 |
|
|
|
|
|
| |
Origin >
How to Build a Brain >
Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
Permanent link to this article: http://www.kurzweilai.net/meme/frame.html?main=/articles/art0527.html
Printable Version |
|
| |
|
Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
The Deep Fritz computer chess software only achieved a draw in its recent chess tournament with Vladimir Kramnik because it has available only about 1.3% as much brute force computation as the earlier Deep Blue's specialized hardware. Despite that, it plays chess at about the same level because of its superior pattern recognition-based pruning algorithm. In six years, a program like Deep Fritz will again achieve Deep Blue's ability to analyze 200 million board positions per second. Deep Fritz-like chess programs running on ordinary personal computers will routinely defeat all humans later in this decade.
Published on KurzweilAI.net October 19, 2002
In The Age of Intelligent Machines
(MIT Press, 1990), which I wrote in 1986-1989, I predicted that
a computer would defeat the human world chess champion by the end
of the 1990s. I also noted that computers were gaining about 45
points per year in their chess ratings whereas the best human playing
was essentially fixed, which projected the cross-over point at 1998.
Indeed, Deep Blue did defeat Gary Kasparov in a highly publicized
tournament in 1997.
Now with yesterday's final game, we have the current reigning computer
program, Deep Fritz, able only to achieve a 4-4 tournament tie with
world chess champion Vladimir Kramnik. It has been five years since
Deep Blue's victory, so what are we to make of this situation? Should
we conclude that:
- Humans are getting smarter, or at least better at chess?
- Computers are getting worse at chess?
And if we were to accept the latter, should we conclude that:
- The much publicized improvement in computer speed over the past
five years was not all it was cracked up to be? Or,
- Computer software is getting worse, at least in chess?
The specialized-hardware advantage
None of the above conclusions are warranted. To gain some insight
into these questions, we need to examine a few essentials beneath
the surface of the headlines. When I wrote my predictions of computer
chess in the late 1980s, Carnegie Mellon University was embarked
on a program to develop specialized chips for conducting the "minimax"
algorithm (the standard game-playing method that relies on building
trees of move-countermove sequences, and then evaluating the "terminal
leaf" positions of each branch of the tree) specifically for
chess moves.
Based on this specialized hardware, their 1988 chess machine HiTech
was able to analyze 175,000 board positions per second and achieved
a chess rating of 2,359, only about 440 points below the human world
champion.
A year later in 1989, CMU's "Deep Thought" increased
this capacity to 1 million board positions per second and achieved
a chess rating of 2,400. IBM eventually took over the project and
renamed it "Deep Blue," but kept the basic CMU architecture.
The version of Deep Blue that defeated Gary Kasparov in 1997 had
256 special purpose chess processors working in parallel, which
analyzed 200 million board positions per second.
An important point to note here was the use of specialized hardware
to accelerate the specific calculations needed to generate the minimax
algorithm for chess moves. It is well known to computer systems
designers that specialized hardware generally can implement a specific
algorithm at least 100 times faster than programming the same algorithm
as conventional software on a general-purpose computer. ASICs (Application-Specific
Integrated Circuits) require significant development efforts and
costs, but for critical calculations that are needed on a repetitive
basis (for example, decoding MP3 files or rendering graphics primitives
for video games), this expenditure can be well worth the investment.
Prior to the time when computers could defeat the best human players,
there was a great deal of focus on this milestone, so there was
support for investing in special-purpose chess circuits. Despite
some level of controversy regarding the rules and procedures of
the Deep Blue-Kasparov match, the level of interest in computer
chess waned considerably after 1997. After all, the goal had been
achieved, and there was little point in beating a dead horse. IBM
cancelled work on the project, and there has been no work on specialized
chess chips since that time.
Computer hardware has nonetheless continued its exponential increase
in speed. Personal computer speeds have doubled every year since
1997. Thus the general-purpose Pentium processors used by Deep Fritz
are about 32 times faster than personal computer processors back
in 1997. Deep Fritz uses a network of only eight personal computers,
so the hardware is equivalent to 256 1997-class personal computers.
