This kind of AI composer would be just perfect for games, especially indy games. Most indy game developers can't afford good original soundtracks and if there was a service that provided them for cheap...

That's an idea for a stratup.

Indeed. Cope's musical grammar would be a great place to start hacking around. This reminds also of the iMUSE (http://en.wikipedia.org/wiki/IMUSE) from the SCUMM game engine, which would improvise soundtrack transitions based on motifs and in-game events.

Perhaps. On the other hand there's a lot of music on the creative commons nowadays, if you drop the requirement of originality.

However generating music that adapts to the game may be more than worthwhile.

I'm fascinated by this. It's interesting that in that context the creativity/passion/humanity of somebody like David Cope doesn't count. If he's writing software then it's just cold, heartless automation.

> it's just cold, heartless automation

An alternative view would be that he has captured part of the 'warm, emotional' essence of a few of the algorithms that are at work in musician's brains when they compose music is making them available to the world. Instead of just delivery a 'golden egg' he is trying to create the goose itself. To me that is a far more spectacular act of creativity/passion (no I don't see 'humanity' as being synonymous with these). It is probably worth noting the the silicon chip he runs the algorithm on probably operates at higher than the temperature of a human brain so perhaps 'warm, electronic automation' would be a better description.

To be clear, my comment was made in surprise that that his creativity/passion goes unappreciated - I agree it's an amazing accomplishment.

In general though I think the binary of warm/emotion vs. cold/heartless is kind of useless. As in his quote : "I can understand why it’s an issue if you’ve got an extremely romanticized view of what art is" Art is amazing stuff people make, be it software or music.

This guy has certainly met a lot of resistance in his day, but I'm tempted to assume that most of it has been from the pre-Atari generation (certainly pre-Internet).

The story mentions that he's contracted with a "pop group" -- if it's anyone noticeable, and if the music is at all good, I don't believe we're exactly going to have a teen revolt on our hands.

(Quite the contrary -- real creativity is rare in pop, and if a computer can bring it, this could catch on just as autotune did. Maybe even bigger in the long run.)

Perhaps more people will take to composing. If with the help of this technique, I can turn a small idea into a mediocre piece of music without too much work, I'll definitely play around with it. (I say mediocre, because I guess you'll probably still need skill to produce good and original music with it. But I will show interest even in mediocre stuff, if it's my stuff.)

It would be interesting to port Emily Howell to CUDA and copyright the billions of "promising" songs generated; thereby establishing something of copyright infringement troll corporation.

(Strictly as a thought experiment, of course, more in an attempt fix copyright than to break it.)

On the one hand, the function space is unfathomably huge; on the other, the "stereotypical" punk song is written using various permutations of only 3 chords.

Really it makes me question how to reconcile the difference between "discovering" and "creating" music. It would appear one can do both. Though can one imbue discovery with emotion as one can creation?

If we consider time of composition as a variable though, then the function space is even unfathomably huger.

What I mean is that even if those 3 chords have been played countless times in the past, right now may be the time to hear them again: when to play is an artistic choice as much as what to play is, because art relates not just with itself as a static object but also with the world around it.

This seems so significant for music as artifact, but less so for music as communication. Analogy: even if you get an AI to generate emails in the style of my friend, it still won't convey any content my friend wants to express. (Provides great insights into their writing style though.)

And what if your friend puts content they want to express as input for the AI? Did the content disappear simply because it's gone through a computer?

Now we're talking! :) I think the content is preserved in this case, but you see it still requires my authentic friend to generate it. (Unless we build deeper maybe...)