Now and Then: Is it really John Lennon's voice or is it AI?

Analysis: how the Beatles released their 'final' song with a little help from AI, decades after the death of John Lennon

By Andrew Hines, UCD

Now and then something happens that attracts a lot of attention in the media about Artificial Intelligence and whether it is being used appropriately: The Beatles releasing a new song in 2023, four decades after John Lennon's untimely death, was inevitably going to do exactly that.

'Now and Then' is the Beatles new single. It came about as a result of Yoko Ono, Lennon’s wife, giving a cassette to Paul McCartney in 1994 with a recording of Lennon singing a song and playing along on the piano in his New York apartment. Lennon's vocals were audible but sometimes masked by the piano and had other auditory quality issues that a studio recording would not have had, including tape hiss, mains hum, and room acoustics. As a result, it remained a work in progress; a demo that could not be morphed into a studio quality-single worthy of one of the most successful bands in history.

As McCartney says in a newly-released short film, in 1994 we "didn’t have the technology to do the separation… the piano clouded the picture". But when Peter Jackson, famous for the Lord of the Rings movies, worked on the retrospective Beatles music documentary Get Back, his team developed a new AI-based application to do what is referred to as 'source separation', which can take a music track and de-mix it into individual tracks or stems, containing each instrument or voice. Source separation is not new – computer algorithms have done this to greater or lesser extent for decades. However, with the acceleration of AI technology, it has gotten a lot better.

We need your consent to load this rte-player contentWe use rte-player to manage extra content that can set cookies on your device and collect data about your activity. Please review their details and accept them to load the content.Manage Preferences

From RTÉ Radio 1's The Ray D'Arcy Show, Dr Martin Clancy, Musician, Academic and industry consultant on the Beatles' new song

How does source separation work?

McCartney’s description of the "piano clouding the picture" is a good one. Imagine a photo of a piano with Lennon playing it: we can see his face and body but his hands and feet are obscured. Imagine now you want to separate the photo into a photo of the piano without Lennon and Lennon without the piano. It is easy for the piano: nothing obscures it. With Lennon to complete the picture, we need to imagine parts – is he wearing shoes or is he barefoot? This can be done, but the more we know about the individual the better the reconstructed picture would be. It is the same with music – once mixed, de-mixing it requires us to fill in the gaps.

So is it really John's voice or is it created by AI?

Here’s where the debate starts – is it really Lennon’s voice or is it an AI generated voice? One like Grimes’ Elf.tech where you can use her voice to release songs. It’s a fuzzy line for sure, between restoration and synthesis, but the world of visual art, for example, has for a long time acknowledged that restoring a painting vs creating a forgery are different things.

Consider how human perception works, how our sight and hearing converts light or pressure waves in the air into a perceived image or sound in our mind. Science has shown us that what we perceive is a mixture in our minds of the sensory input combined with our brain filling in the gaps, with data from our prior experiences. In a way, we’re using AI to do the same thing: fill in the gaps to produce an audio track of John Lennon singing, cleaned up to studio quality and remixed with vocals and instruments to produce something the Beatles could be proud of and share. As John’s son Sean commented, "my dad would have loved that, because he was never shy to experiment with recording technology."

Is it the Beatles? Absolutely - it contains original content from all of the fab four, restored by AI, then mixed and engineered to produce a sound that is both new and very familiar. Is it still art if AI was involved? Well, whether it is music to your ears is for you to decide.

The views expressed here are those of the author and do not represent or reflect the views of RTÉ