What does it mean for an algorithm to make music on its own ?
When the first computers were built, composers began to experiment with the new interactive medium. Max Mathews and Yannis Xanakis were pioneers in these early experiments. The earlier computer musicians were classically trained and were inspired by number sequences, rule based systems and stochastic processes for their digital melodies. If I am not mistaken something like open source was also conceived of by these folks, because sheet music for all practical purposes is in the public domain.
While this was going on, the new electronic instruments however had a much bigger impact on the music industry. People seem to forget that electronics has a long history with music. Radio, telephones, movies ... all had to deal with sound in one way or another. Vacuum tubes from analogue computers are still being used in guitar amplifiers!
Instruments like the Moog, Yamaha changed the face of rock music and drum synthesizers like TR 909 created whole new genres like Techno, DnB, House … samplers went on to lay the foundation for Hip Hop.
Every 90s kid and 2000s teen fondly remembers chiptunes from software cracks. This evolution we see in digital music is rooted in hacking around special audio chips designed for early video game consoles like Atari and Amiga and pushing creative programming to the limits.
By the late 80s from the digital side, you had the first musical programming languages emerge, along with the first music protocol MIDI.
Broadly speaking you can classify music programming languages into three types,
Visual - Pure Data, Reaktor
Textual - Csound
Live - Super Collider, Impromptu, Overtone
You don't need special programming languages to make music, as the demoscene community has shown - you just need an old amiga and 1337 assembly skills. You can hard code your melodies and wav files into C code directly too!
Stochastic experiments in these languages by Aphex Twin and others laid the foundation to IDM.
By the late 90s computers became powerful enough to completely disrupt hardware workflows of recording, mixing and effect processing. Physical modelling algorithms too were decent enough in many cases as compared the actual instrument, leading to a whole class of virtual instruments. This lead to nimble home producers who could make professional sounding albums - classical, rock, electronic - with nothing but a home studio and virtual instruments.
What now ?
We have robots making music and artificial neural networks making digital forgeries. You also have an increasing number of web based experiments with audio. A new medium VR dawns on us.
To what extent is this Art ?
30 minutes of randomness / ML / genetic algorithms produce 5 minutes of good music IMO. I don't think software, engineering and mathematics are art forms. I believe this goes back to Immanuel Kant who in his critique of art defined it as something that invokes the sublime. Going by that ... then yes, mathematics, engineering and programming are capable of achieving the sublime but they can achieve just that.
The essence of art is the emotions behind it. Humour, pathos and joy are important too. The intention of the artist is absolutely relevant. You can't make music in an emotional vaccum. The average listener can't wish away their emotional reactions. Going by this definition, yes Lou Reed’s monumental album Metal Machine Music is Art.
Ada lovelace asks, can computers surprise us ? Why stop at Ada Lovelace, when will Artificial consciousness replicate Lord Byron, her father in spirit ? Is that even a remote possibility ?
I think recommendation systems and the current state of AI has recreated obsessive compulsive disorder right down to the paranoia and intrusive thoughts. You could call it obsessive categorisation disorder. By feeding advertisements, clickbait, addictions, internet drama polygons, horrible content and biases to filter out ... how can the neural network not resemble artificial insanity ?
When an AI System can,
detect sarcasm
prank you
say sorry
I suppose we are getting somewhere with A.I, but the question still remains, what does it mean for an algorithm to feel something ?