Slashdot Mirror
Two Technologists Create Black Metal Album Using An AI (theoutline.com)
An anonymous reader quotes a report from The Outline: Coditany of Timeness" is a convincing lo-fi black metal album, complete with atmospheric interludes, tremolo guitar, frantic blast beats and screeching vocals. But the record, which you can listen to on Bandcamp, wasn't created by musicians. Instead, it was generated by two musical technologists using a deep learning software that ingests a musical album, processes it, and spits out an imitation of its style. To create Coditany, the software broke "Diotima," a 2011 album by a New York black metal band called Krallice, into small segments of audio. Then they fed each segment through a neural network -- a type of artificial intelligence modeled loosely on a biological brain -- and asked it to guess what the waveform of the next individual sample of audio would be. If the guess was right, the network would strengthen the paths of the neural network that led to the correct answer, similar to the way electrical connections between neurons in our brain strengthen as we learn new skills.
This kind of reminds me of an old Microsoft project that didn't pan out. The "commercial" is pretty cringe inducing and not terribly interesting, but it gives you an idea of what it's supposed to do. The real fun was when people fed vocal tracks for popular songs into to see what it would do, often to comedic results: Queen, Johnny Cash, and Motörhead produce some amusing outcomes.
I don't listen to much black metal, but these results seem a bit better, or perhaps just far less silly.
Sorry, but did the person writing this story up actually listen to the thing? Not only are the songs totally indistinguishable, but they also suck. They are noise. The drums sound awful. The mix is harsh. The music, if you can call it that, is lame and boring. Granted, I don't have any familiarity with the Krallice album they used to train the AI - perhaps all those descriptors apply - but this hardly seems like any kind of real achievement. You could just slice up an existing black metal album into granules and then stick them all back together at random and achieve similar results.
This indicates to me that they're not really there yet. If you really want to compose something interesting, train the AI in a wide range of music first, then have it focus on a particular artist to mimic, with a goal being to produce songs that sound like they were produced by the same artist. Of course, to be really good, you would need to also handle lyrics, not just the music.
That has to be some of the worst shit I've ever tried to listen to. It's worse than hip-hop or country. I could put random notes on a page and it'd sound better.
If that's AI, then we have nothing (or everything) to fear.
They must have used the same AI to come up with names for the songs.
Congratulations, this is the first time AI has decomposed.
PlanetVulkan.com
The linked sample track is sold for 7 USD. I am not sure I would accept to listen to it again if I was given that money.
I've been involved with generative music since the 1980s. You could say it's a serious hobby of mine.
Making a generative black metal album like the one in the article is trivial, and could easily have been done a decade ago.
If you want to hear the state of the art in generative music, I'd recommend checking out the most recent album and the earlier four iOS apps from Brian Eno. His latest, Reflection was some of the best music in any genre released on record in 2017. It's light years ahead of this junk, which is an insult to the many talented black metal musicians.
You are welcome on my lawn.
Oh, my disk drive left me
Optical I lack,
when it comes to networks, well
I ain't got jack!
I got the Retina blues...
but I get good reviews.