Slashdot Mirror

← Back to Stories (view on slashdot.org)

How Artificial Intelligence Is Changing Music

Posted by Soulskill on from the i-think-i-saw-this-episode-of-TNG-already dept.

mbone writes "Ever wonder how Jimi Hendrix would cover Lady Gaga? Whether you do or not [I'm guessing not], you may be about to find out. Writing for Wired, Eliot Van Buskirk describes North Carolina's Zenph Sound Innovations, which takes existing recordings of musicians (deceased, for now) and models their 'musical personalities' to create new recordings, apparently to critical acclaim (PDF). The company has raised $10.7 million in funding to pursue their business plan, and hopes to branch out into, among other things, software that would let musicians jam with virtual versions of famous musicians. This work unites music with the very similar trend going on in the movies — Tron 2.0, for example, will clone the young Jeff Bridges. If this goes on, will the major labels and studios actually need musicians and actors? In the future, it could be harder to make money playing guitar with all of the competition from dead or retired artists."

4 of 261 comments (clear)

  1. Re:It's a shame, but I'm ok with it by theheadlessrabbit · · Score: 5, Informative

    It'll totally be worth it if it gets rid of the Nickleback derivatives.

    Since all Nickleback songs sound the same, does Nickleback count as a Nickleback derivative that will also have to be gotten rid of? That would be something all music fans can hope for.

    Oh, and as a Canadian, I'd like to apologize to the rest of the world for Nickleback. We're not happy about them either. Sorry.

    --
    -I only code in BASIC.-
  2. zenph does not play new pieces by pikine · · Score: 4, Informative
    Not to mention, currently all they do is to extract the timing of notes and the velocity of an existing performance from an old recording, and then play it using a player piano. Their technology doesn't play new pieces. From TFA:

    As things stand now, Zenph’s technology looks at actual old recordings to find out how a performer played a certain song, and is not capable of figuring out how a musician would play a new part.

    All they do is digital signal processing, not artificial intelligence.

    --
    I once had a signature.
  3. Re:It's a shame, but I'm ok with it by Nick+Number · · Score: 3, Informative

    That would be my thought as well.

    Fogerty v. Zaentz

    --
    Promote proofreading. Don't mod up sloppy posts.
  4. Re:roll over, beethoven, by Haxamanish · · Score: 3, Informative

    What halting problem? Never assume your reader knows everything you do.

    The halting problem: it is not possible to wite a program, let's call it P, which takes another program as its input and then tell if that program will stop or go into an infinite loop.
    To understand that this is impossible, imagine you would write a shell script which calls P and passes its own argument to P. Next the shell script would enter an infinite loop if P says its input will end. If P says its input would generate an infinite loop, the shell script would end. Now run the shell script and let it pass its own source code and the source code of P itself (for all practical purposes, P and the script form together a single program) as input into P. Now you get a paradox: if the shell script ends, it goes into an infinite loop and if it goes into an infinite loop its has to end...
    I second the advise on reading Hofstadter's GEB.