Predetermination

Imagine for a moment that robots have taken over the world. OK, I know this is kinda scary coming from someone who has declared himself "the guy who will put robotic servants in every home around the globe by 2020." But I promise it's only hypothetical. Humor me.

So robots have taken over the world. They have systematically destroyed all traces of human life, moved in to our homes, and are now going about their own robotic lives in what used to be our communities.

There is only one problem. The robots have been programmed to believe that the earth represents the entire universe. They have no concept of anything off the surface of the earth, and they do not know that there are still a handful of humans still living on board a space station orbiting overhead. It's up to these final few to defeat the robots and reclaim the planet for the human race. But how can they do it?

(OK, we assume by the time the robots take over the planet, that a space station has a handful of staff instead of 2 on the International Space Station, and working escape pods. We also assume other things that make it feasible. Humor me.)

Here is the point. The space-faring humans can rescue the planet by hacking the robots. All they need to do is (1) find a way to read the contents of each robot's central computer, (2) build a massive computer simulation of their own, and (3) extrapolate out exactly what the robots will do when. Armed with this knowledge they can plot their comeback and even test different scenarios by altering the computer simulation. Ultimately, the space station workers are successful in taking the planet back from the robots, and have a marvelous time repopulating the human race.

But is that really possible? OK, most of the story is not really possible. But the bit about creating the uber-computer-simulation is theoretically possible. "Theoretically" possible because it would be really hard to actually do, and would probably require more energy to simulate than to simply let reality play out and watch what happens. However nothing about it is impossible per se.

How do you write a computer simulation for everything that happens on the planet? I would start by breaking out indepent autonomous systems, and then finding ways to simulate interaction between them. For instance, we already have computer programs that predict weather patterns by analyzing current weather conditions and extrapolating out based on formulas that predict that sort of thing. You could call a weather pattern an independent autonomous system. Naturally, we would expect the robots' behavior to be influenced by the weather, so the weather simulation would have to be one part of the greater Equation for Calculating Everything.

Of course robots' interactions with each other will also impact their behavior, so we factor that in to the Equation too. We will create one autonomous simulation for each robot, and then merge them all together with the weather simulation. Now we can already reasonably predict where the robots will be and what they will be doing at any point in time into the future. Given the complexity of the Equation, our accuracy will likely decrease the further out in time we go.

To make it more accurate, we can incoporate a greater number of less-significant autonomous systems into the Equation. For instance, if the robots left any life on earth after they killed the humans, then that should be taken into account. Otherwise, you could have a siberian tiger come out of nowhere unexpectedly while you're battling a robot. So we throw together a quick computer simulation of what remain's of the Earth's Living Ecosystem. Test it out, and merge it into the Equation with the weather and the robot simulations. What's left? Seismic activity? I guess our simulation ought to be aware of meteors from space too in order to be smarter than the robots.

Once we have integrated all possible variables and factors into our great computer simulation of the Equation for Calculating Everything, we add a few exobytes of RAM, and start crunching out information. What will this robot eat for breakfast in 4 days, and what time will it eat it? Will that robot meet a life mate next month? Who will win the robots' next presidential election?

To further improve accuracy and visiblity further into the future, we can refine our simulations by breaking "independent autonomous systems" down into their actual specific components. The more detail our simulation takes into account, the higher resolution the image it produces of the future will be. For example, if we actually modeled every individual molecule in a weather system, we would have a substantially more accurate and long-term vision of the weather pattern's future. Today's models look at overall molecular composition (and temperature, pressure, trajectory, etc.) of weather systems, and as such lose accuracy the further out you extrapolate them.

The point is that all of these things are theoretically predicatable. That is, they are pre-determined. To make this point, I also like to think of comets in space. Comets travel our solar system (and probably many others) on paths that are defined perfectly by concise laws of physics. It is possible, and even relatively simple to predict a comet's exact position in space at any given point in time given knowledge about its current position and trajectory. If we get this information for lots of comets, and planets, and moons, we can start to predict impacts and observe craters being formed. Many were awestruck, but few surprised when the Shoemaker-Levy comet struck Jupiter in 1994 because such events are relatively easy to predict with existing technology.

In fact, the same principles apply on a much smaller scale too. Robots do not even have to take over the world for this to be the case. Even human behavior is predetermined and predictable given the right information and computer models. What we say and do is driven by neuro-psychological reactions to what we perceive happening in the world around us, or in response to activity within our own brains. It would be much easier to comprehensively simulate the robots' artificial intelligence in our Equation only because that's already in the form of a computer program. Human behavior however could also be modeled in detail given enough information about individuals' brain anatomies and chemistries.

Do what you will with the implications of that. I have strong opinions on the subject, but will reserve those for another blog. I will end today with my own prediction that within 50 years, students will be taught the principle of predetermination in high school science class alongside the likes of molecular biology, evolution, neurology, and quantum physics.