Compare that to Deep Blue, which used 256 specialized chess processors,
each of which were about 100 times faster than 1997 personal computers
(of course, only for computing chess minimax). So Deep Blue was
25,600 times faster than a 1997-class personal computer for computing
chess moves, and 100 times faster than Deep Fritz. This analysis
is confirmed by the reported speeds of the two systems: Deep Blue
can analyze 200 million board positions per second compared to only
about 2.5 million for Deep Fritz.
Thus the primary problem with Deep Fritz is that it is much slower
than Deep Blue. However, the reason for this is the use of specialized
hardware in Deep Blue, and the lack of it in Deep Fritz. This reflects
the relatively low priority we've given to chess machines since
1997. The focus of research in the various domains spun out of artificial
intelligence has been placed instead on problems of greater consequence,
such as guiding airplanes, missiles, and factory robots, understanding
natural language, diagnosing electrocardiograms and blood cell images,
detecting credit card fraud, and a myriad of other successful "narrow"
AI applications.
Significant software gains
So what can we say about the software in Deep Fritz? Although chess
machines are usually referred to as examples of brute-force calculation,
there is one important aspect of these systems that does require
qualitative judgment. The combinatorial explosion of possible move-countermove
sequences is rather formidable.
In The Age of Intelligent Machines, I estimated that it would take
about 40 billion years to make one move if we failed to prune the
move-countermove tree and attempted to make a "perfect"
move in a typical game (assuming about 30 moves in a typical game
and about eight possible moves per play, we have 830
possible move sequences; analyzing one billion of these per second
would take 1018 seconds or 40 billion years). I noted
that this would not be regulation play, so a practical system needs
to continually prune away unpromising lines of play. This requires
insight and is essentially a pattern-recognition judgment.
Humans, even world class chess masters, perform the minimax algorithm
extremely slowly, generally performing less than one move-countermove
analysis per second. So how is it that a chess master can compete
at all with computer systems that do this millions of times faster?
The answer is that we possess formidable powers of pattern recognition.
Pattern recognition incidentally is my principal area of technical
interest and expertise, and is, in my view, the primary basis of
human intelligence. Thus we perform the task of pruning the tree
with great insight.
After the Deep Blue-Kasparov match, I suggested to Murray Campbell,
head of IBM's Deep Blue team, that they replace the somewhat ad
hoc set of rules they used for this pruning judgment task, and replace
it with a well- designed neural net. All of the master games of
this century are available on line, so it would be possible to train
these neural nets on a considerable corpus of expert decisions.
This approach would combine the natural advantage of machines in
terms of computational speed with at least a modest step towards
more sophisticated pattern recognition. Campbell liked the idea
and we were getting set to convene an advisory group to flesh out
the idea when IBM cancelled the project.
It is precisely in this area of applying pattern recognition to
the crucial pruning decision that Deep Fritz has improved considerably
over Deep Blue. Despite Deep Fritz having available only about 1.3%
as much brute force computation, it plays chess at about the same
level because of its superior pattern-recognition-based pruning
algorithm.
So chess software has made significant gains. Deep Fritz has only
slightly more computation available than CMU's Deep Thought, yet
is rated almost 400 points higher.
Are human chess players doomed?
Another prediction I made in The Age of Intelligent Machines was
that once computers did perform as well or better as humans in chess,
we would either think more of computer intelligence, or less of
human intelligence, or less of chess, and that if history is a guide,
we would think less of chess. Indeed, that is what happened. Right
after Deep Blue's victory, we heard a lot about how chess is really
just a simple game of calculating combinations, and that the computer
victory just demonstrated that it was a better calculator.
The reality is slightly more complex. The ability of humans to
perform well in chess is clearly not due to our calculating prowess,
which we are in fact rather poor at. We use instead a quintessentially
human form of judgment. For this type of qualitative judgment, Deep
Fritz represents genuine progress over earlier systems.
Incidentally, humans have made no progress in the last five years,
with the top human scores remaining just below 2,800. Kasparov is
rated at 2,795 and Kramnik at 2,794.
