Artificial Intelligence, or AI, is being used everyday for a wide variety of uses online, so it’s no surprise that music is being generated using it. While it might not win any awards, in most cases it’s at least passable. But music is mathematical by nature, which makes it a good candidate for AI creation. Lyrics, on the other hand, are a different kettle of fish that AI still can’t get its arms around.
The whole idea of AI is that it learns by repetition. On a streaming platform, if you keep on selecting songs by a particular artist or a certain genre of music, it learns your choices and then begins to feed you similar music. With music creation, if you feed it examples of a type of music, it learns from that and uses it as a basis for creation something similar.
Packt Publishing via a post on Hacker Noon recently released a step-by-step guide showing how to build an AI neural network that can be used to generate new lyrics supposedly in the style of any artist. The larger the training dataset, the better the results, so the authors used a file of lyrics from 10,000 songs, which seems like plenty. What it spit out wasn’t exactly Bernie Taupin, but it was funny.
Yeah, oho once upon a time, on ir intasd
I got monk that wear your good
So heard me down in my clipp
Cure me out brick
Coway got baby,
I wanna sheart in faic
I could sink awlrook and heart your all feeling in the firing of to the still hild, gavelly mind, have before you, their lead
Oh, oh shor,s sheld be you und make
Oh, fseh where sufl gone for the runtome
Weaaabe the ligavus I feed themust of hear
A Medium writer used an open source network called textgenrnn to generate lyrics like Taylor Swift, and got this back:
i ‘ m not your friends
and it rains when you ‘ re not speaking
but you think tim mcgraw
and i ‘ m pacing down
i ‘ m comfortable
i ‘ m not a storm in mind
you ‘ re not speaking
and i ‘ m not a saint
and i ‘ m standin ‘ t know you ‘ re
i ‘ m wonderstruck
and you ‘ re gay
The thing about AI neural networks is that the more they’re used, the smarter they get. That means that while today’s lyrical output might not yet be radio ready (or maybe it is?), after a million more inputs next ear it might be turning out music’s equivalent of Shakespeare. In the meantime, it’s good for a laugh. (There’s more in this article).