CONFRONTING THE BEAST

SOME THOUGHTS ON KASPAROV VS. DEEP BLUE

by Steve Lopez

It's been just over 24 hours as I sit and type these words and I'm still waiting patiently for Doomsday to hit. As I look out the window of my basement office, I can see that the leaves are finally out on the two maple trees in the yard and that the grass is still green (and badly in need of cutting. "Doomsday" at my house comes when I see giant anacondas slithering through the tall grass and I realize that I can't put off mowing the lawn any longer). The view of the yard looks the same as it did two days ago, before Kasparov lost Game Six, and I expect that it will look much the same tomorrow. I can infer from this that Doomsday ain't coming.

I've been avidly reading the posts on the Usenet chess newsgroups for over a day now. Those that are posting fall generally into two camps: those that think humankind irretrievably lost something yesterday when Garry tipped his King and those that don't understand what the fuss is all about. Count me in the latter category.

I have a somewhat different take on computers than many people. I guess I can chalk that up to the time in which I was born. I'm old enough to remember what it was like when nobody except universities and large corporations had access to computers, but young enough to not be intimidated by the dang things. I bought my first PC in 1992; in fact, I'm typing these words on the same PC right now. I bought it for one reason (and never mind what I told my wife): to play chess. But a month after I bought the box, I had an epiphany of sorts while reading a book on information theory. I realized that a computer, by definition, by design, is a tool. Nothing more, nothing less. It's a bit more complex and versatile than a screwdriver but you can pretty much lump it into the same category. If I have a job to do, I need the correct tool. If I need to put together bikes for my kids I use a screwdriver. If I need to put together an analysis sheet on the hideous opening Kasparov played in yesterday's game I use a computer.

Now before someone takes me to task for saying that the World Chess Champion just lost a match to an electronic screwdriver, let's approach the thesis from a different angle. Bear with me here and keep the "tool" analogy in mind as we go along.

THE BIRTH OF A NEW TOOL

No, it's not the Binford Rocket Boy 9000 speed drill with four-speed reverse motor and adjustable chuck. It's something even more manly. Let's all grunt together as we read the following quote:

"By 'augmenting man's intellect' we mean increasing the capability of a man to approach a complex problem situation, gain comprehension to suit his particular needs, and to derive solutions to problems. Increased capability in this respect is taken to mean...that comprehension can be gained more quickly; that better comprehension can be gained; that a useful degree of comprehension can be gained where previously the situation was too complex; that solutions can be produced more quickly; that better solutions can be produced; that solutions can be found where previously the human could find none. And by 'complex situations' we include the professional problems of diplomats, executives, social scientists, life scientists, physical scientists, attorneys, designers -- whether the problem situation exists for twenty minutes or twenty years. We do not speak of isolated clever tricks that help in particular situations. We refer to a way of life in an integrated domain where hunches, cut--and-try, intangibles, and the human 'feel for a situation' usefully coexist with powerful concepts, streamlined terminology and notation, sophisticated methods, and high-powered electronic aids."

When Douglas Englebart penned those words in 1963, a new tool was born. "Congratulations, Mr. Englebart...it's a processor!"

Computers had been around for close to twenty years when Doug Englebart wrote "A Conceptual Framework for Augmenting Man's Intellect" but nobody was quite sure what to do with them. Sure, one could do complex mathematical operations with a computer, one could do mindbending statistical analysis, but what were the machines really good for besides crunching numbers?

Englebart's article defined what a computer was good for and even dared to go beyond that simple definition. He envisioned a world in which humans and computers worked together to solve problems that humans couldn't solve unaided. That world is defined with startling clarity in the paragraph I just quoted. Go back and read it again. Once the comprehension of what he's saying hits you, I defy you to prevent a Tim Allen-esque grunt from passing your lips.

But do you know the sad part? This brilliant vision was either ignored or scoffed at for close to another two decades. It took a couple of hippie-types named Steve to bring the concept to the world's attention in the early 80's. Nowadays many of us are so closely wedded to our computers that we can't imagine life without them. What took us so long to figure it all out?

