The musical AI is now working on its debut album(s)—and wants to do the Beatles better than the Beatles

Image: Reuters/Mario Anzouni
We may earn a commission from links on this page.

Researchers at Sony’s computer science laboratory in Paris recently put out a set of pop songs composed by an AI system, which scans songs from a database to compose entirely new pieces in certain musical styles. The final results were, unexpectedly, fairly catchy tunes.

Expanding that idea, Sony CSL now wants to use its system of machine-learning algorithms—which is named Flow Machines, and has been in development since 2012—to write and publish entire albums. Using an algorithm to write songs ”makes all the elements of experimentation easier, which otherwise would have been too time-consuming or meticulous,” says the lab’s director and artificial intelligence expert François Pashet.

Does a music-making AI violate the intimacy and emotion of the songwriting process? Many a (human) musician would likely say yes; Pashet and the rest of Sony CSL’s six-person team argue, though, that having machines serve as composers actually opens the music world up to a “new creativity” that isn’t possible otherwise.

It’s not like the music industry isn’t already heavily machinated: tech tweaks like auto-tuning, as well as conscious efforts to repeat certain riffs and musical themes have turned the business of pop music into one more of than mechanical strategy than artistic inspiration. Ask any fan of K-pop—music can be pretty robotic already.

Flow Machines is currently being used to write several albums, including one in the style of Beatles music and another adapting the style of current pop hits.

So a Now! compilation album of the future could feature such hitmakers as Katy Perry, Drake, Beyoncé, and Sony MusicBot AI-0177181.

Watch this next: “The Millennial Whoop”: The same annoying whooping sound is showing up in every popular song