Rock, robot rock

The music industry has long suffered at the hands of technology. Musicians were terrified by the idea that recorded music would replace live performance. Famed composer John Philip Sousa in 1906 wrote an indictment of recorded music, worrying that:

“when a mother can turn on the phonograph with the same ease that she applies to the electric light, will she croon her baby to slumber with sweet lullabies, or will the infant be put to sleep by machinery?”

The Federal Radio Commission (the predecessor to the FCC) regulated broadcasters of recorded music and in 1927 was fining them if they didn’t state a song was recorded before playing it stating that the airing of unlabeled “canned music” could be a “fraud upon the listening public.” One wonders what they would have thought of OpenAI's Jukebox which "generates music, including rudimentary singing, as raw audio in a variety of genres and artist styles."

The history of musicians and technology is a history of battles -- from the early resistance to recordings, to fights over digital formats that could be shared online (Napster) to streaming services such as Spotify. Now we have the latest battle with the Recording Industry Association of America (RIAA founded in 1952) targeting artificial intelligence as the threat to the well being of musicians.

As with image generation popularized by DALL-E, Stable Diffusion, and Midjourney there is an interesting intellectual challenge. These mathematical algorithms are "trained" with existing art and music, and thus are able to make sense of what a Matisse should look like or an Elvis should sound like. But should we consider the resulting novel images or songs "derivative works?" Even harder, another emerging category of AI for music promises to "make your songs sound (almost) as good as your favorite artists."

Every young musician starts out by playing the music of artists that they admire and we often refer to later work by those artists as being "influenced." David Bowie for example says that he was influenced by The Beatles, The Velvet Underground, and Little Richard amongst others. So should David Bowie have been paying royalties to those other artists?

Alternatively what is the test that should be applied to a specific piece of music to determine that royalties should be paid? With "sampling" there is now an established way for musicians to get clearance and pay royalties, but this practice allows an artist to identify a specific piece of music that has been used in their own work. What happens when an artists says "I've used a compendium of all recorded music..."

How we got here: Musicians perhaps more than other categories of professional artists have had to adapt multiple times over the last century to the changing economics of their field due to changes in technology. From recording, to Internet sharing, to streaming, and now artificial intelligence we've had to reassess the role of different participants in these ecosystems: performer, producer, broadcaster... Artificial intelligence may present a threat to the way things are done today but some common senses about how musicians actually make money suggests that AI might not be as big a change as organizations such as the RIAA (which defends copyright holders not musicians) would suggest. Live performance 28% and teaching 22% are the largest categories, neither of which will be immediately threatened by AI. Rock, robot rock (lyric sampled from Daft Punk).