FIGURING OUT THE NEW TOOL

The personal computer really is a remarkable tool. I can do all sorts of interesting and valuable things on it. I can store an entire encyclopedia on a compact disk and search for specific information on nearly any topic quickly and easily. I can simulate an entire English League soccer season in a matter of seconds. I can write letters and send them electronically without a pen having touched a paper or a tongue having licked a stamp. I can "talk" to friends the world over in real time by typing words on my keyboard, words which they see on their screens a mere fraction of a second later. I can control my personal finances, do my taxes, and even send my 1040 to the IRS electronically.

And then there's chess, the reason I bought the infernal device. I can play a game against a computer any time of the day or night, without shame or embarrassment if I should lose. Many years ago, I made part of my living as a bar-room chess hustler and never had a problem getting a game. Nowadays I take my set out and hit the town and I may have to go to three or four places before I find an opponent. Everybody who wants to play is at home either challenging their computer or playing online.

It's a far cry from the days of my childhood. Back then, if I wanted to play chess I had to either find an opponent or be stuck playing through games from books. The closest I could get to anything competitive in the latter case was Al Horowitz's "Solitaire Chess" feature in Chess Life. These days, I just turn on my PC and fire up any of my 50+ chessplaying programs. I can get a game at any level, from the no-brain play of Clueless to the razor-sharp tactics of Fritz4.

When my game is done, I can use that game I've played to help me improve. I can fire up ChessBase and do a database search on the opening I played in my game against the computer. I have access to opening trees and games with grandmaster commentary as well as hundreds of thousands of unannotated games. I can find out quickly and easily where I strayed in the opening and whether or not there's a better line that I could have played.

After that, I can have a strong chessplaying engine analyze every move of my game and show me exactly where I went wrong. And, after several of these analyzed games, I can see patterns developing in my play. I can isolate fundamental gaps and weaknesses in my chess knowledge and concentrate on improving these weak points.

Then I can go back and test myself in another game against the computer, starting the process anew. This is exactly the kind of "integrated domain" envisioned by Doug Englebart. I use the computer to test me and to educate me. I use it as a tool to find the answers to complex questions about chess, to test out ideas that I have, to help me identify strengths and isolate weaknesses. In short, the computer is my partner in the quest to improve my chess, rather than an enemy to be feared and reviled.

Just where did this idea of "computer as enemy" come from, anyway?

ELECTRONIC BRAINS: THE COMPUTER IS YOUR ENEMY

It's ironic that one of the world's leading "tech heads", Arthur C. Clarke, did so much to engender society's general mistrust of computers.

By the mid-1960's, most people had a mental picture of a computer as being some kind of large cabinet with flashing lights and spinning tape reels and which could perform mathematical tasks at a prodigious rate. Computers were interesting, a bit off-putting, but essentially benign machines in the minds of most.

Then came HAL.

HAL9000 from 2001: A Space Odyssey (and, to a lesser extent, the female-voiced computer on Star Trek) helped to define computers and the concept of artificial intelligence for a whole generation of people. And the picture HAL painted wasn't pretty.

In the film, a strange black monolith is uncovered at the bottom of the crater Tycho on the moon. It beams a high-powered signal outward to the vicinity of Jupiter. The government sends a team of scientists to Jupiter to investigate the recipient of this transmission. The spacecraft carrying these scientists is controlled in large measure by an intelligent supercomputer named HAL.

Making a long story short, two of the scientists decide to scuttle the mission. HAL has been programmed to carry out the mission at all costs. So he (it?) murders all of the scientists save one, who manages to disconnect HAL just as the ship is approaching Jupiter.

Charming picture, eh? And I find it ironic that the story was penned by one of technology's greatest champions, the father of the communications satellite, Arthur C. Clarke.

