This article is part of Demystifying AI, a series of posts that (try to) disambiguate the jargon and myths surrounding AI.
The robots are coming for your job—that’s the general perception of where artificial intelligence is headed today. While that is (in my opinion—and that of many others) an overblown statement for the moment, it’s true that AI algorithms are transforming the employment landscape and changing the way we’re performing tasks.
However, some domains are considered of purely human nature and less threatened by AI disruption. Naturally, creative arts, music and painting are among the first candidates that come to mind. And to some extent, this assessment is true. Artificial intelligence has a ways to go before it yields the next Mozart, Bach or Da Vinci.
But AI will affect artistic endeavors in its own ways, just as every technological breakthrough has. There are plenty of companies that—PR BS aside—are applying AI in the creation of works of art. While true artists shouldn’t fear about robots taking their jobs, here’s what you need to know about how this often-overhyped technology will affect the artistic landscape.
How AI performs tasks
The main technology behind recent applications of AI in different domains is machine learning. Before machine learning, most software were created by specifying a set of rules that defined the program’s behavior. In contrast, machine learning algorithms are given large sets of relevant data and are told to figure out the rules for themselves. This approach helped automate tasks that could not be defined with rigid rulesets.
Image classification, face and voice recognition, transcription and translation, and even more complicated tasks such as medical diagnosis and job application assessment are some of the fields where machine learning has made serious progress.
However, while AI has proven to be much faster and more efficient than humans in the tasks it performs, what we currently have is narrow artificial intelligence, AI that can perform a single or limited set of tasks. You can’t expect a recommendation algorithm to start scheduling meetings or make traveling decisions. However, juggling between tasks that are seemingly unrelated is what humans excel at, and what gives us the edge over AI.
Until scientists create technologies that can mimic human intuition and decision-making, AI algorithms will only be capable of performing routine tasks, both cognitive and physical. Even performing seemingly trivial tasks such as navigating a restaurant or a warehouse are feats that current AI technology struggles with.
How machine learning affects artistic creations
At first thought, we think of music and arts production to be works of pure imagination, a spark in the mind that cannot be quantified or described in terms of data. But in many cases, what our imagination creates is the reordering of things that we already know, or in AI terms, the reorganization of existing data in different patterns.
This is model of work that machine learning is especially good at. Earlier this year, I had the chance to examine a deep learning algorithm that created Irish folkloric music. Researchers at Kingston University and Queen Mary University of London had trained the algorithm with a repertoire of 23,000 music transcripts, and had then let it loose to create its own unique tunes.
The algorithm doesn’t have an understanding of music. Neither does it have an ear to define what good music sounds like. What it does is it examines all of the music transcripts and tries to find common patterns that define professionally composed Irish pieces. It then tries to apply the same pattern to create sequences of notes that don’t exist in the repertoire. It might be much different from how a human composes music, but the results were stunningly convincing.
I spoke to Irish musician Daren Banarsë, and he confirmed that the tunes had the Irish feel to them, though most needed reworking or tweaking to sound convincing. He also pointed out that the algorithm’s mistakes sometimes ended up yielding some interesting tunes.
However, both Banarsë and the creators of the algorithm believed it would complement the efforts of human composers rather than replace them. “As a composer, I have no worries about my job being replaced in the near future,” Banarsë told me. “But as the technology progresses, who knows? I hope there’ll be a stage before that, where the computer can assist me in my job. I always find it daunting when I have to start a large scale composition. Maybe I could give the computer a few parameters: the number of players, the mood, even the names of some of my favourite composers, and it could generate a basic structure for me. I wouldn’t expect it to work out of the box, but it would be a starting point.”
Big companies such as Google and Sony have their own independent research units that are working on the intersection of artificial intelligence and arts. And a number of startups are offering services that create music using AI.
While it’s fair to say that none of these efforts are a threat to the careers of Ramin Djawadi or Hans Zimmer, they do provide some interesting opportunities. For instance, Jukedeck, a music generation website, uses AI algorithms to generate tunes. Users have to specify and tweak a few parameters such as genre, mood and tempo and the system provides a unique music. This can be a low-cost solution for people who want to add background music to YouTube videos, corporate presentation and games without violating copyright laws.
AI Music, a British startup, uses AI to make smart adjustments to music, such as syncing the tempo with the movements of the video it’s being played on.
When it comes to visual arts, things get a bit more complicated. As it happens, AI is much more adept at understanding the context of images than creating images on its own.
In 2015, Google researchers developed DeepDream, an AI algorithm that uses deep learning to generate images. While the images look more like hallucinations rather than works of art, it’s interesting to see how AI perceives the world.
Where AI really shines is in helping humans artists. One of the interesting projects is Google’s AutoDraw, an application that helps turn rough sketches into rich drawings. The name is inspired from autocomplete, the feature that is found in most typing applications these days and tries to guess what word you’re typing. When you use AutoDraw’s tools to draw, the applications tries to guess what you’re trying to draw and provides you with the option to replace it with a fully drawn version of the object.
Other applications of AI include auto-coloring objects in pictures, turning brush strokes into drawings, or helping with the smart placement and arrangement of objects in a poster.
It’s obvious that for the moment, artificial intelligence and machine learning aren’t yet ready to independently create works of art. But they are helping us better understand how we create art. And they’re proving to be good sidekicks to artists and an efficient tool to make artistic work more accessible for everyone.