New Definitions of Skill

 

 

The Rifts album was a ‘Turing Test’ for music to see whether a human with meager skills at the piano can improvise in a mode better than an algorithm. The next step would be to code the algorithm to do the same thing. Will the machine create better music in terms of the ‘free will’ that humans use when playing music? Will it understand the conceptual elements or metaphors? When humans improvise music the ‘rules’ are changing in real time. As of now I don’t know, aside from Markov Chains, what is used to create this kind of real-time improvisation that is somewhat listenable. 

The major hurdle for assisted art is the size and accessibility of the data set. It would have to be oceans of data to be the equivalent of the ‘data’ humans use for creativity. And how is that ‘ocean’ defined’?    

As far as music is concerned, I’d rather listen to Miles Davis or John Coltrane’s modal improvisations, even if a bad performance, rather than try to solve a problem in music that it never had. It’s a solution in search of a problem.

9/10/2017

[9/10/2024: Skill is an interesting concept in AI-generated content. Where is the skill? The skill is in finding the art in something. Mere generation of content isn’t art necessarily. For example, how can generated music have genres? What’s the avant-garde in AI music? How could a new movement arise from it? What’s “punk” and “grunge” in AI music, aside from using it in the text prompt? Piano improvisations are now easy by typing words on a computer keyboard, not a piano keyboard. If there was a 21st century John Cage who composed a 4’33” sitting in front of a computer, would the sounds we heard in our heads be generated by AI?]