2001: A Space Odyssey didn't shape our perceptions about computers and AI single-handedly. It had a lot of help from the endless imitators that followed, from the computer-as-menace no-brainer film Colossus: The Forbin Project to the highly derivative David Gerrold novel When Harlie was One. The grand tradition continues to this day, with luddite films such as Ghost in the Machine appearing with alarming regularity (said film having the singular distinction of being the only movie I have ever walked out on. And, speaking as a connoisseur of "B" movies, that really says something about how totally tasteless and moronic it was). Even good science fiction programs (such as Star Trek: The Next Generation) suffer an occasional lapse and dredge up the hoary old cliche of the "maniac out of control" computer. It's a hard image to avoid, even if you want to, even if you try to.

Advertisers of computers even get lazy and fall back on this stupid stuff. I recall one ad in particular that showed a professional woman sitting at her PC, waving a floppy disk in front of the monitor, and telling the machine in a chiding tone, "Now I want this back..." I vowed on the spot to never buy a computer from the company that dreamed up that idiotic ad.

Unfortunately, the net result of this "computer as menace" blitz is that the message has been hammered into our heads over and over: "Computers are evil, computers are taking over the world, computers are bent on human destruction". It lurks in the backs of all our minds, eevn the minds of those of us who know better. No wonder thousands of people are still today afraid to touch the dang things, even after they've spent thousands to purchase one. Nobody really believes that these devices are sentient beings bent on achieving our downfall, but the message from the films that I mentioned nag at the backs of our minds.

Consider for a moment the fear that many first-time PC users display, just from having to work with a desktop Pentium 133MHz computer with 16 MB RAM and a 1 gig hard drive. Now imagine what these folks would do when confronted by a (*GASP*) mainframe?

WHITHER GOES IBM?

Deep Blue is one intimidating piece of hardware. It looks like two black refrigerators sitting side by side. The power cables for it make standard 220 cord look like angelhair pasta. Plus it requires a specialized environment: Deep Blue craves air conditioning as badly as the main character in H.P. Lovecraft's story "Cool Air". It calculates and evaluates chess moves so quickly that any attempt to display all of its thoughts on a monitor simply results in a cursor blur. And it accesses a database of over a million games so quickly that ChessBase looks like a snail by comparison.

This stuff is a bit overwhelming even for those of us who don't buy into "HALphobia". I can imagine what this beast looks like to someone who watched Colossus three too many times.

So why did IBM spend millions of dollars on a computer that would ultimately play chess better than any human in the world? Was it an insidious attempt to destroy humanity? (If you answered "yes" to that, turn off your computer right now and go turn on your TV. You're missing The X-Files). Was it an altruistic gesture designed to entertain and enthrall chess fans the world over? (If you answered "yes" to that, go turn on your TV as well; there should be some Fantasy Island reruns coming on soon).

First of all, they wanted to make buckets of ducats. That goes without saying (but I said it anyway). They received tons of PR value from it. CNN ran pieces on the match all week long; you can't buy publicity like that.

But they had another motive too, one which may also translate into tons of greenbacks for Big Blue. It was a foray into the realm of "expert systems", sort of the poor man's Artificial Intelligence.

Picture this: you're a doctor and you have a patient who is suffering from some kind of "mystery" ailment. You've consulted with your colleagues, read stacks of medical references until your eyes are red, and you're still baffled.

Wouldn't it be great to just fire up a computer, type in some information, and have it find a solution for you? And wouldn't it be awesome if it could do it interactively? The "conversation" might go like this:

You (the doctor) type in a list of symptoms. After a database search of a few seconds, the program presents you with a list of fifty possible maladies. It then asks you to get more specific: "Is the patient running a high fever?"

You answer "no" and the program eliminates the Malaysian Jungle Fever as well as 34 other afflictions. Next it inquires, "Does the patient display a nasal discharge?"

You reply "yes," and 10 more illnesses go out the window. The program then shows the remaining five diseases and asks, "Does the patient suffer from sinus congestion?" You say "yes" and BOOM! -- we're down to two.

