Orchestration?
Dramatically different arrangements can be very effective at reusing old patterns. I suspect this is basically why those AI's are making impressive results.
Enjoy the benefits of a premium membership!
Upgrade Account
-
TalkBass has been independent since 1998. Add your voice.
Create a free account to reply to discussions, view embedded media, and browse with fewer display ads.
Join free Log in
Want zero display ads or expanded classifieds tools? Compare plans.
Artificial Intelligence (AI) generated music
- Thread starter wintremute
- Start date
Orchestration?
Dramatically different arrangements can be very effective at reusing old patterns. I suspect this is basically why those AI's are making impressive results.
This gets back to the Borges story I mentioned, and the notion of "vast" numbers: Yes, there is a finite amount of possible note and chord arrangements, in the strictest sense of the word -- it is not genuinely infinite -- but the amount is so vast, such a huge (and hugely incomprehensible) number, that for any practical consideration it may as well be infinite.
Folks much better at math than I can probably express this in more compact terms (e.g., am I describing a factorial?) but consider the first two consecutive notes that a musical phrase can begin with...including repeated pitches, and including rests. How many possibilities exist? Now, how many different rhythmic variations describing the placement of the first two consecutive notes that a musical phrase can begin with are there...including rests (of any length) both before and/or after the first note? Now how many possibilities exist?
And that's just for the first two notes. As I was suggestiing in my previous post, any computer/bot that could actually generate every possible combination of notes would create a database that was so unweildy it would be impractical to search it. Or possibly even to simply store it.
Some students shared an article on AI music a few years back, which is where I learned about it. It's very interesting. I'm sure that, like sampling or synthesizers, it's something that's here to stay as a part of how we produce music, especially for commercial purposes. And like sampling (in particular), it's probably going to drive shifts in the music industry and copyright law.
The thing about it is, though, at least at the current state of the technology, is that it is derivative by nature, and that's going to both define and limit its place in the music world. You can program an AI to generate music that sounds like Nirvana. What we can't do - not yet, anyway - is program an AI that will do what Nirvana did to the music scene when they first came out. If you had this technology in 1990 and told it, "Write a rock hit," it would have sampled all the rock hits of the prior ten years and come up with something that sounded just like all the stuff that had been hits up to then. What it would never have given you was "Smells Like Teen Spirit."
Sort of, but there's no reason you couldn't program flubs into the algorithm in a way that the listener wouldn't realize it was AI if they hadn't been told. You can program a saxophone on digital software with random breathing variations as though a human saxophonist was taking breaths and no one would know.
But I think of it this way: a few years ago, IBM's supercomputer Watson competed on Jeopardy against the show's biggest all-time champion players, and mopped the floor with them, even though it sometimes spat out some absurd answers. A sufficiently advanced computer can beat humans at Jeopardy, or at chess, on a regular basis. It was fascinating to watch - one time. But I ask myself, would I want to watch a show where every episode was three computers answering trivia questions? Or two computers playing each other at chess? Not in a million years. Because what I really want to see is not a perfect trivia expert, or a perfect chess player - and not even a really good chess player with an occasional flub. What I really want to see is a human being striving at those challenges.
There are environments where we use music and the human element is not really important. Movie soundtracks, for instance, or elevator music, or a lot of advertising. As a market, that will probably entail some contraction of the demand for human composers and players, just like you had less string orchestras getting used in pop music once synthesizers became common. But the core appeal of music is not just to hear something, but to hear people striving at making it. I don't think AI touches that.
I was a theater major in the 80s, at a time when live theater and Broadway seemed to be dying, losing ground to movies and TV. It has since recovered. But at the time, what really struck me was the introduction to a book we were reading, where the author envisioned such a collapse of the commercial stage, and basically said, "So what?" Even if the economic model collapses and there are no more professional theaters, people will still dress up and put on skits in their living rooms and act out around bonfires in a field, because people want to act, and like to watch each other do it. I think something similar is true with music.
The thing about it is, though, at least at the current state of the technology, is that it is derivative by nature, and that's going to both define and limit its place in the music world. You can program an AI to generate music that sounds like Nirvana. What we can't do - not yet, anyway - is program an AI that will do what Nirvana did to the music scene when they first came out. If you had this technology in 1990 and told it, "Write a rock hit," it would have sampled all the rock hits of the prior ten years and come up with something that sounded just like all the stuff that had been hits up to then. What it would never have given you was "Smells Like Teen Spirit."
Sort of, but there's no reason you couldn't program flubs into the algorithm in a way that the listener wouldn't realize it was AI if they hadn't been told. You can program a saxophone on digital software with random breathing variations as though a human saxophonist was taking breaths and no one would know.
But I think of it this way: a few years ago, IBM's supercomputer Watson competed on Jeopardy against the show's biggest all-time champion players, and mopped the floor with them, even though it sometimes spat out some absurd answers. A sufficiently advanced computer can beat humans at Jeopardy, or at chess, on a regular basis. It was fascinating to watch - one time. But I ask myself, would I want to watch a show where every episode was three computers answering trivia questions? Or two computers playing each other at chess? Not in a million years. Because what I really want to see is not a perfect trivia expert, or a perfect chess player - and not even a really good chess player with an occasional flub. What I really want to see is a human being striving at those challenges.
There are environments where we use music and the human element is not really important. Movie soundtracks, for instance, or elevator music, or a lot of advertising. As a market, that will probably entail some contraction of the demand for human composers and players, just like you had less string orchestras getting used in pop music once synthesizers became common. But the core appeal of music is not just to hear something, but to hear people striving at making it. I don't think AI touches that.
I was a theater major in the 80s, at a time when live theater and Broadway seemed to be dying, losing ground to movies and TV. It has since recovered. But at the time, what really struck me was the introduction to a book we were reading, where the author envisioned such a collapse of the commercial stage, and basically said, "So what?" Even if the economic model collapses and there are no more professional theaters, people will still dress up and put on skits in their living rooms and act out around bonfires in a field, because people want to act, and like to watch each other do it. I think something similar is true with music.
- Oct 16, 2014
- 6,942
- 29,551
- Disclosures
- Endorsing Artist: Langstrom Carrot Farms
The only thing that surprises me about this, is that it didn't get figured out a decade ago.
"Socialise the costs, and privatise the profits" is a succinct way of putting it, and isn't just limited to this field, but is rife within the arts.
