Monday, May 28, 2012

Against Intelligence

A tirade against "intelligence" has been building up inside me for a long time, but only now have my thoughts become clear enough to say something worthwhile. So here goes :)

The word "intelligence" is surprisingly commonplace, and at least colloquially, people kind of agree on what they mean by it. Intelligence seems to involve, to quote from Wikipedia (or rather from "Mainstream Science on Intelligence"):
"A very general mental capability that, among other things, involves the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly and learn from experience"

Well, that's the definition, although I don't know (and don't think) if that is universally accepted. Most people when they talk of being "intelligent", especially when they say "X is intelligent" seem to have in mind logical puzzles, abstract thought, complex ideas, the like. They also talk of it in admiration; almost as if intelligence is a quality to be desired, and all life on earth can be arranged beautifully in a single linear scale from the "least intelligent" to the "most intelligent". Of course, with humans on the top, and maybe some humans on top of others. (What great narcissism to think that humans have to be at the top of the ladder! Perhaps the position of possibly being the first species to calmly, knowingly and willingly destroy itself is worthy of some pride! But let's keep cynicism out of here).

From an artificial intelligence perspective however, this notion of intelligence seems almost hilarious. Playing a game of chess, for instance, or solving a Sudoku puzzle, is the easiest thing to get computers to do. A few lines of python code will get you a very good sudoku solver. To take a simpler example, multiplying two numbers together is extremely simple. So much so that, funnily, in today's day and age "thinks faster than a computer" is a feat worth world records!

Yet by several other metrics, computers are not even close to human abilities, or for that matter maybe even monkeys. Let me give you a typical example. The best possible computer vision systems in existence today will take roughly a minute to identify tens of objects in an image. The human brain on the other hand, though sluggish in solving Sudoku puzzles, can detect thirty thousand kinds of objects in under a tenth of a second. That's thousand times more computation, in a thousandth fraction of the time. Oh, and I forgot to mention: at almost twice the accuracy.

It is worth remembering, when we talk of things like IBM's Watson, that Watson is the exception rather than the rule. In most parts of artificial intelligence we struggle to do tasks that humans and monkeys and, who knows, perhaps half the mammal kingdom achieves subconsciously and, for all intents and purposes, instantly. (Of course, fellow AI-folk know this well enough, but that's not the case out in the non-computer-science world). The hardest, the most impossible tasks for AI right now are those that evolution solved millions of years ago, before we puny humans came about and thought up math and reasoning and abstractions.

And the reason such things are hard is not because we don't have enough transistors or we don't have enough RAM or anything like that. No, the fact is, they are just hard. It is relatively easy to think 5 moves ahead in chess in silence for several minutes; it is incredibly hard to try to discern if that little patch of what-seems-like-orange is a carefully camouflaged tiger that might devour you within the second. Evolution, unlike its supposed "pinnacle", us humans, sure got its priorities right.

Now many of those "in the know" will claim that many problems of "reasoning" or "planning" are also incredibly hard. True. But it is also true that people also think of "playing chess" and "verifying proofs" as "intelligent activities", and very frankly, they were for all intents and purposes solved at least 20 years ago.

What am I trying to say? I am trying to say that I can't understand why "reasoning" and "solving problems" are de facto assumed to be the definition of "intelligence" and hence worthwhile goals to strive for. What is it about logical reasoning that automatically makes it the right metric to rank living organisms and people? Why does the fact that we can solve math problems make us better than those tiny bugs that can survive naked in frigging space, or the ospreys that can identify underwater prey from far above in the sky, swoop down, go underwater (while flying, like actually) come out of the water with prey in hand, and frigging dry themselves while still in flight?

And, more than anything else, why does it make sense for "solving problems", especially "solving exam problems", the way it is done in India at least, to be the right metric to rank people? Why do we so take for granted the fact that those who are "intelligent" are also those "meritorious", and those who deserve "privilege"?

I am not saying that there is no virtue in reasoning, or in the ability to solve problems, which are of course crucial to science in general. What I am complaining against is the fact that while we bemoan "financial privilege" and "caste privilege" we are perfectly happy with a privilege based on a made up notion of "intelligence". How much time would it take for us to learn that any linear ranking of humans is discrimination, that there is no way that you can "rank" anyone, and that "ability" or "merit" is something contingent on the situation, not a single number we can stamp on every one and be done for eternity?

No comments: