Sunday, April 8, 2012

The Deep Blue Error



In a small room in the Pennsylvania Convention Center, 3 p.m, 10th of February, 1996, a historic occasion was about to emerge. Positioned in the middle of the room, there was a Chessboard. 

A Chess match of strange dimensions is going to take place here.

On one side, smartly dressed, sat down Garry Kasparov, Russian, and by agreed-upon accounts: the greatest Chess player of all times. Mr. Kasparov, by that time, had held the Chess championship for straight 10 years and had surpassed many of his chess grand masters peers by extraordinary lengths. To put it short: his accomplishments in that realm were unparalleled.

On the other end of the Chessboard, sat a gentleman by the name Feng-Hsuing Hsu. Yet he wasn't Kasparov opponent. Curiously, he was a computer scientist by profession, and his role at that room that day was to assist Kasparov real opponent: Deep Blue. Deep blue wasn't, or couldn't, be there.

It's noteworthy that Deep Blue wasn't intimidated by Kasparov ravishing reputation. Deep Blue, as it turns out, is an IBM computer.

This match, as rightfully hyped by the media, was the ultimate match between men and machines. The event was the stuff Hollywood movies are made off. The Terminator Vs. John Connor kind of match. There was a romantic flare about it. Human dignity was on the line as Kasparov mentioned prior to the game. The tension level in that room must have been off the roof.

The magnitude of inequalities in favor of Deep Blue was humongous; Kasparov, for example, knew it would be pointless to utilize his mental tricks with his adversary. But the most telling disadvantage was the fact that Kasparov, with his human mental capacity, could only calculate 3 moves ahead per second. Deep Blue calculated more 100 million moves per second. Deep Blue, in the language of human psychology, was overwhelmingly talented.

The game was set for 6 matches; the opponent with most winnings will be declared the winner. The game lasted 8 days.It captivated the world. In the first match Deep Blue beat Kasparov, putting a dint in the champion's record, and a larger one at the spectators' faith for a human to win over machine.
However, by February 17th, Kasparov had won two matches, and drew two others. The score was tied. The audience breath had lessened as the concluding match started. And by the end of the match, Kasparov emerged victorious.
Humanity prevailed.

Here is a collection of noteworthy questions: how did that happen? How did a limited functioning brain, with all the disadvantages accompanying it, had beaten a peerless performing machine? We know it wasn't luck. Luck is when you win the toss of a coin. It wasn't intelligence; Deep Blue, metaphorically, had a staggering intelligence comparing to that of Kasparov. It could calculate per second 100 million moves against Kasparov's 3 moves. So what was it?

The answer to this question is a lot less straight forward than we might think. There might have been lots of factors that groomed the victory of Kasparov, but for the purposes of this post, I'll focus on one. And it requires the abandonment of some false believes we've accumulated. But we are getting ahead of ourselves.

First thing we need to understand is that, contrary to old convictions, success in chess doesn't come from intellectual analytical prowess. Rather, it comes from memory.

Of course, I realize, there's something counterintuitive about that statement. But let's listen to the accounts of the performance studies expert, psychologist K. Anders Ericsson; he tells us that extraordinary performances aren't talent based, they're based on years of deep, purposeful deliberate practice. Factors like passion, patience, and the willingness to improve play a major role in transforming the human mind, thus distinguishing the expert from the novice.

The most distinctive factor about experts is their minds' mechanism of interpreting given information. It's unique. Upon several tests, it showed that experts when looking at human endeavors, Chessboard for example, they don't only see Chess pieces and black and white squares; they translate them into familiar patterns, words and pictures; their minds recall the images of previous experiences and it aids them to extract meaningful information that novices wouldn't comprehend. 

But non-experts face another problem.
Harvard psychologist George Miller in his famous essay "The Magical Number Seven" states that human ability to process information and make decisions is fundamentally constrained by memory; we can only process seven things at a given time, plus or minus two. When we encounter a situation, our working memory registers only around 7 facts. That's why phone numbers are only 7 digits. It's our built-in processing capacity. Working memory, it should be pointed out, is affected by emotions like arousal, intimidation, and pinking; which could impose further crippling constrains.

But is the working memory limited? Are we truly condemned to the number 7?

Ericsson has marvelously illustrated what it takes to un-limit that constrain. In his human performance lab, he brought in policemen; some were fresh out of the police academy, and some were experienced veterans. In his lab, there's an almost wall size screens displaying situations where a culprit was misbehaving. Ericsson's goal was to compare how the fresh grads and the experts dealt with the disturbing situation. The results were astonishing.
It took the fresh grads longer times to handle the situation, and in most cases, the culprit has escaped; they couldn’t make sense of the situation and what needed to be done. Experts, on the other hand, processed the situation differently; they weren't constrained by 7 factors, they had the advantage of reading situations in the way only experts can see: pattern recognition. They sized up the situation quickly because somewhere in their unconscious memory they recognized certain clues and cues.

This is the ultimate gift of practice, it doesn't matter of your brain can process 3 moves or 100 million moves per second, your mind will leap beyond the trap of number 7 and react based on the stored data of thousands of hours playing Chess or chasing criminals. Their memory has expanded beyond the normal capacity. That’s something that can't be learned in a text book or a lecture.

Kasparov-Age 11, Vilnius, 1974
After reading this, it makes sense to understand the unforeseen victory Garry Kasparov over Deep Blue. Let's quickly view the chronicles of Kasparov. He started playing chess at the age of 10 under highly experienced couches. That was in 1964. And over the years he accumulated banks of knowledge and expertise that in 1980, he became the World Junior Champion. And since that time, he scored a cascade of championships.

In his match with Deep Blue, Kasparov was counting on more than his analytical rigor, he was relying on a bank of information that he started depositing in over 30 years ago.
Experts view the world differently. Ericsson states that they depend extensively on their vast knowledge, pattern-based retrieval abilities, and planning mechanisms acquired over many years of experience in the associated domains.

Here is Kasparov: "I simply understood the essence of the end game in a way the computer didn't. I'ts computational power wasn't enough to overcome my experience and intuitive appreciation of  where the pieces should go."

Hard work and purposeful practice are much more important than talent. We're not imprisoned by our level of intelligence. We can escape that. We can escape the limits of number 7. Experts from different domains; movie actors, athletes, computer programmers, and countless other endeavors are living proofs of that.

Counting on intelligence alone could be crippling; experience counts. That was the Deep Blue error. It didn't posses that savviness. But IBM had realized that shortcoming. On May 3rd, 1997, Deep Blue faced  Kasparov again. This time, humanity failed.

How did Deep Blue beat Kasparov? It could be debated that IBM had upgraded the computer processing capabilities from 100 million to 200 million moves per second. But there was another factor they added and tipped the scale in favor of Deep Blue. IBM enriched the program with Chess grandmasters games with over 100 games. Deep Blue suddenly became more than just a machine; it had something it didn't have back then: vast experience.