The program asks "Is it ragweed season?". You reply "no" and one more affliction is eliminated. The screen now displays simply "Common cold". You click a button marked "Treatment" on the screen and after another database search, you finally get the answer you need:

"Chicken Soup"

Aside from the fact that you're now quacking like a duck, "Doctor", I'm sure you see the potential here. Such a computer/program would be an invaluable tool to the medical profession. My kid nephew is extremely sick right now, has been for months, and nobody can figure out what's wrong with him. I would love for the doctors to be able to utilize an interactive tool like this one and cure this kid today.

Of course, we saw how this "miracle" device works. Through a combination of database searching and interactive questioning (an active function of the program's set of instructions, or "heuristics") you get a sort of "doctor's helper", able to unravel the knotty problems facing members of the medical profession.

We call these computers/programs "expert systems". It's a way to digitize expert knowledge in a particular field of endeavor and make that knowledge available to less skilled individuals.

With a little thought, we can imagine other applications for this technology. Engineers can use expert systems to better design everything from stronger bridges to more aerodynamic NASCAR vehicles. Palentologists could type in the mineral composition of a plot of ground as well as the depth at which to search and come up with a probability of fossils being discovered there. The same process could be used in the search for fossil fuels. The possibilities are astounding.

CHESS EXPERT SYSTEMS?

Deep Blue used the two functions we've discussed (database search and heuristic questioning) to defeat the world's strongest human player. It asked itself questions to evaluate positions and simultaneously conducted database searches to look for similar positions.

Can this primitive expert systems technology be adapted to other, more important fields of endeavor? Only time will tell but the future looks bright.

But what about the chessplayers who feel let down and cheated by the fact that a big chunk of silicon just trashed the world's best carbon-based chessplayer? What's going to happen to chessplayers after IBM packs up its chess toy and moves on to these more vital areas of research?

I have news for you, my fellow player: we're all going to get better.

Allow me to digress for just a moment. I mentioned earlier how I couldn't play chess as a kid unless I managed to find a friend or relative who would sit down and give me a game. I learned the game of chess at the age of four but I never progressed to a competent level until I was over 30 years old. A big reason for this slow progress was a lack of opportunities to play chess.

I still remember (back in 1992) the first time I saw the combination of Knightstalker and ChessBase University. I told Don Maddox, "If I'd had this when I was a kid, I'd be a much better player today!" And it's true -- if I'd had these tools then, I might be halfway good now.

The strength of chessplaying programs has increased deamatically over the years. In the late 1970's it was considered a big deal if a program beat a Class B player. These days you can buy a chessplaying program for your PC that'll tear the head off of many grandmasters.

The technology is now in place and most grandmasters have embraced it rather than shied away from it. Many of them are overjoyed that they finally have a training/analysis partner available 24 hours a day. Vladmir no longer wakes up in the middle of the night with a great new novelty running through his head, picks up a phone, and yells "Nikolai! You must come over right now! I have an idea I want to test!" He instead makes a pot of coffee, sits down at a computer, fires up a database program to see if the idea's been played, and then uses a strong playing program to analyze the resulting position and to test the idea in practical play.

The net result: better grandmasters playing more beautiful chess games for all of us to enjoy.

This approach works for players of all levels. I remember (from my pre-computer days) a position from a book that was driving me nuts. The book evaluated the position as being won for White. I searched and searched and couldn't see a win. Then, after getting a PC, I first did a database search for the position. One game came up on the list, a game which I replayed, sitting back and enjoying the entertaining King hunt. Then I let a playing program chew on the position, setting it up to play game after game. White never lost, and now I finally understand the analysis.

SO WHAT'S MISSING?

The tools are all here: a primitive type of heuristic questioning in the form of chessplaying programs plus the database search functions. The number of different types of database searches has increased as well over the last couple of years. Now in ChessBase you can do position searches that include what's on the board, what's not on the board, and what might be on the board (for example, White's King on either f1 or g1). We can do material searches, right down to opposite-colored Bishops. We can even search for material sacrifices.

