Making music is one out of very few endeavors that require human creativity – though it might not remain that way for long as artificial intelligence is starting to catch up on music composition as well. In fact, one of the most recent instances of AI delving into creating music happened just last December 2017 when a piece of software with machine learning capabilities had put out a black metal album. Given that the said metal subgenre typically relies on its human creators to convey a dark, atmospheric sound, it’s rather surprising that an AI program is showing potential to do the same. But regardless of genre, here’s how artificial intelligence is changing the future of music as we know it.
How Does Artificial Intelligence Create Music Anyway?
While not as prevalent at the moment as its various other applications in the real world, artificial intelligence has been crossing paths with music for quite a while, the most notable instance being when former University of California, Santa Cruz professor David Cope had developed a program called “Emmy.” The said AI program was able to put out a whopping 11,000 compositions patterned after the style of German Baroque composer Johann Sebastian Bach.
So how did “Emmy” as well as the AI program above from three decades later create music? Since they aren’t capable of sitting down and writing a bunch of notes on paper, here’s how they did it:
1.) The creator of the AI program feeds it with existing pieces of music as performed by humans.
Contrary to most people’s perception of artificial intelligence, an AI program can’t simply go about writing music all by itself. Its creator would first have to input existing musical creations as performed by humans for it to learn. Come to think of it, an AI program’s creator feeding it with human-recorded music isn’t that different at all from teaching someone how to play a musical instrument by letting them practice using the same music.
2.) The AI program would then deconstruct the music that was given to it by its creator.
For an AI program to start getting into music composition, it would have to break down the human-recorded music that its creator had given to it into bite-sized chunks of audio. Otherwise, it won’t be able to generate any newly composed music at all.
3.) The AI program would pick which pieces of deconstructed audio it wants to use and then put them back together to make an entirely new composition.
Every AI program has a neural network patterned after the human brain. The said series of algorithms would attempt to decide which two audio segments would blend as seamlessly as possible. Each decision that the music-generating AI program makes is purely guesswork, so mistakes are bound to pop up here and there until it comes up with better guesses that eventually lead to musical compositions that sound as if humans had made them.
The three-step process above might sound deceptively simple at first. But previous attempts at making AI-generated music have proven to be hit or miss. Even the creators of the AI program that had put out a black metal album had struggled with replicating the said metal subgenre at first as it had initially come up with layers of textured noise. Other than obstacles like that though, it won’t be long before AI programs start sharing space with musicians both in the recording and performing circuits. However, there’s no need to panic over music-generating AI programs. After all, even with artificial intelligence delving into music composition, it can’t replicate the heart and soul that a human composer puts into their work.
You May Also Like: