January 6, 2009

Cozying up with Deep Blue: A SentDev Classic

"Advanced Chess" pitting computer-human teams against each other shows how humans can avoid obsolescence through symbiotic relationships with technology

Several weeks ago, while bored on a commuter train, I decided to pull out my Palm Pilot and play a game of chess. Seeing as I had no one to play against, I decided to try my hand against the computer. I was quite confident that I'd have little difficultly keeping up—it's hardly Deep Blue, after all.

I arbitrarily picked an average difficultly level and proceeded to get my ass kicked in frighteningly short order. Somewhat discouraged, I then tried at the easiest level. Once again, I suffered an embarrassing thrashing.

With my dignity soiled, I vowed to improve my chess skills. I wasn't going to let some puny Palm Pilot beat me at chess. I dusted off an old chess manual and practiced some standard openings and strategies. I can now proudly say that I can beat my handheld at level 5. My goal is to beat it at level 8, maximum difficulty.

Playing a computer at chess can be rather humbling. As you're waiting for it to make its move, watching the "thinking" progress bar move from left to right, it's daunting to consider how many moves it's evaluating. I'm happy if I can think three to four moves ahead. The computer can contemplate thousands every second.

I'm sure Garry Kasparov felt the same way back in 1996 when pitted against Deep Blue. Now that computer could crunch the numbers. Written in C and running under the AIX operating system, Deep Blue was a massively parallel, 30-node, RS/6000, SP-based computer system enhanced with 480 special purpose VLSI chess processors. Odds are those stats are meaningless to you, but this one shouldn't be: This mother could crunch 100,000,000 positions per second.

100,000,000 positions per second!

It's a wonder that Kasparov could play against it at all. Of course, there's more to chess than just raw computation. It's a game of subtlety, nuance and sophisticated psychology and strategy—elements that are far beyond the capabilities of even the most powerful computers. In fact, prior to Kasparov's defeat, some chess experts maintained that computers would never be capable of defeating grandmasters. But thanks to Deep Blue and its successors, we all know that this is in fact possible.

Kasparov's loss was indeed a deep shock to the chess world. It was a significant milestone in the history of chess, not just because a reigning world champion finally lost against a computer, but because of the ramifications to the game itself. Did Kasparov's loss signify the beginning of the end for meaningful human interaction in professional chess? Would future tournaments see humans as mere spectators to machines?

More broadly, did Deep Blue's intrusion into a previously sanctified human realm represent the beginning of a larger trend? If computers could now defeat even our grandmasters, what else might they be capable of? Indeed, the steady onslaught of Moore's Law and breakthroughs in parallel processing has some fearing the rise of AI and the subsequent delegation of human minds. Are Homo sapiens poised for obsolescence and even replacement?

Well, if Kasparov has his way, the answer is no—and not because he feels that humans can continue to compete with computers. Rather, Kasparov believes the future of chess can be advanced through the cooperation of computers with humans. Consequently, Kasparov's idea of Advanced Chess, where human-machine teams compete against other human-machine teams, offers an effective framework for how humanity as a whole should manage its ongoing relationship with its advancing technologies. To avoid replacement, we need to establish a symbiosis with our technologies and create something greater than the sum of its parts.

Computer chess vs. human chess

In all fairness to Kasparov and other expert chess players, computers still aren't able to consistently defeat their human counterparts. After losing to Deep Blue in the first game, Kasparov rebounded by winning three games and drawing two, defeating it by a final score of four to two. Kasparov lost the 1997 rematch, but managed a draw against its successor, X3D Fritz in 2003. Similarly, grandmaster Vladimir Kramnik tied Deep Fritz in an eight-game tournament a year earlier. As it currently stands, the tables are quite even in terms of what the best computers can do against the best players.

But what's interesting is not so much the parity; it's that humans and machines play chess so differently yet still come up even. Computers and humans have unique weaknesses that are clearly offset by their strengths.

It's generally acknowledged that computers are superior calculators, while humans are better at long-range planning. Computers cannot be psychologically intimidated (something Kasparov does very well against his human opponents), nor are they capable of suffering from fatigue or other physical problems (during the 1984 World Championships, for example, Anatoly Karpov lost 22 pounds and was hospitalized several times as he battled Kasparov in a protracted tournament that saw them play well over 30 games). Computers are also immune to making silly mistakes (Kramnik lost game five against Fritz after making a severe blunder).

Humans, on the other hand, can plan, bluff and, most importantly, adapt. Kasparov, in all his encounters with computers, tends to finish more strongly than he begins. Even in my own clashes against my Palm Pilot, I have noticed that my computer opponent gets quite messed-up when I open with the Queen's Gambit. Consequently, that's now my standard opening against it. The Palm, on the other hand, cannot learn from my mistakes, and has no idea that I fare very poorly in end game scenarios.

Computers are also quite poor at recognizing when something is irrelevant. During its first match against Kasparov, for example, Deep Blue eliminated an inconsequential pawn at a critical point in the game. It's thought that Deep Blue sensed no threat from Kasparov at the time and that the move wouldn't detract from the attack it was developing at the other side of the board. It was merely being mindlessly methodical by claiming the material.

