The Synthetics are coming
Originally published 2006 in Atomic: Maximum Power Computing Last modified 03-Dec-2011.
Some people are wondering whether computer-generated actors will ever replace real people in the movies and on TV. Some people are saying they never will.
To my mind, this sounds like the people who wondered whether heavier-than-air flight would ever be good for anything - or, a bit earlier, whether it was even possible.
We are, I remind you, already living in the crazy sci-fi future. The CGI in The Last Starfighter can, of course, now all be done on a desktop box.
More relevantly, people watching Starfighter at the movies in 1984 probably would have considered it mere sci-fi gibberish if someone in the movie referred to a problem with "overly shiny cube maps on ATI hardware". Today, that's a routine Half-Life 2 bugfix.
And yet we still pay temperamental people lots of money to prance around in front of cameras with makeup on.
Backgrounds and effects are becoming more and more digital, creating weird hybrids like Sky Captain and the World of Tomorrow and the Star Wars prequels. People have complained about the crummy acting you get when people have to perform in places that don't exist and with props that aren't really there, but if Patrick Stewart can respond believably to an X on a stick that represents the Borg cube on his viewscreen, I reckon people paid twenty times as much can probably figure this stuff out.
They'd better, or their lunches are, most certainly, going to be eaten by synthetic actors.
Philosophers can - and do - yammer on for years about the significance of real people representing quasi-real people with the help of a team of other unseen, but real, people schooled in the tricks needed to make fake things look realer than real. It makes your head spin to think about it.
Won't it be so much simpler when it's all straightforwardly synthetic?
Most objections to these ideas seem to me to be the same sort of irrational emotional attachments that people have to other obsolete technologies. Vinyl records, film photography (OK, OK, that isn't quite obsolete), steam trains; they all have their charms, go ahead and enjoy them if that's your thing. But don't insist that because the new tech isn't quite there yet, it never will be.
Computer generated actors at the moment can be affecting and emotive (I saw you crying at the end of Monsters, Inc, don't try to deny it), as long as they're not trying to look like people.
Computer generated humans are, at the moment, right down there in Uncanny Valley - the part of the realism-versus-emotional-response graph where something's human-ish enough to trip our people-detecting circuits, but not realistic enough.
At the moment, the only way out of the Valley is by using human puppeteers to keyframe-animate your computerised characters, as was done by Andy Serkis in his digital Gollum suit.
If you think there's no way for wholly digital actors to move their bodies and faces like Sean Connery or Judi Dench, though (Clint Eastwood might be an easy starting point for the facial animators), I think you're in the same boat as all of the people who confidently predicted that computer chess programs would never beat human grand masters. Some of them kept predicting it right up until Garry Kasparov lost a game in 1996.
Later, Kasparov commented on how human Deep Blue's strategy seemed. Not at all the undirected flurry he'd seen from previous basically-brute-force computer opponents. Deep Blue was still the same kind of program, though. It just had faster hardware. A really big quantitative difference looked just like a qualitative one.
I see no reason not to think that the same will happen in the movie world.
It's not as if the money isn't there, after all. Rob Schneider's salary would buy a lot of processor cycles.
And the sooner it does, the better.