Where we go from here? Now that computer chess is relying on software
running on ordinary personal computers, they will continue to benefit
from the ongoing acceleration of computer power. In six years, a
program like Deep Fritz will again achieve the ability to analyze
200 million board positions per second that was provided by Deep
Blue's specialized hardware. With the opportunity to harvest computation
on the Internet, we will be able to achieve this potential several
years sooner (Internet harvesting of computers will require more
ubiquitous broadband communication, but that's coming too).
With these inevitable speed increases, as well as ongoing improvements
in pattern recognition, computer chess ratings will continue to
edge higher. Deep Fritz-like chess programs running on ordinary
personal computers will routinely defeat all humans later in this
decade. Then we'll really lose interest in chess.
| | |
|
| |
Mind·X Discussion About This Article:
 |
|
|
 |
Re: Having raised the Chess bar...
|
|
 |
|
Although computer versus man is an interesting sidebar issue, chess, tennis, the Super Bowl, etc., is a competition between humans, and all the associated facets of being human that go with it. Someday a machine might be made that can retrieve and return a tennis ball better than any human alive, and summarily beat any and all comers. While technologically significant, it lacks "sporting" appeal. Likewise, a robot arm can be made that obtains the world arm wrestling championship. Etc., etc. |
|
 |
|
 |
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
> All possible combinations in a chess game (assume that it continues 40 moves) are far far more.
Doesn't matter at all!
Be cause we (orga & mecha) computers, can group large amounts of moves under one hat, one case . That is what Ray calls - pattern recognition. Some instant chunk solutions, for a zillions of moves.
How many moves are possible, when only two kings, two white bishops and two black knights are on the board?
More, than there are atoms in the Moon. Yet, there is about 1 KB long (mecha) program, which perfectly plays, what almost no human is able to. That's white (with the bishops) wins, after more than 200 hundred moves!
Well, this case is a clear computation matter. Which chess pieces must be introduced, that there is "more than a computation case" - Brigadira?
If you can't answer me that, you have answered - everything! :)
- Thomas
|
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
Among the rumors about the 1997 Kasparov-Deep Blue match that I've heard:
1) Kasparov played "human chess." He resigned in a drawn position that his counterparts pointed out. Kasparov also made a complex blunder on move 6 of one of his games in a Caro Kahn line that he later admitted was pure foolishness give he had studied that line extensively.
2) Deep blue had master games on its system, including a great deal on Kasparov himself. Knowing a player's weaknesses is an incredible advantage, period.
Bottom line, you can train a computer to think fast and human like, but it will be a long time before computers will be able to beat the best human players, if ever. Eventually we will have computers analyzing trillions of positions per second and have databases of grandmaster games that would rival any archive. Then they might win....
But consider this: when a human is forced to think for long periods of time he gets tired. When he's under stress he gets fatigued easily. This, and only this is why Kasparov lost. The computer was not a superior player to Kasparov in 1997. As we've seen, players like Kasparov are indeed amazing, and their experience and skill have not allowed another loss to a computer by a World champion. Computers are as smart as humans make them, and that is an encouraging thought for the chess world.
|
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
Deep Junior is running on a SMP machine with 8 1.6 GHz P4s... significantly more than a laptop.
Also, the way that computers prune the tree is actually a weakness. Perhaps it's something that will be corrected, but it might be very difficult.
The problem is something like this: The computer threatens a piece. Assume there's a Knight on g5, the computer just played h6. It's calculated that if the Knight doesn't move, it can capture it. After looking at the capture, most programs use a technique called null-move heuristic. It gives the opponent a free move, checks that things are still bad, then discards the move it was testing. So in this way, it "assumes" that the Knight will move. Instead the human plays h4, defending the Knight with a pawn. The computer sees this as a free knight and captures. Unfortunately, after the moves it pruned due to the null-move heuristic, there's a serious threat. No checks, but a few pawns advance and suddenly the computer's King is unable to go anywhere and it's over.
There's a few games on chessbase's site, and there's a book by Ernest F. Pecci. Md. that demonstrate this.
If you remove null-move pruning, or in general, stop pruning every "bad" move then you're back to calculating for 40 million years.
