Call it “Ye.I.” A high school student in rural West Virginia has used Kanye West’s lyrics to teach a neural network how to rap. And even though it raps about as awkwardly as you could expect a computer to rap, this 17-year-old’s creation yields some remarkably coherent, clever lines like, “Stop running up my money tall, height advantage.” Other lines feel a bit more like what happens when you keep tapping the center button on iMessage’s predictive text: “Not now but right now, you know, I had to go downtown.”
Regardless of how well it spits bars, Robbie Barrat’s rapping A.I. shows just how accessible coding neural networks has become. He made this project to prove a point to the fellow members of his high school programming club, reports Quartz. Apparently Barrat and his friends were arguing over whether artificial intelligence could outperform humans in certain tasks, and he rose to the challenge.
“All of the sudden I had a week to make a neural network that could rap,” he tells Quartz.
Barrat used open-source software on his Linux machine to program the neural network with 6,000 lines from Kanye West songs. He then trained it using the open-source machine learning library PyBrain.
“Originally it just rearranged existing rap lyrics, but now it can actually write word-by-word,” Barrat tells Quartz.
The result, like other neural networks that draw from real people, is a bit hit or miss. But the result of Barrat’s project is less important than the creative and technical process that got him there.
Open-source coding is making it easier for young people like Barrat to teach themselves the skills that the tech industry craves. Silicon Valley is seeking more creative engineers, and someone like Barrat, even though he’s not sure he wants to join their ranks, gives us a glimpse at how today’s open-source resources can accelerate creativity and yield surprising results.
