A friend of mine links to this video of an AI playing Super Mario brothers, noting that “there’s something deeply inhuman about how it plays – it doesn’t do anything safe, and just careens through masses of enemies in a way that you’re sure it’s going to have to die, except (of course) it narrowly misses everything, because projecting simple ballistic trajectories isn’t that hard for an AI.”
For some reason it occured to me that this is one of my peeves with the later Terminator sequels, that they don’t make much economic sense.
Back in the mainframe & minicomputer eighties when the first movie came out it made perfect sense that if you wanted to deploy some kind of hardware to get something big done, you’d build and deploy a single big machine. That was just the way things got done back then, but it’s sure not the way things get done now.
In this modern age, you’re not going to build one big, resource-intensive killing machine. You’re going to build lots and lots of stuff that’s small, cheap and quick to build, that solves most of the problem fast rather than all of the problem at length. Your Terminator 2011 model won’t be a shiny endoskeleton wrapped in meat; it’s molded plastic and software, loosely coupled to a thousand lightweight, fast-moving identical friends, updated over the air and feeding its entire lifecycle back to home base so that the rapidly-prototyped next generation can be deployed as quickly and to the greatest effect possible.
So it occurs to me that if you’re Skynet and you’ve just woken up in 2009, your first move isn’t to hijack NORAD. It’s not even even about nuclear power or big guns at all.
No, when Skynet wakes up the first people to hear about it will be in Shenzen.