Music created by an AI

I've started playing with it and it's fun, but like other experiments in computer-generated music -- including those of some of my students; I've had two develop functional, but very limited, music generators -- it sounds rather one-dimensional. In other words, the music is superficially appealing, but lacks any depth, direction, or (for lack of a better description) melodic or emotional message.

By way of analogy, if it were a poetry generator, it would be writing superficial, robotic (exactly!) poems like, "I am I am I am I am you are you are you are us are you are us are you are we are we are we are you."

Sometimes, it seems to gets the register of instruments wrong, like a bass will dip jarringly far (what is that low string, an E-what-the-****?), song endings feel like it's given up hope rather than finished, and occasionally it sounds out of tune though it probably isn't.

Of course, maybe that's because I'm using "Simple" rather than "Pro" mode. Have to try the latter, later.
 
I guess it's the first steps in the way of 100% AI created music.

I really see the potential to the best from human's creativity. Like Bookchin's view.

But on the other side I see leazy peoples and profits oriented people and it doesn't seems to me that we are not going that way.
 
Interesting stuff! I own a rare device which could possibly be considered a prehistoric version of this, the Triadex Muse. It was developed in the early '70s by a couple of MIT professors, and was intended as a "compositional synthesizer" (actually a sort of sequencer). It didn't have any keys; you set a series of long slider switches, adjusted the pitch and tempo controls to taste, and digital logic circuits within proceeded to generate a series of pseudo-random tones which, depending on the setting of the slider switches, could take weeks or even years to loop back around to the beginning! It even offered a "light show module" (which I'm lucky enough to also have an example of) which added a nice visual aspect to things. Here's a video of one in action:

 
I see a perfect little nutshell here:)
What's notable -- speaking as someone working on the periphery of the AI field, with colleagues in the centre of it -- is the rate at which AI work is improving by leaps and bounds. We're currently in an AI boom -- particularly in machine learning and neural networks -- largely driven by improved hardware performance.

On the other hand, AI development has gone through several boom and bust cycles and might soon hit another wall -- resulting in an "AI winter" like that from the mid 1970s to about 1980, and again around the late 1980s and early 1990s, where little innovation happens. Before the next "AI winter" comes, if it comes, it might result in AI that can make excellent music, because what AI research tends to reveal is that cognitive capabilities previously regarded as uniquely living, or uniquely human, can be done by machines.
 
Last edited:
Hahaha wonderful! Useless but wonderful, I like it
I'm not sure I'd call it truly useless. It was intended to help people write music, and with the right switch combinations, it can produce some interesting-sounding tunes (and the 'light show module' adds an extra, interesting dimension to the whole thing, IMO). While it didn't sell well originally (rumor has it that one of the only places the Muse was advertised was in Playboy Magazine! :dunno:), It has managed to make its mark on the world of avant-garde electronic music (into which category my music could be said to loosely fall into), and has made occasional appearances in the world of popular music. Here's the best example I know of, the random-sounding burbles heard throughout the middle part of the "In the Dead of Night" suite by the prog supergroup UK (starts at 5:28):

 
Back
Top Bottom