Can Computers Play Chess?
A guest post by Icepick
[Ed. note: Icepick agreed to my request to promote this to a post in J’s honor. As I told him, J was riveted by chess and was quite a strong player, although — as with writing and jazz — his true genius was as a generous and penetrating appreciator of other people’s brilliance. He could watch high-level chess play for hours. He didn’t play more because he feared falling into chess and never coming out. In the words of an old Russian proverb in a book of chess quotations he treasured: “Chess is a sea in which a gnat may drink and an elephant may bathe.”]
I have seen computers play pretty mean games of chess. Wait, let me check….
I have to date played 7605 games against the BulletB program on the Internet Chess Club. I’m currently +950. Disappointing, but those are 1 1 games, and I’m just not that good with a mouse, so my current ratio is probably about as good as I can hope for long term.
Based on my experience of playing a few thousands of games against people live, and another several thousand against people over the internet (still looking for a day where I can play against people on all seven continents in one day – personal best is five continents in one day) I think I can be a reasonable judge of whether or not a computer can or has played chess.
And the judge SEZ – Computers can and do play chess.
As for the Turing Test – too few human beings can pass it.
* * *
A further elaboration:
Machines have been built that do most PHYSICAL feets better than humans. (I believe John Henry got his ass whupped long ago.) Autonomous machines (robots) started getting used in manufacturing decades ago. And adding machines have been faster than (most) humans for a long time now.
But what happened over the last few decades with chess playing computers was a different matter altogether. Chess was chosen for a reason – namely that it was a very difficult game that didn’t come down to a simple formula. (Unlike, say, tic tac toe.) The variety of different pieces also separated the game in complexity from checkers. (Checkers HAS been largely beaten by the computers. Only Marion Tinsley could match them.) Chess requires memory (a specialty of computers), calculation (another specialty of the silicon monsters), pattern recognition (not a strength – or at least it wasn’t a few decades ago), and positional judgement.
That last was somewhat mysterious – the good chess players knew better than the poor ones when a position had something in it. The very good players had a better sense than the good ones. The very best were better still, and the true greats leave everyone baffled. But it’s hard to define what the “it” is. It was more than just pattern recognition, although that was part of it.
I remember an anecdote about Bobby Fischer. He was playing in some open tournament somewhere in his youth. He walked by the board of an acquaintance, looked at the position for maybe five seconds, and wandered off. Years latter he ran into the man again and asked him, “Did you win that game against so-and-so?” The man replied that he hadn’t won, so Fischer quickly explained to the man what he should have done to win the game. A mere glance at a complex position was enough for Fischer to judge the position accurately and formulate the winning plan. FIVE SECONDS. Such powers mystify us regular players.
Then there was the story of Capablanca explaining an endgame to some other masters. He stopped by an analysis session and started explaining the position to them, one they had been working on for some time at that point. He said something along the lines of “this piece should be here, and that one there, and then White wins easily.” They asked for variations and he replied that he didn’t need them – once he knew where everything went the rest was elementary. I’ve heard similar stories about (amongst others) Bisguier (an under-rated American GM), Kasparov, Tal, and especially Karpov. Even Kasparov (the greatest chess player ever – human player anyway) seems in awe of Karpov’s ability to just know where the pieces belong.
So this positional judgement thing was a great challenge. It was a perfect problem for the programmers.
In the early years they hoped to mimic a human approach to the game. That was the great hope of early pioneers, including Mikhail Botvinnik, the greatest of the Soviet players (and their longest tenured world champion). Botvinnik also happened to be an engineering whiz, and he hoped to construct a machine to mimic the human mind’s functions. He failed miserably.
Early on the programmers realized that the strongest element of a computer’s ability was the calculation aspect. So they put more and more effort into algorithms to evaluate positions along strictly mechanistic means, and to calculate as many positions as possible. No gestalt in this approach, nothing “holistic”. This approach bore fruit.
And it was ultimately this approach that led to the creation of programs that could best all but the strongest humans. And as it stands today, the humans really can’t hope to win a match save by glitch in the machine. They might tie (the match, not any individual game, which they still win on occasion) with best play, but that’s all they can really hope for. In about five years I doubt they will have any hope at all of even winning a single game off the best programs. It’s changed that much.
In addition, these programs have shown people new ways to play the game. Their calculating abilities have revealed any number of flaws in old approaches to specific positions. About ten years back they even started occasionally started showing new strategies and positional motifs in certain positions. (I remember a game with a Re3 lift that kind of blew everyone’s mind – computer programs weren’t supposed to DO that.) But it wasn’t understanding, it was more and more calculations.
These days the best of the younger grandmasters show unmistakable computer influence on their style of play – not just in the opening preparation but throughout the game. (I thinking most notably of Magnus Carlsen’s relentlessness and Hikaru Nakamura’s tactical wizardry.) The computer programs have changed the game. (The old-timers generally don’t seem to like it – they believe it has removed much artistry. I don’t entirely agree, as their is much beauty in the new stuff as well. Times change.)
But the sad upshot of all of this has been that computers don’t think like people when playing chess. The more … call it … holistic approach never worked. Botvinnik would have been disappointed, as is his latter successor Kasparov. The big loss isn’t for chess, it’s for understanding how human thought works, and determining if it can be duplicated. Chess just didn’t hold the answer.