The hardware has made a difference too. Five years ago, when a 486/25 was considered a miracle of technology, 50,000 games was a big database. These days it's considered to be laughably small. Databases of more than a million games can be compiled and accessed quickly, easily, and cheaply. On my Pentium 133 I can do a search of 600,000 games off of a CD (which is slower than accessing them from a hard drive) in less time than it takes to go get a cup of coffee.

So what's missing? What keeps all of this from being a complete expert system?

One school of thought says "nothing". In Englebart's original vision, humans and computers worked in an integrated domain, sharing the workload. Humans defined the problem, computers helped them find the answers.

The tools to do that are in place. But researchers in expert systems want to take this one step further. They want the computer to be able to define the problem.

Let's look at an example. Suppose you're playing the White side of the Caro-Kann Advance and you keep getting your butt kicked game after game. The logical steps here are to do database searches to see what strong players are doing with the opening plus have playing programs do analysis of your games to see where your weakness lies.

Suppose, though, even after doing that, you're still unable to interpret the data and figure out the problem. Wouldn't it be wonderful to feed your games as well as coleected GM data to a computer and have it spit out an answer in plain English?

That's the second school of thought, the missing piece of the expert systems puzzle. Remarkable as the present tools are, they still can't slap one in the face and say "No, stupid! Not that -- this!!!" They can't take the workload completely off one's shoulders.

But help is lurking somewhere just over the horizon. Instructional software is getting better and better. I had my first look at Chess Mentor last week and I was really impressed. It's not every piece of software that can throw you a position, ask you what you'd play, and then explain why dang near any move you made is right or wrong. True, the analysis is pre-written, but I can see a day coming, possibly in my lifetime, when the computer will actually generate the analysis and dialog right on the spot. It will be able to analyze our weaknesses and provide us with corrective measures, improving our play.

Something a bit like this exists now for beginning to low-intermediate players. The program Power Chess will allow you to play a game against it and then provide you with spoken analysis of your play immediately following the game. The analysis is a bit superficial for stronger players but it's still a slick little program and worth the $18 bucks just to see the spoken analysis in action.

Programmers have taken their first baby steps towards giving us expert system chess tools for our PCs, but they've taken a big jump forward this past week toward proving that an expert program can be a match for the acknowledged best in a specialized field of endeavor. When a computer can teach a human something beneficial, the event becomes a cause for celebration, not a reason to regard the machine in the same light as Arthur C. Clarke's fictional creation.

No, Kasparov losing to Deep Blue isn't the end of the world. To crib from Aldous Huxley, I believe it's the start of a brave new world. If it's a world that will help us to become better at what we do, whether it's playing chess or performing chemical research or hunting for Apatosaurus remains, I for one am ready to embrace that world wholeheartedly.

READ ALL ABOUT IT

If you're the least bit interested in the topics we've just examined, here are some suggestions for further reading:

How Computers Play Chess, by David Levy and Monty Newborn; Computer Science Press, 1991. Probably the best mass-market book on the subject.

How to Get the Most from Your Chess Computer, by Julio Kaplan; R.H.M. Press, 1980. Hard to find, and some of the information is a bit dated, but an extremely helpful book nonetheless and worth the search.

The Brain Makers, by H.P. Newquist; Sams Publishing, 1994. How the commercial AI industry drove itself to bankruptcy promising what it couldn't yet deliver.

Expert Systems: Principles and Case Studies, edited by Richard Forsyth; Chapman and Hall, 1984. Very technical stuff, but interesting for the ideas alone if you care to wade through the math and technical jargon.

Artificial Intelligence and Human Learning: Intelligent Computer-Aided Instruction, edited by John Self; Chapman and Hall, 1988. Like the previous book mentioned, this one is very technical but the ideas discussed will make your head spin.

The Improbable Machine, by Jeremy Campbell; Simon and Schuster, 1989. What AI research has told us about ourselves, and why the human brain is so danged hard to understand. A very entertaining book.

I'd be very interested in reading your opinions of this (or any other) issue of T-Notes, as well as my book Battle Royale. I invite you to post comments to our ChessBase Users Group.