Artificial Intelligence and Music

Artificial Intelligence and Music

Will AI produce really good music or even music that is merely a commercial success in the moment? Can computer brains put composers and musicians out of business? Is the day of amazing guitarists like Jimi Hendrix gone, replaced by computer generated perfection guaranteed to gain the widest possible audience?

What affect will an AI system such as Jukedeck have on new artists, when people have access to endless variations of compositions by their favorite artists, continuously customized just for them?  Just listen to a sampling of some of Jukedeck’s  AI compositions.  Remember that this is still cutting edge technology, still in its infancy.  Where will it be, tomorrow?

Already, musicians who know a bit about what is going on are concerned.  Their questions range from the more positive “How is AI going to enhance our own, human creativity?” to “Will AI replace us entirely?”

Alex Da Kid, a music producer from Wood Green, London, used IBM’s Watson Beat to write “Not Easy,” which hit number 48 on the Billboard Rock Airplay chart in 2016.  Watson Beat, called a “creative assistant” by an engineer from IBM, listens to music and scans social media for trends that could spark new ideas for commercially successful releases.

What is a bit astonishing, even a little scary for artists, is the song “Daddy’s Car,” composed entirely by Sony Computer Science Laboratory’s Flow Machines, which received more than 1.5 million views since it was posted on September 19^th, 2016.  Considering that we truly still are only at the nascence of this technology, it is hard to imagine just how far it will go.  Naturally, as with all uncertainty, there is a certain amount of fear mixed in with such imagining.

Baidu AI Composer generates music by “looking” at an image, then creating music based on the image.  While startling at first glance, it is in fact designed using image recognition software to translate the content, tone and mood of images into music.

One key issue that perhaps is all too easy to forget or ignore, especially when we’ve grown up on a diet of books and movies where autonomous robots turn on their creators, is that AI simply isn’t autonomous.  Hence, in fact, the term.  Artificial intelligence is something that has the appearance of intelligence, such as silk flowers often look startlingly real.  However, it and they are not.

The following definition of AI, translated from a post[1] on the subject in French states:  “At the origin of research work, artificial intelligence is based only on: mathematics, algorithms (i.e. programs that a computer executes via programming languages), and semantics (i.e. the study of language and the meaning of words).”

The reality of AI is that it is much more than math, but it is still artifical by definition, in that it requires input from humans.  Whether it is the type of weak AI such as you find in robotic lawn mowers or simple chatbots, or strong AI, such as is used for facial and voice recognition, human input is required.  To date, computers are not capable of autonomous (i.e. self-generated) thought.

Autonomous thought requires a conscience, something that science does not even understand, let alone have the ability to create.  Unless and until science is able to learn how to create true consciousness, computer intelligence will always be controlled and defined by humans.  It will always, therefore, be artificial, just like a silk rose, having a very strong resemblance to true intelligence, but nevertheless artificial.

With that noted, fantastic leaps have been made in AI in the last few years that are just beginning to make an appearance in music.  There is no doubt in my mind that AI is going to make a huge impact on the music industry.  It has already greatly improved the tools available for composers, with software such as Orb Composer by Hexachords.

What AI is doing for composers is making it much simpler to create and test full compositions with a wide array of instrument sounds, while also correctly creating and recording the musical score in the way real humans play music.  This is light years from early software that created musical scores from midi files, which rarely were written the way composers would actually write.  This made it very difficult for example, for another artist to play a composition as intended by the composer from such a score.

Today’s AI allows artists to play with their ideas in real time, even offering hints and suggestions based on the embedded intelligence drawn from a very deep knowledge and analysis of music that has now become possible.  Technical knowledge that once took great masters many years to achieve is now almost instantly available to modern composers, thanks to the advances in AI in recent years.

Will AI actually replace artists?  That is debatable.  The following two links beautifully illustrate one major difference between an AI composition and the same composition reworked by a human composer.

Here is Morning Birds, an AI composition.  Not bad, right?

Now here is Morning Birds, reworked by Gael Tissot, a human composer.  Can you spot the difference?

The two videos offer an excellent graphical representation of the difference between AI composed music and music created using AI by a human composer.  Color!  In case the video spoiled the affect, listen to both versions, again, but this time with your eyes closed.  Note how similar and beautiful they both are, yet the second one is much more, isn’t it.  The human conscious adds something that AI can never add, and that is original color.   AI can only put in what it has already learned, but it cannot create something truly new.  It is still artificial.

The fact that must be remembered when we begin to dream of buying a computer program that will make us a fortune from new hit songs while we vacation is that all of the AI music to date comes from human creativity.  Baidu AI Cpmposer, Hexachords Orb Composer, Jukedeck, Google Magenta, IBM Watson and Sony Flow Machines all work based on the parameters set by their creators.  They do not act autonomously.

Computers are, as they are so named, machines that compute.  They use processors to analyze, sift, annotate, calculate and process information that is fed to them, one way or another, according to the way they have been programmed.  Rather than replace humans, this frees up humans to be more creative than ever.  The machines then do the heavy lifting, in a sense, sifting through all the vast quantities of information available, recording our ideas, and in the case of musicians, writing musical scores, testing new ideas input by the artists, adding layers on request, etc.  While the machines are doing the “grunt work,” the artists are free to dream of new ideas, feed them to the machines for processing, and quickly weed out ideas that are not necessarily viable.

With that in mind, there is still no need for people to worry about the machines turning on us!  Artists need not worry that AI will start composing the music that will put them out of business.  Instead, AI is a tool that they can embrace to enhance their creative abilities, creating new opportunities to try ideas and techniques they may never have been capable of trying in the past.

[1]    Qui sont ces IA qui composent de la musique?  Tour d’horizon!  https://mbamci.com/les-ia-qui-composent-de-la-musique/