The end of Music composition as we know?

Yesterday I came across a BBC article about AI music composing.

Interesting article, talking about SKYGGE, the new musical project of French composer Benoît Carré.

He’s been in the composition business for a long time and now he just created this album using the help of Flow Machines, “the world’s most advanced artificially intelligent music program”.

Flow Machines analyses and interprets every musical score input it receives to create musical patterns, harmonies, chord progressions and melodies to deliver back some suggestions so that you can create your “own” music based on these interpretations.

Having these inputs of great successes analyzed and reinterpreted by the machine is great. Logic tells that you’ll end up having one major hit yourself.

Since Music is pure Math, it makes total sense. Artificial Intelligence has been used to help to create classical scores, Irish Music and all sorts of music in all sort of styles.

Take Melodrive, for instance. Instant music creation using AI.

But here’s where things get interesting.  By only having this, you’ll have “more of the same” hit in your hand, right?

Exactly.

AI is great, all data input is great, all algorithms and processes are great.

But here’s where it stops.

You’ll need some human to gather all this info and put together in a proper way so that you won’t end up having something that sounds great but also that’s been “already heard” in some way.

Also, although Music is pure Math, the unpredictability is what makes music great.

Having a chord progression change where it is unexpected.

Having different voice (or instrument) melodies over the same chord progressions used before.

Creating silence, tempo changes, sounds, processes, and everything that can take you and your music apart from the obvious.

Here is where AI fails.

At least for now. Maybe if we input also these different creations into the machines, they’ll learn how to make them and surprise us someday.

__________________

Here’s the album.

The end result is interesting. Great production, but sounds like something’s missing.

Maybe in the eagerness of doing something different, Benoît ended up doing some too-much-different choices, just to prove the point that AI won’t only deliver the expected.

Looks like somebody sent him a link to download/access Flow Machines (don’t even know if it is downloadable) with the message – “Go crazy!”.

🙂

Cheers,

Billy.