Steve Collins and Sarah Keith delivered a session titled Music and AI: Creativity, Control, and Copyright at this week’s Centre for Media History conference at Macquarie University’s Sydney city campus.
In their paper they explored the idea that there is no longer a single music industry but “many music industries.”
Similar to the history of radio, where manufacturers and retailers of radio sets needed content to encourage people to buy their products, so to, music technology companies need content to sell their products. This is reflected in the current trend of supplying music making technology cheaply or free to creators on the internet, but making them pay to extract their musical creations once they have used the software. In the past you bought the software or hardware, but this old business model is changing for music production tech companies.
“The IT sector is now entwined in the music industry… there is decentralisation of production… AI is transforming how music is created, shared and enjoyed,” said Collins and Keith.
AI tools can also analyse the millions of songs that have been hits around the world over time and determine the characteristics of a hit in different markets, languages and cultures. Once that analysis is done “AI can analyse a new song to determine if it will be a hit.”
Other issues discussed by the pair include whether AI used in song creation will infringe copyright and who owns AI generated music works. While there are current legal positions on both these questions, new technology may force a rethink of copyright law over time.
In relation to the business model of music streaming companies, Collins and Keith commented that Spotify “pays 70% of its revenue to rights holders,” so over time “there may be a strong financial incentive” for Spotify to produced its own AI generated music and prioritise it in its playlists, or to acquire the rights to musicians’ back catalogues to create new hits in their voice if they die or are no longer able to sing.
The pair mentioned an amusing example of how AI and voice cloning technology, Johnny Cash singing his version of the novelty song Barbie Girl in his traditional country style, but long after his death.
The pair recently published an exploration of AI in the music industry, in a book titled Slave to the ‘Rithm , The AI Turn in the Music Industries.
Getting famous, notorious and infamous people to sing songs they've never sang has been illustrated in this article where Johnny Cash sings the lyrics of Aqua's "Barbie Girl" to Johnny Cash's "Folsom Prison Blues".
Bugs Bunny sings Dean Martin's "On An Evening In Roma"
https://youtu.be/qTOd1IIZ6Q0?si=-Jc5n0HAqJQrU_3W
An evil dictator also known as the Austrian Painter who died in 1945 sings Bambee's "Bumble Bee", from 1999.
https://youtu.be/usKSp9yPM2Q?si=PRcHezPGV87bjjAK
For the Johnny Cash version of Barbie Girl, AI and statistical analysis techniques can be used to isolate instruments and vocals and substitute other vocalists.
In other situations AI and statistical techniques can be used to isolate musical instruments and vocals from an original mono recording to produce a stereo recording.
Example Eric Records has a CD album of "Hard To Get 45s". Some of these songs debut in stereo.
Example Strawberry Alarm Clock's "Incense and Peppermints" from 1967.
https://www.ericrecords.com/htf_17.html
In addition, anyone can have AI programs compose music to lyrics.
For example suno.ai lets one enter lyrics into the web interface, nominate a style and generate a tune.
The lyrics could be your own, or lyrics from existing songs. Experiment to your heart's content.
Examples
* Our National Anthem's lyrics rendered with a new melody based on a 1940s big band.
* The lyrics of the theme to "Gilligan's Island" sung with a new melody in a Gregorian chant.
However one has to question the source of the artists and vocalists used by the suno.ai program is violating copyright as suno is counter claiming that it is legitimate under the "fair use" doctrine in the US.
https://www.riaa.com/record-companies-bring-landmark-cases-for-responsible-ai-againstsuno-and-udio-in-boston-and-new-york-federal-courts-respectively/
The case is ongoing.
Thanks
Anthony, Strathfield South, in the land of the Wangal and Darug People's of the Eora Nation.