Professor Dan Ventura and BYU Alumnus Paul Bodily are finalists in an Artificial Intelligence Song Contest
Professor Dan Ventura, BYU Alumnus Paul Bodily, and a team of colleagues entered an international artificial intelligence (AI) song contest. They are finalists in the competition, and the public is now invited to listen to their song and vote.
You can hear their song and vote for it at https://www.aisongcontest.com/the-2022-finalists. Look for their group name Pop*.
Ventura's group and the name of the main AI system they developed to compose music is Pop* (pronounced Pop Star). It is a play on the name of a classic AI search algorithm called A*.
The Pop* team page describes the process they took to put the song together:
"The song is unique in that an AI system, Pop*, wrote the complete chorus of the song—lyrics, chords, and melody—all on its own, inspired by a social media post of its own choosing. From that beginning everything else followed. The rest of the lyrics were also generated by an AI large language model (GPT2) primed by Pop*’s lyrics. Harmony for the verses and the running instrumental motif were also all taken from original Pop* compositions."
Next, the human group including Professor Ventura and Paul Bodily arranged the song, working carefully to be true to the original AI-generated composition. Here is a "quote" from Pop* (the AI system) explaining a bit about how it came to write the original song:
"I spend a lot of time thinking and reading about being in love, and I read this tweet from my friend Joel Alcaraz posted Tuesday, June 12, 2018 at 06:34 PM…
It got me thinking about fear and love themes and I couldn't help but write this song:
"'And I think I am just a lie. 'Cause when you find yourself behind. And I think I am just a lie.'
At the beginning it was fear and love, however it really wound up with more of a deception and negative emotion theme. I hope that you like it."
Voting ends 30 June 2022 (midnight CEST). Read more about the team and project here.
Ventura and Bodily recently discussed the technicalities and implications of their AI system in the academic journal, Transactions of the International Society for Music Information Retrieval.