Another note is that chess programs are generally designed based on established classical chess theory. What if it's possible to ignore all of the classical ideas in chess and slowly build a crushing attack? Most anti-computer ideas seem to be based on something similar. Get a closed center, slowly buildup pieces against the computer king, advance pawns, even use the king to attack if necessary... computers fail because they don't understand what's going on. |
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
This is probably an oversimplification, but it reminds me a lot of when I was at one small company. In the break room was a cabinet video game. Specifically, Street Fighter II.
While nowhere near the level of sophistication or strategy of chess, it held a very interesting lesson, one that others pointed out, one Kramnik knew and Kasprov didn't: Playing a computer is still different than playing a human.
While button mashing during break times, I was getting quite frustrated at the game, until I decided to replace away complex strategies and maneuvers for simple shortterm gains. The computer had very little learning ability, and the different paths it took were very recognizable. A human would break the pattern quickly, but I was able to defeat the game using this simple set.
Of course, there was someone there that was a master at the game, and had perfect countermoves for any strategy I had. And the only way I could sometimes, just barely, defeat him was to use psychological tactics, to telegraph my old attacks and rely on his counterattack. Such tricks would never have fooled the computer opponent, which did not have the biases and assumptions.
This has been said almost ad nauseum, but one point of contention seems to be that chess can be brute-forced in a very unhuman manner. One possible solution is to change the game.
Disenchanted ( http://www.disenchanted.com/dis/technology/solved.html ), which is how I found this article, mentions Go, another thinking man's game. If Chess is a battle simulation, then Go is a global campaign simulation.
It's very heavily dependant on pattern recognition. Winning is defined by the area a side sorrounds, either partially or fully, but does not occupy. Direct invasions simply won't work.
Even the resolution is less than binary. There is no King to check or mate. There is always an open spot and the limitation of moves is less rigid. Instead, the game ends with both sides pass and agree that they cannot improve their position. The number of possible outcomes is literally infinite.
As such, from what I've heard, current Go alogarythms, even well-funded ones are limited to beginner and intermediate levels, unable to comprehend the not just strategy, but philosophy.
Maybe that's the next step. Computers that, after digesting the games can explain (Graphically with peices at least, let's not get into chat-bots) what are good moves, and why in a way that makes sense not just statistically, but logically. We don't need just good players, but good teachers. |
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
I think Ray Kurzweil made very interestic topic. But chess is not so easy as it looks. Why? Let's look.
1) We have an example how Ken Thompson from Belle Laboratories used his head and made great job about discovering ALL possible position with 5 pieces on the board (ex. (1)White King + (2)Black King + 3 pieces). Using these endgame databases a computer will play them absolutely perfectly ("like a God"). In any position with the given material on the board it knows instantly whether it is a win, draw or loss and in how many moves. Often it will announce a win or mate in over fifty moves (!). On the losing side it will defend optimally. Now Thompson is currently generating with 6 pieces.
And now very important thing: This was possible because he can used the fastest computing power. And if he has had discover what will be the final score with 7 pieces he will be have to take 10 times faster machine then we had.
And I should add that that was ONLY 6 pieces and the more the pieces - the faster exponential progress. It means that computing all is not possible (not in nearest 1000 years).
Maybe when we programmed machine to recognize the pattern then it will be playing much more stonger than today (Top machines like Fritz and Junior reached 2700 ELO level! = like 5% of top GrandMasters!).
I think if it will be done, when machine will be World Champion next day after programming it.
2) Humans can plan and machine can't. Both ways (computing) and planning LEAD to the same result! But if we (human) can plan our game is some cases we can do it up to 40-50 moves ahead (it depends from position, sometimes we can plan it even to hundred! I mean Top Players), then the computer has no chance with it. It must compute. Now the comp is "thinking" 13-14 moves ahead and it gives him power of play 2700.
3) Humans are fragile units. They are emotionally senitive and they are tired, nerves, sad, etc. Computer has not it! That will be one of the most important reason why comp will be the strongest, not his "geuine" play!
4) I don't think that people will lost interest after the situation when World Chamopion lose to comp. The same case was in 1997 with Kasparow, when he lose to Deep Blue. And what? Now he's competing against israeli program Junior in New York. And even if he lose the interest in chess will not (dramatically) decrease.
