We are well into the era when what we expect to see computer graphics characters in a live action film. Unlike characters in animated films, CGI characters in live action films are expected to be at least plausibly fit into the real world — although we are usually talking about a fantasy version of the real world.
We are not yet 100% there. This image of Smart Hulk from Avengers Endgame will attest to that. But we are getting closer, and within a few years CGI characters will look completely realistic.
Generally speaking, such a CGI character is driven by a real actor — in this case Mark Ruffalo. Teams of digital artists then convert that actor’s performance into a representation of a person (human or otherwise) who doesn’t actually exist.
Of course the possibilities will eventually go beyond the realm of fantasy. An actor with the talent of Paul Giamatti could play a character who looks exactly like Humphrey Bogart or Alain Delon.
In time, nearly all aspects of filmmaking will be pushed into post-production, including sets, lighting, costumes, make-up and camera. This shift from on-set production to post-production has the potential to make filmmaking far more fluid, flexible and inexpensive.
Yet the one thing that cannot be pushed to post-production is the actual performance of the actor. Nobody has any real idea how to use a computed substitute to replace the nuances of human performance — and the intelligence of the decisions that go into that performance.
To do so would, in a sense, be to solve the “hard A.I.” problem — the replication of human intelligence itself. Machine Learning can do many things, but it can’t even begin to do that.
In a way this is comforting. The technology we use to create art can continue to enhance communication between humans, but cannot replace it.