This neural network can change the style of music

Is there anything the AI can’t do? Starting from editing photo to smelling explosive to detecting cancer, there are a lot more things can be done by computer. And now it can even change the style of music as well. Researchers from Facebook AI Research (FAIR) just unveiled a neural network, which can convert music from one genre, style, and set of instruments used to another.

From now on, if you want to change the style of any music, just ask the AI and it will do that for you. For instance, if you choose a pop song as an input, the neural network can turn it into a heavy metal one.

“Our results present abilities that are, as far as we know, unheard of. Asked to convert one musical instrument to another, our network is on par or slightly worse than professional musicians. Many times, people find it hard to tell which is the original audio file and which is the output of the conversion that mimics a completely different instrument,” the team’s whitepaper says.

Unsupervised learning method

Although there are lots of research going to teach the AI to do things like humans, FAIR is the first team that developed an unsupervised learning method to recreate high-integrity music using a neural network. the team reached the integrity level by educating the neural network about how to auto-encode audio. For the AI it’s not a style transfer, just a conversion of a bunch of noise into another bunch of noise.

The research team says, “We distance ourselves from style transfer and do not try to employ such methods since we believe that a melody played by a piano is not similar except for audio texture differences to the same melody sung by a chorus. The mapping has to be done at a higher level and the modifications are not simple local changes.”

The auto-encoding is quite a complex method, which lets the neural network audio from inputs it’s not aware of. So, the network doesn’t change the music by memorizing notes or by matching pitch. Instead, it’s unsupervised learning method changes the music given as input using high-level semantics interpretation.

In the past few years, artificial intelligence has come so far. It can now even clone human voice, book a restaurant for you, and what not. The music FAIR’s AI created could be mistaken for actual humans’ creation. It’s likely we are heading towards a world where we won’t be able to believe our own eyes, voices, and even ears.

Artificial Intelligenceauto-encodingFacebook AI ResearchFAIRFeaturedhigh-level semantics interpretationneural networkunsupervised learning method

Originally published at thetechnews.com on May 23, 2018.