Tomoko hears and feels the humanity expressed in the piano
compositions she plays. But now artificial intelligence can play and compose music,
even symphonies. What, if anything, is lost in the process? Could there be a
future classic?
As one example, overseen by the Director of the Karajan institute
in Salzburg, an international team fed a speech recognition program Beethoven’s
nine symphonies and preliminary work on Beethoven’s 10th as well as other pieces by his contemporaneous
composers. Excerpts can be heard at https://www.youtube.com/watch?v=EAgnSr5t1VI.
As the video demonstrates, Google was able to use algorithms
to enable Google users to input a few notes that could be manipulated by an
algorithm to “compose” a piece in Bach’s style. Chi Cao’s senior theses consisted
of a software program he developed called “AI Classical Music Composer.” This
program drew upon music theory and knowledge, which Chi Cao transformed into
hundreds of lines of code. In this way, more people could have the opportunity
to experience the joy of composing music.
On a commercial basis, Aiva Technologies created the software
program called Aiva: Artificial Intelligence Virtual Artist. The inventors used
deep learning techniques to “teach” the computer to compose music. The generated
pieces have been used for film, ads and gaming soundtracks. The company
released a album “Genesis,” and Aiva has been given the official legal status
of Composer, including copyright status. As such, Aiva is the first AI to gain
this recognition.
The question remains: where is the heart of AI-generated
music? Maybe it rests in the ears of the beholder. The Turing test states that if
one cannot tell if a person or a computer is responding, then it may be
considered artificial intelligence. That process can easily be applied to discerning
the difference between human-generated and AI-generated music. Of course, such
a test probably says more about the listener than the composition. Tomoko would
know the difference….