OTOH, we have observed two things in the meantime. First, that computers CAN solve extremely complex problems through systematic refining of brute force algorithms. Second, that humans can learn to mimic that approach, within limitations. So some things have been learned.
But there is still hope that the computer guys can do something different. The hope now rests on coming up with good Go programs, as that game turns out to be much more difficult for programmers to figure out. I’ve heard they hope to have some success with poker playing programs, but I expect that to be a bit less worthwhile. Poker can be reduced to mathematics fairly easily (for a computer at least), and some clever application of game theory ought to insure enough variability that I expect such a program to be better than pretty much all humans in the not distant future. (Assuming it hasn’t happened yet. I only follow that peripherally.)
But the upshot has been this – the brute force “materialist” approach works. And I mean that both from a chess esthetics viewpoint (computers, like Victor Korchnoi, are suckers for grabbing all the material they can get) and as a programming solution to complex problems.
* * *
(The BulletB program I mentioned is one that has been “lobotomized” to play at around a certain strength at certain time limits. The Crafty program upon which it is based is a pretty solid free-ware program developed primarily by Dr. Robert Hyatt. Interestingly for me, the program’s origins date back to 1968, the same year as my birth. I’ve been lucky enough to play a few games (fewer than five, I think) where I have managed to draw a non-lobotomized version of Crafty. )
* But the humans are catching up. There’s a reason that we have so many people who’ve become grandmasters at 13 these days. Playing programs and especially databases have drastically altered the learning curve. The kids are getting programmed too these days.
* * *
H. G. Wells’ description in this piece is wonderful. Here’s the best extract:
The passion for playing chess is one of the most unaccountable in the world. It slaps the theory of natural selection in the face. It is the most absorbing of occupations, the least satisfying of desires, an aimless excrescence upon life. It annihilates a man. You have, let us say, a promising politician, a rising artist, that you wish to destroy. Dagger or bomb are archaic, clumsy, and unreliable–but teach him, inoculate him with chess!
I must say, however, that the game does have some charms. The Dutch GM Hans Ree once said, “Chess is beautiful enough to waste your life for.” If you don’t understand that statement, if you don’t feel it in your soul, you aren’t really a chess player.
About three years ago I started entering all my existing live game scores into a database. I forget what the total is, but it’s around 300 games, if memory serves. Those are just the live tournament games, plus some games from high school matches, and together they span 27 years now. Not a great big number, not by any stretch. I’d play every weekend, given the opportunity, but they haven’t always presented themselves. But I played regularly with a couple of diehards back in my Maryland days. One of them has played in over 655 EVENTS since late 1991, and the other has played in over 1273 EVENTS in the same period of time. They’re both lifetime Class B players at best. But they’re out playing every chance they can get. And that’s just the rated events they PAY to play in. I can find one of them online just about every day, looking for more action. I just checked a recent tournament they both played in, and I recognized every player in the tournament. Dedication, baby!
So the short bit is that yes, it can dominate/ruin one’s life!
* * *
Another proverb from J’s book of chess quotations, this one German:
No fool can play chess, and only fools do.
What You May Not Know About Your Health Insurance
A guest post by Icepick
Recently I heard President Obama once again claim that his health care/health insurance reform is needed to stop the growing costs of health insurance. He has repeatedly implied that health insurance costs are rising mostly because health insurance companies are raising rates to increase their profits. This is ill-informed at best. Health insurance costs are rising because health CARE costs are rising.
I know this because of personal experience. I used to have the fancy title of Senior Benefits Planning Analyst at a Fortune 100 company. I did the financial modeling of that company’s employee benefits costs for ~54,000 full-time employees, including the medical costs. Our department did a lot of research into the costs of health care, in part so that we could make decent forecasts, and also so we could give corporate leadership recommendations on how much we should charge for medical coverage. That’s right, the company set insurance rates (sometimes after collective bargaining, sometimes not), not some evil insurance company. That’s because the company was self-insured.
Something that the President never mentions is that most large companies self-insure their medical plans – they pay most or all of the medical expenses of their covered employees and families themselves. Let me explain it another way. If a company purchases insurance for an employee, the company only pays for the insurance which is typically a fixed cost. If that employee’s medical expenses are less than the insurance, the company loses money on that employee’s insurance. In a self-insured plan, the company pays those medical costs itself. Once a certain threshold in size is passed a company often finds it less expensive to self-insure.
Most people working for a self-insured company won’t know that the company has self-insured. Usually a company will hire an insurance company to handle the actual claims. Companies do this for several reasons. First, privacy concerns and HIPAA law mean that a company does not want direct access to its employees medical information. Second, most companies do not have the expertise to handle insurance claims. Third, insurance companies typically have more power to negotiate favorable rates with local providers. So a self-insured company will pay an insurance company to handle the administrative side of things, while the company pays for the actual medical expenses.
When I worked for such a company, we analyzed our medical and insurance costs in excruciating detail. And you know what? Our medical costs were increasing at roughly 9% a year for at least a decade. And the insurance company costs had little to do with it. Our costs were going up at that huge rate because medical expenses were going up that fast.
So all the talk of insurance companies fleecing the public is at best a side issue – insurance costs have increased in recent decades because medical expenses have increased. The current bill, which will no doubt be passed this weekend, does nothing to truly address that issue.
Ambiance 