..And now the conclusion…

OK, so I’ve explained the Trash TV singularity, and you’re probably thinking “who gives a crap?”  Fair point.  I wrote it, and quite frankly, even I don’t give a crap.  But here’s something you should give a crap about:

THE TECHNOLOGY SINGULARITY

Proponents of this idea believe that one day soon, computers will achieve essentially infinite intelligence, at which point all bets are off (though personally I bet everyone’s brain will explode). 

How could anyone arrive at a bullsh*t conclusion like that?  Good question.  As I understand, it goes something like this:

“One day we will be able to build computers that are much smarter than us.  If our intelligence is such that it can create superior intelligence, than it follows that this newly created intelligence will have a similar ability to create intelligence greater than its own, and so on and so on to infinite intelligence.  Not only that, but because computers can think so much faster than us, this process, once started, will happen extremely quickly.”  

If the concept of infinite intelligence seems fishy to you, it should.  I’m not sure what it’s supposed to mean.  I guess adherents to this idea would probably say that infinite intelligence is a concept with no meaning, so most would likely settle for saying the intelligence will be so great we can’t possibly comprehend it with our puny monkey brains, and thus the consequences are utterly incalculable.  This brings us to the notion of a singularity, which in math is a point outside the normal rules of math (a loose approximation of the idea, anyhow.  It’s similar in physics, where a singularity generally refers to a gravitational singularity, known by normal people as a black hole, which is a point at which density becomes infinite, thus causing the known laws of physics to break down).

So why do I think this is a bullsh*t idea?  Well there are a few reasons.  One of the pillars of the theory seems to be moore’s law, which predicts that the number of transistors that can be placed inexpensively on an integrated circuit will double every 2 years.  Essentially, this is predicting that computing power will double every 2 years, resulting in exponential growth.  If this “law” holds indefinitely, it seems plausible to argue that eventually, computers will be so fast and so powerful, they will surpass the human brain in “thinking power”, whatever that might be.  Not only that, but shortly after computers surpass us, they will quickly double in intelligence, and then double again, and again, and again, and very rapidly turn us into insects by comparison.

But moore’s law isn’t really a “law” in the traditional sense.  It works now, sort of, but common sense should tell you that just because something has increased in a certain way for some period of time does not mean it will increase forever.  Take a tree.  Suppose for a moment that a tree grows the same number of inches taller every year for 100 years.  Does it follow that the tree will continue to grow forever?  Obviously not.  The fact that the tree grew a certain way for so long may say something about how much it is likely to grow next year, but it tells you almost nothing about how tall the tree will eventually get.  Similarly, there is no reason to think that just because we’ve been able to double computing power every 2 years for several years, that such a growth rate will continue forever.  As an anonymous contributer to the Economist illustrated in a tongue-in-cheek graph, (http://www.economist.com/node/5624861?story_id=5624861 unfortunately you need to be a subscriber to see it.  I got the story from Wikipedia: http://en.wikipedia.org/wiki/Technological_singularity#Criticism) we might as well use histroical trends to show that the number of blades available in disposable razors will continue increasing on toward infinity.

If that doesn’t convince you of how stupid the technology singularity is, then keep reading.  A central argument is that if we can build something smarter than us, that thing will have to same thinking abilities as us, namely the ability to build something smarter than it, and therefore we can conclude that this process of creation will continue to infinity (or at least to some kind of super intelligence).  But this does not follow at all.  You can’t simply assume that as intelligence increases, newer iterations will have the same exact ability to increase on their intelligence as their predecessors.  Why not?  An example may be illustrative here.

Lets say I create a robot which is physically like a human (2 arms, 2 legs etc.), but 2 times bigger than me, using only my bare hands.  Setting aside considerations of this robot’s skills at designing other robots, does it follow that this robot could physically construct another robot 2 times as big as itself?  Possibly.  Does it follow that this process can continue very long?  Not at all.  I can build something 2 times bigger than me in the same basic form factor without having to worry too much about whether or not it can support its own weight.  However, if you start to get much bigger, you’re going to run into structural problems.  If the dimensions of an object scale up by 2, the volume of the object (and therefore its mass) will scale up by a factor of 2^3=8.  But the strength of a material scales with the area of a cross-section of that material.  So if a robot doubles in size, it’s strength will increase by a factor of only 2^2=4.  In other words, as something grows, its mass increases faster than its strength (this is why insects are able to effortlessly lift objects many times their own weight).  This means the mass of a robot growing in size will soon outpace the strength of its legs, and it will collapse.  Now, the robot could try to come up with ways to keep the thing standing, maybe by adding more legs.  But now we’re asking it to step outside the rules I created in my original robot.  Not only that, but its own strength will not scale with the mass of the robot it’s building.  It may only be strong enough to build something 1.5 times its size.  

Or consider what happens when the robot’s height becomes so great that it has to start taking gravitational effects into account (a robot the size of the ocean, for example, would have to deal with the stresses caused by the moon’s orbit).  I didn’t have to deal with these forces at all, so why would anything I created or any of my creation’s creations have the ability to deal with these forces?  Similarly, a super smart computer may run into obstacles that we can’t even contemplate, obstacles which could slow the rate of intelligence improvement.  It may even run up against a theoretical limit.

Physics has a nasty way of screwing up our hopes for technology.  Consider the fact that if we keep shrinking the size of transistors to squeeze more of them onto a chip, we will run up against an atomic barrier (once you get down to the size of atoms, you can’t get smaller because there is nothing smaller than an atom to build with).

Now, this does not preclude a breakthrough in the underlying technology which may revolutionize computing as we know it.  Perhaps silicon-based transistors will be replaced with something that doesn’t have the same limitations.  Certainly there people already working on the next big thing after silicon-based chips, possibly utilizing optics or quantum effects.  But can moore’s law really predict a breakthrough of that nature, much less how computing power will improve after a breakthrough?  I don’t think so.

OK, so maybe infinite intelligence isn’t on our doorstep, but what about a computer that is merely just as smart as a person?  Surely that can’t be too far away.  I mean, in less than 10 years we went from giga pets to robotic dogs, surely human-like intelligence in a computer can’t be far off?  Maybe.  I have my doubts though.  The human brain is incredibly sophisticated and complex, and it’s not easy to unlock its secrets.  We really don’t know what makes it function the way it does.  There’s no instruction book to go by, and physically mapping the millions of cells that make it up is not really possible right now.  The best we can do is look at brain regions and try to figure out what general tasks they are associated with.  Kinda cool, but “the Matrix” is still a loooong way off.

So if we’re nowhere close to creating a blueprint for how an actual human brain works, is there good reason to think we’re anywhere close to duplicating the functionality of a human brain?  No.  Creating intelligence is more than just building a fast computer.  You have to have the right program.  And if we can’t even figure out how our brains are programmed, then how are we supposed to design something better?

I don’t know, maybe SkyNet is upon us, and robots from the future are going to descend on us any minute to terminate the future leaders of the inevitable rebellion by future humans against our soon-to-be robot overloards.  Maybe.  That would rule.  I’m not holding my breath though.