“Now and Then” is sweet and has a good backstory, but this is not the last Beatles song.
2024: Sean Lennon finds a tape of his dad whistling a fragmentary tune and muttering a few lines. The remaining Beatles and Giles Martin run it through AI trained on John Lennon’s music that hallucinates the rest of the melody and harmonies, and then has another AI sing John’s vocals in his voice. (These AI voice impersonations already exist, but it seems the YouTube videos of the ones I found when I blogged about AI music have been taken down due to copyright claims, and when I search for “AI generated John Lennon voice”, of course I get thousands of news stories about this official song.)
2027: the Beatles run a population of AIs in virtual reality that are exposed to skiffle music, post-war hardship in Liverpool, and family tragedy, and develop into virtual John Lennons and George Harrisons. The remaining Beatles and Giles Martin select the virtual John and George that produce the best Beatle-esque songs and performances, expose them to recent events, have them chat with actual Paul and Ringo, then prompt them to write new songs. (Large language models can already generate do-re-mi melodies, chord notation, and/or sequences of pitch and duration numbers.)
2030: everyone who wants to has their own set of John, Paul, George, and Ringo neural weights to write songs for them.
Creating music is like transferring our feelings into a song. It’s a beautiful way to express emotions and connect with others on a deeper level.