I described to a friend last night my recent post about a hypothetical “all star cast” of a play, consisting of brilliant actors who have played its various parts across different decades. My friend immediately replied “well of course that will be coming soon”.
What he meant was that we will, as technology progresses, become ever better at simulating the appearance of any actor, even those long deceased. I sort of agreed and sort of didn’t.
I completely buy the appearance part. Sooner or later computer graphics will allow us to create a virtual Audrey Hepburn or Cary Grant or Humphrey Bogart that appears completely indistinguishable from the original.
The problem comes when you talk about the performance itself. Any such simulation will necessarily involve a certain level of puppetry. Even if we run a machine learning algorithm to analyze all existing performances by this actor, and simulate their vocal performance, facial expression and body language accordingly, those algorithms cannot know about the higher levels of meaning in any new scene.
A great actor never quite repeats a previous performance. They use their face, body and voice as instruments, to reveal the unique meaning in each new psychological situation.
Our hypothetical “puppeteer” will need to possess an equivalent deep understanding, and then be able to use that understanding to drive a highly nuanced simulation of the actor in question.
In other words, while it may become fairly straightforward to bring Humphrey Bogart back to life as a mediocre actor, it will take a lot more work and talent to bring him back to life in a way that is true to the glorious original.