5) In last years computers with specialized chess programs helps people (the most the Top Players) to compete with people. It is not disturbing that computers are better than 99% people in the world. People are still playing. Why? Because they like this game. They do not know what will be the outcome before they play. And the game surive because it was needed (like a spoon) and people wanted to play it! If the solution will be known (ex. White started and wins in all possible continuations in 1350 moves), then we can talk what will be next. But this will not be in next 1000 years (exept when we made significant progress in AI in chess, like pattern recognizion) and even if it (ever) be people will be competing still. I am quite sure they will be. Most of them want to play and some of they love this game. |
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
I, too, believe that deep fritz will be able to defeat a world chess champion after ten years or so; however, I don't really think that a person like Vladimir Kramnik is the best chess player in the world. Just look at his recent results! he is hiding from his opponents. no linares! no big tournament?! Everybody knows that after beating kasparov in london, Kramnic avoids playing with his fellows. look at his result in Corus 2003. He lost 3 or so games!! That's funny but he really played bad against deep fritz match. He could've won 5.5-2.5.
If we could skip all the claims the author makes in his article, we can't definitely skip his last sentence. "we will lose interest in chess.", this sencence is shallow! We, chess players, whether weak, like me, or strong like Kasparov, play chess for its fun, its are, and maybe money. We are not gonna lose interest in chess just because we can't beat computers. chess is a game for men not for machines. Let's not forget this.
|
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
This article is suffering from a lack of understanding of the computer chess world.
First, there is no way to justify saying that Deep Fritz plays chess at about the same level as Deep Blue. Yes, Deep Fritz is stronger than Deep _Thought_ was, but you have to put that into perspective. Fritz is programmed by Franz Morsch, who has been a professional chess programmer for decades and has been refining his program the entire time, often with the help of strong chess players. On the other hand, Deep Thought was an experiment, cobbled together in a short amount of time primarily by a student (F. H. Hsu) who was very bad at chess. It suffered from severe bugs and a simplistic evaluation function. Nobody should be surprised that Fritz is stronger when searching the same number of positions per second (nodes per second, NPS).
Deep Blue was almost completely different from Deep Thought. Its evaluation function was far more sophisticated, and it was tuned (and partially designed) by a chess grandmaster. Hsu estimated that a software implementation of Deep Blue's evaluation function would require 40,000 instructions per position, whereas Fritz uses 1,000 to 2,000. During informal testing, a Deep Blue chip was slowed down to search the same number of NPS as a top PC program (possibly Fritz) and it still won 10-0.
In other words, there's every indication that Deep Blue's "pattern recognition algorithm" (evaluation function, which has little or nothing to do with tree pruning) was far superior to Deep Fritz's and Deep Blue was a significantly better chess player.
Regarding using neural networks in chess programs, several people have tried and the attempts have been abject failures when compared to conventional programs. Changing programs to behave more like humans has always perplexed me, as almost all humans are much weaker than programs...
Another issue is the implication that people are losing interest in chess. Perhaps the general population does not have the chess fever that they did back in '96 and '97, but United States Chess Federation membership has been steadily increasing and because of the open source movement and the Internet, the number of people writing chess programs has increased exponentially over the past few years. Contrary to Kurzweil's assertion, some people are currently working on chess chips (with FPGAs). If anything, interest in chess and computer chess has been increasing, despite the fact that computers have been able to beat the vast majority of humans for more than a decade (another point that was not acknowledged by the article).
-Tom Kerrigan |
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
There is nothing wrong with a computer Grand Master, although other aspects are important in a commercial computer program, and DJ probably does not yet play as well as the results of the games would let us think.
I use Fritz 8 and have it on my (single processor) laptop computer. It calculates 700,000 moves per second and plays at a level far above my own. This has not reduced my interest in chess. On the contrary, having constant access to a much better player has improved my chess. At least I hope so.
Fritz 8 can do other things which help me improve my chess. It can analyse my other games and suggest alternate, better moves. It talks to me during the game. It probably insults my play less than it could. Clearly, there are other important issues in commercial computer software, and focus on ELO is only one aspect.