Assistive devices

In consideration of these differences and unique strengths, it's safe to say that the best chess playing entity in existence today is neither a computer nor a human, but rather a computer and a human working together. As Albert Einstein once remarked, "Computers are incredibly fast, accurate and stupid; humans are incredibly slow, inaccurate and brilliant; together they are powerful beyond imagination."

Indeed, computers have changed the face of chess—not just because they have proven to be formidable opponents, but because they can also act as potent assistive devices. Grandmasters now use them extensively for planning and practice. Exhaustive hash tables have been generated by computers that map virtually all end game scenarios involving up to five pieces. Scenario analysis is now possible at an unprecedented scale, including backward analysis (starting from a position with a large edge and moving back to a starting position) to find new branches worth analyzing, and multi-variation analysis mode to examine alternate tries worthy of analysis.

Simply put, not using computers to assist in chess play would be as silly as not using calculators to help us do math. Further, when looked at as prostheses, computers clearly expand human capacities, helping us take our activities and disciplines to the next level. They enable us to partake in endeavors that were previously cognitively impossible.

Recognizing this, Kasparov proposed a new form of competition during the late 90s. Inspired by his matches against computers, Kasparov felt that humans and computers should cooperate instead of contending with each other. Called "Advanced Chess," the new style of play would see human players team-up with a computer and compete against another man-machine unit.

Kasparov got the ball rolling by organizing a six-game Advanced Chess match against Veselin Topalov in June of 1998, with Kasparov using Fritz 5 and Topalov using ChessBase 7.0. The match ended in a three-three draw. Kasparov commented afterward, "My prediction seems to be true that in Advanced Chess it's all over once someone gets a won position. This experiment was exciting and helped spectators understand what's going on. It was quite enjoyable and will take a very big and prestigious place in the history of chess."

Since this initial match, Advanced Chess tournaments have been scheduled annually in Leon, Spain. Grandmaster Viswanathan Anand, the winner of three titles, is currently considered the world's best Advance Chess player. After losing to Kramnik in 2002, Anand commented, "I think in general people tend to overestimate the importance of the computer in the competitions. You can do a lot of things with the computer but you still have to play good chess...I don't really feel that the computer alone can change the objective true to the position."

Expanding on Anand's point, advocates of Advanced Chess argue that the strength of a player does not come from any of the components of the human-computer team, but rather from the symbiosis of the two. The combination of man and machine results in a "player" that is endowed with the computer's extreme power and accuracy and the human's creativity and sagacity.

Ultimately, the combined skills of knowledgeable humans and computer chess engines can produce a result stronger than either alone. Advanced Chess has resulted in heights never before seen in chess. It has produced blunder-free games with the beauty and quality of both perfect tactical play and highly meaningful strategic plans, and it has offered chess aficionados remarkable insight into the thought processes of strong human chess players and strong chess computers.

Cooperation and merger, not obsolescence

With the rise in prominence of computers in the chess world, Kasparov refused to throw up his hands in despair and declare the end of human involvement in the game. Instead, he devised a new activity that would combine the best of what the digital world had to offer with that of the biological. The result was something greater than the sum of its individual parts.

The rest of society should learn from this example. Naturally, people are growing increasingly wary of supercomputers and the potential for AI; it's understandable that people fear a future in which humans are replaced by machines. But as the example of Advanced Chess shows, that's not necessarily what's going to happen. The development of AI and other information technologies will continue to advance based on how we choose to adapt to them and how they adapt to us. Further, human control over where and how advanced technologies develop will have a significant impact on the kinds of collaborative and symbiotic systems that emerge.

Thanks to human ingenuity, our disciplines, activities and goals will continue to change and evolve, taking the human experience to unprecedented places as we become capable of things never before possible.

Like beating my Palm Pilot at level 6.

The article was originally published on March 19, 2005.

2 comments:

Singe said...

Though Kasparov probably has done the most of any person to promote the concept of computer-assisted chess competition, it should be noted that the concept was originally thought up by Vernor Vinge!

It makes a notable appearence in his 1984 book, "The Peace War".

Interstellar Bill said...

Your answering machine doesn't talk, you do.
Deep Blue didn't defeat Kasparov, its programmers did.
Give Kasparov another such computer and he'd be invincible.
The computational theory of mind is an incoherent philosphical disaster hypnotizing the entire 'AI' field with an impossible fantasy -- HAL9000, the newest embodiment of the genie myth.
The inescapable brute fact is that dead things can't think and never will. Thinking computers are even less plausible than chimp ballet stars. At least the chimps can dance, however terribly.
Just as AI having had zero accomplishments in 50 years only fires up the faithful even more, so too will the next 50 years of abject failure make no difference. Bad philosphy trumps facts every time.
Kiss C3PO and the Singularity goodbye, just as you let go of Santa Claus when you reached the age of reason.
Calculators and computers marvelously assist human thinking but will never produce it.