The year 2023 saw the world receiving a rare and unexpected gift. The remaining members of The Beatles—Sir Paul McCartney and Ringo Starr—collaborated on their last song where all the Fab Four played together. And it couldn’t have happened without artificial intelligence.
After The Beatles broke up in 1970, each member went their own way. Some had successful recording careers, but a reunion or a comeback seemed to be way out of the question.
Imagine
John Lennon was particularly prolific during this period and recorded and released various solo albums. He recorded a song on cassette tape in 1977 titled “Now and Then,” which was basically just him singing and playing on the piano. Lennon would be murdered in 1980, ending any possibility of band reunion.
Lennon’s widow, Yoko Ono, gave the remaining Beatles a copy of the recording in 1994, hoping they could play on it and release it as a tribute to her late husband. The surviving bandmates tried to make a go of it—they recorded parts around the Lennon section, but the quality of the recording made it impossible for them to lift just the vocals without bringing in the discordant piano. The project was shelved for 24 years.
Machine learning to the rescue
Director Peter Jackson (The Lord of the Rings), who helmed the band’s extraordinary Get Back documentary used his MAL machine learning technology to remaster the Lennon recording and successfully extract his vocals.
With this problem solved, McCartney and Ringo Starr added the bass and drum parts while using archival recordings of Harrison’s lead guitar as well as an entire string orchestra to round out what would be The Beatles’ last song. “Now and Then” was released as a single with “Love Me Do,” the band’s first song, as the bookends to a phenomenal music career.
AI as a remastering tool
Much has been said about AI’s use to create deepfakes, disseminate fake news, and misrepresent artists. Remember “Heart on my Sleeve”? The Weeknd and Drake song that they never really recorded?
Still, the technology offers staggering potential in recovering and recreating lost music, photos, and movies with surprising accuracy.
A 2008 fire in Universal Studios in California razed through three hectares of property and destroyed thousands of master tapes of hundreds of Universal Music Group artists. Also lost were various films and TV shows. The new breakthroughs in AI can, theoretically, upscale or enhance copies of these lost archives and bring them back to life.
AI for preserving culture
Aside from The Beatles, there are countless musicians whose unreleased works have languished for decades because they aren’t in good condition. Thankfully, the technology to solve this is now available.
If used responsibly, AI technologies like machine learning could unlock long lost works of art and preserve them for future generations. I’m personally expecting to see more of these applications bring lost treasures back to life.
Words Gadjo Cardenas Sevilla
This article first appeared in Speed’s December 2023 issue, which you can read here.
.jpg)