The programmers of Deep Junior have agreed that DJ plays at a level of around 2500-2600 but never tires and never makes an "error", which probably brings the level of play of the machine to 2800. Looking at the games played so far, in two games, Kasparov played arguably better than DJ but either drew or lost following an error on his part. My feeling is that we are not yet at the point where the computer plays better or even as well as the human player; we are at the point where the computer can just barely hold the game until the human makes an error. Of course, this is a fine line, given the importance of errors in chess.
As an aside, I'd like to mention that Fritz and Junior are *not* the only chess programs out there. Dozens of high quality programs are available and are being actively worked on: Hiarcs, Chess Tiger, Crafty, Chess Genius, Shredder, Nimzo and I apologize if I haven't mentioned yours, as I should have. The chess software programming community is very alive.
I can only hope that chess computers will improve to the point where they play all aspects of the game better than humans. I also hope that software companies will integrate options into their commercial software which helps me improve my chess. There have always been people who play better chess than myself: what is the difference if I have one at home ? I personally doubt this will kill *research* into computer chess software, but it certainly will not kill chess.
|
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
I believe what Kurzweil said about "losing interst in chess" is from the computer science
perspective, not from the human enjoyment of playing the game.
Human GMs have much more chess knowledge than Deep Blue, Deep Fritz, or Deep Junior
combined! This is reflected in the fact that they can defeat these machines at all.
I believe the day is near when humans will not be able to defeat the machines. But, I also
believe if human GMs play for draws, computers could not beat humans, now or in the
future.
I would be much more inpressed with computers that only evaluate one or two
moves per seconds (as humans) and could still defeat the GMs of the chess world.
Or, to create a general purpose AI computer that could decide that it wanted to learn
the game of chess, teach itself how to play, study games on the Net, have a fear of
losing to a human, and still be able to defeat the top GM's of the chess world!
May the Forced mate be with you!
|
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
Some interesting comments, though I didn't get to them all. Here are some points to ponder...
1) Will we get to the point where there is some type of combination between chess players and computers, meaning that the chess players consult their own computers when they play?
2) If it ever gets to the point where programs are installed in our brains increasing our caluclating power, will that be a victory for the machine or the man?
3) What are the chances of humans beating computers at other kinds of chess, i.e. Fischer chess where the pieces are placed randomly, and will these games become more popular?
4) I've played chess off and on over the years, and when I was younger, I used to wish that there were people to play with, a problem that most people my age who didn't grow up in New York can attest to. How much better are humans going to get when they have a Deep Fritz to play every day?
Also, I think someone put it best before saying that the man vs. machine is not really man vs. machine, but man as performer vs. man as toolmaker. What's interesting in the whole man vs. machine debate is that it's not a debate between machine and men, but with men of different philosophies. How will this all affect chess? If humans are getting smarter, then it will only benefit chess as it is doing now. Where else can man and machine truly battle for supremacy? I can envision chess becoming much more popular with future eras of junior high school tournaments having thousands of kids playing thousands of machines, or perhaps teams will combine to play against the computer that the other team constructed. Perhaps grandmasters will let go of their invididual pride and realize that the only way to beat the next generation of computers will be to combine their intelligences. Maybe math teachers will realize how much computational power and pattern recognition chess has and they will begin to teach the game to their students. Point being, I think the popularity of chess will only grow as it seems to have done since 1997.
C |
|
|
|
|
|
|
|
|
some points (which changed into essay unfortunately).
|
|
|
|
1. might be "we" is the AI community.
2. But! Boby Fischer is known he said: "I liked when the opponent's ego had been broken". Who played the chess in the teenage knows what the feeling is about. Frankly I stopped to perfect chess when I realized there are tons of openings to memorize just to play at a mediate level. Some others were more gifted in chess playing and I withdrew into other activity area. Otherwise chess takes too much time and is not so useful for the life. I note that I was not pressed by parents or anybody else to make such decision. I simply moved into another areas where I could be more successfull. May be it is the basic human algorithm to find the best niche for each self. If I knew from the beginning that I cannot beat the best computer may be I would not start with chess at all (even if more talented). So this is the reason of ego and perhaps something other on the background too.
One of the human flaws is they need to be feeded by food and money which are passed to the wife and similar modern civilization inventions. How many top proffessional chess players is the humanity wishing to support in order to defeat the computer? Not many I guess. This is the reason of economy.
