of ‘robots taking jobs’ in such industries started to gain traction and validation, there were a few artists and creative folks out there quietly chuckling at the thought. That was, until they realised that they too were in AI’s path of domination.
In fact, a recent collaboration between economic modelling firm AlphaBeta and the ABC saw the development of the ‘Could a robot do your job?’ calculator, of which determined Art professionals to be in the top half of Australian workers whose jobs are most susceptible to automation - on par with factory line workers and keyboard operators. But can art and more importantly human creativity really be broken down to an algorithm able to be taught to a computer?
The idea of constructing a program or computer capable of ‘human – level’ creativity isn’t so wild when we look at the fact we have already developed machine learning algorithms to understand content, for example speech recognition technology. Disney Research has taken it a step further and recently developed AI that is capable of evaluating short stories written by humans to determine their quality and popularity.
"The ability to predict narrative quality impacts on both story creation and story understanding," said Markus Gross, vice president at Disney Research regarding the research. "To evaluate quality, the AI needs some level of understanding of the text. And if AIs are to create narratives, they need to be able to judge the quality of what they are producing."
The type of machine learning required for creative output is pretty unique. It uses what is known as a ‘generative model’ which means that it learns to mimic the data that it has been trained on. But, you could potentially train AI to not only replicate a piece of art, but to use its knowledge to create something unique within certain parameters. CEO of machine learning startup Somatic Jason Toy explains the process to IBM in a recent blog post “If you feed it thousands of paintings and pictures, all of a sudden you have this mathematical system where you can tweak the parameters or the vectors and get brand new creative things similar to what it was trained on.”
Think of it like learning to play a guitar. You can learn the 4 chords required to play Wonderwall at your next house party, but when the angry crowd lays down some new parameters by demanding you to play something else, at the very least you will be able to rearrange these chords to play a new song.
A better example may be Google’s Magenta project. According to the lead researcher Douglas Eck, the project’s main goal is to develop algorithms that can learn how to generate art and music, potentially creating compelling and artistic content on their own. Google premiered Magenta’s first attempt at laying down a banger at Moogfest in 2016, and while it’s probably not going to be making its way into the charts just yet, it does show that their AI has the ability to create something unique based on whatever knowledge of musical composition it has been trained on.
Nonetheless, there is still the big question amongst AI experts that while we may be able to develop algorithms capable of learning the parameters of art, can AI truly develop its own sense of creativity? The fault here may not lie with the robots, but in our own inability to define exactly what creativity is.
Director of IBM Research Arvind Krishna distinguishes teaching AI to be creative from a function such as diagnosing cancer, stating “I think teaching AI what’s melodic or beautiful is a challenge of a different kind since it is more subjective, but likely can be achieved. You can give AI a bunch of training data that says, ‘I consider this beautiful. I don’t consider this beautiful.’ And even though the concept of beauty may differ among humans, I believe the computer will be able to find a good range. Now, if you ask it to create something beautiful from scratch, I think that’s certainly a more distant and challenging frontier.”
So rather than attempting to replicate something we don’t really fully understand, what if AI could be used to help us understand more about our own creativity?
According to the Association for Computational Creativity, two of the main goals of art-inspired AI should be:
• To better understand human creativity and to formulate an algorithmic perspective on creative behaviour in humans
• To design programs that can enhance human creativity without necessarily being creative themselves
A lot of AI experts tend to agree with this positioning. In an interview with ABC, Dave King, founder of a creative AI company Move 37 states "One of the most interesting aspects of creativity is that ability to combine ideas or to draw things together… If you have an algorithm that is working for you in the way you want it to it can source and discover lots and lots of different things."
For example, American musician Taryn Southern recently released an album composed and produced entirely by an AI program called Amper. Much like Magenta, Amper learns from analysing existing music and how the elements correlate with different genres and moods to then create new compositions to match these patterns. However it was Southern’s role to determine the parameters of the algorithm. Essentially giving it all the material it needed to develop something original and useful to her.
Southern made clear her views on AI in creative industries at the 2017 TNW Conference, stating “I’m an artist, so I want tools that augment and fortify my creativity… I don’t want anybody telling me I can’t play the game anymore because an AI can do it better.”
So AI may not completely replace artists and render creative industries inaccessible to humans in 30 years’ time after all. Increasing automation may actually make these industries more accessible by inspiring us to seek and develop new mediums, styles and genres alike. I for one can’t wait to see what Magenta and Amper feature on next.