The resurrection of the Beatles for new songs and Sean Connery as the best James Bond ever in new films? This will soon no longer be a problem for the brave new world of AI.
The latest episode of our “Brave New AI World” series takes us beyond the grave: it was recently announced that the Beatles have produced and released “their last song”. Of the “Fab 4”, only Paul and Ringo are still alive, although ironically, the phrase “Paul is dead” was supposedly smuggled into the song “Revolution 9”, recorded in reverse and therefore only decipherable with the help of technical tricks.
AI could help to clarify whether this was the real content of the sound clip in question, or just an acoustic misunderstanding. In any case, Paul McCartney & Ringo Starr have recently managed another impressive trick using AI: filtering out John Lennon’s voice from a lousy cassette recorder recording. The restored vocals were mixed with original bass and drum playing by Paul and Ringo, a guitar recording by the late George Harrison from 1995, and so “Now and Then” was produced.
Paul emphasizes that the new song is “real” and that the AI did not generated John’s voice, but is merely used as a sound filter. However, AI has long since crossed the line between restoring and recreating voices that have gone silent forever. Edith Piaf’s AI-generated voice will guide viewers through an animated film biography of the French chanson singer, which Warner Music is producing. Entertainment giant Disney recently licensed the voice of 92-year-old actor James Earl Jones so that Darth Vader will be able to spread the terror of the Empire for decades to come.
The development of generative AI systems has made giant leaps in the past year and will continue rapidly. This includes, for example, an app – still labeled “Beta” by developer HeyGen – that automatically translates and synchronizes a speaker’s video, both lip-accurate and in the same tone as the original. Having an avatar speak a desired text, including matching facial expressions and body language, is even easier. To give the avatar your appearance, you don’t even have to move away from your desk – a photo or recording from your cell phone camera is enough.
A digital mask can already artificially age or rejuvenate actors in films. That’s practical because, for decades, the film industry has thrived on sequels to tried-and-tested hits like “Indiana Jones”. The fifth episode of “Indy”, just released, is supposedly the last for the now 81-year-old Harrison Ford, who has been digitally rejuvenated by 40 years. In conjunction with special recording techniques, machine learning makes this technique called “FaceSwap” possible. Similar applications are currently less perfect in the area of “deep fakes”, the creation of videos ofr scenes and statements that were never actually made.
Future AI applications will go far beyond what is possible today: More Indiana Jones episodes with Harrison Ford? A new James Bond with all-time favorite Sean Connery, who sadly died a few years ago? The Rolling Stones still on a never-ending tour in 20 years? No problem: AI systems will make such productions possible shortly and convince even the most critical fans. The celebrated ABBA musical with avatars of the band members serves as a taste of things to come. Soon, generative AI will become sufficiently creative to deliver appropriate scripts. This prospect recently motivated screenwriters and actors to embark on Hollywood’s longest strike.
These applications will also find their way into our everyday lives, for lively family videos of our ancestors and polished online presentations for companies and products. On social media, the age of fake and doubt began long before AI. Perhaps AI systems can contribute not only to the further deterioration of communication but also to its improvement: by using AI as a fake hunter.