Despite all of the above I think that I still do appreciate the esthetic of the chess game (not of all plays of course). This may prevail forever and even the computer games produce the beauty sometimes, e.g. the 4th play Kasparov-DJ is extraordinal in some parts (mainly the Kasparov's defense after move 40). Poeple who play chess regularly also shall not stop doing so I suppose.
3. The difference of 100x times power increase used for full tree search represents just about 1,5 half-move ahead improvement in the middle game (I assume 20 possible moves in every half-move during the middle game). Current chess engines like DJ use some kind of selective strategy to search into about 15 half-moves I guess. So what, shall they search into 17 half-moves or just slightly more widely? Yes, they shall evaluate 100x times more positions and some statistics shall get better. But I am not sure how much better. Definitely not 100x times.
The next question regarding power is if general economy restrictions as well as physics restrictions shall allow to make further drastic power improvements. Moore law is just an empirical rule which is fine but no necessity. May be Yes and may be No. My teachers teached me in 1983 that in 1990 there shall be really intelligent computers of the 5th generation programmed in symbolic languages like Prolog. The implementation of Prolog is similar to the chess engine but these computers are still not on the horizon yet.
4. The more promising way is to improve the evaluation functions. If poeple find really good evaluation functions of the position (without the need to investigate next moves) then chess engines shall get essential advantage over poeple to choose the move according to the evaluation of all positions 8 half-moves ahead (accessible depth by brute force full tree search) while human player shall stay to consider rather the current position. This would be really dangerous for human players but I expect rather some mixture of static evaluation combined with the check of the next moves.
5. I also spent 3 years studying neural nets and I am not convinced it is appropriate to be used for chess easily. Chess is more like number theory, cryptography etc. - all these environments are 1 point left or right might evaluate into something completely different (like win xor lose ;-). May be that humans are the best implementation of NN and that enough large and complex NN can not be cooled in silicon and that is our problem. But who knows, NN and AI may be useful in some areas.
6. The number of atoms of the universe is rather irrelevant to the problem. I need not to have 1024 fingers to represent 1024 numbers! I am fine with ten fingers to perform that. But an efficient implementation to operate such numbers might require slightly more fingers.
7. Yes, definitely "we" are alive and our very basic instincts (some of which are likely processed by NN) are developed inherently to this fact. We have got some sense of it in different layers of feelings and ideas, from emotions to the principle of god... We are not the "layered computing system". I cannot prove it but I doubt anybody can prove the opposite and in that case I prefere may feelings. Yes, even feelings and emotion patterns may be psychologically imprinted but nobody can twist neither everything 100% nor forever. Even if somebody proved that (we are layered computing system) I would doubt there is some mistake in the evidence or some kind of hidden agenda. As we do not understand the "dynamic reference" we cannot imitate it with engineering means, the "heart" must be alive whatever the word inside quatas means. Ultimately the goal of living subjects seems to be to keep life or to grow and "the intelligence" of many types is just a mean for that purpose. May be chess shall also stay to be a good training and funny mean for the education of one kind of intelligence. Not to be used in chess eventually.
[ I apologize for grammar mistakes.] |
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
Although this article is interesting and informative, it does lack some chess knowledge, or assumes incorrect views on it, which is sadly typical for writing of this kind.
The fact that Deep Blue lost to Garry Kasparov in 1997 was hugely controversial. Most commentators believe that IBM did everything it legally, and possibly illegally, could to win this match. For instance, Kasparov was unable to put his legendary preparation to use due to having seen none of his 'opponent's' games, whereas IBM had access to tens of thousands of Kasparovs games, and a team specifically trained to focus on his weaknesses. IBM also decieved Kasparov over the nature of the match, which he assumed to be the nature of a 'scientific experiment' similar to that of his compressive win over Deep Blue in 1996. The match conditions where also not to Kasparov's liking, with IBM being pedantic over the arrangements of the games, even making Kasparov battle through spectators to even get to his rest room.
The most controversial aspect of Deep Blue's victory however was the possibility that it had human assistance during the games. Kasparov comprehensively defeated the machine in the first game, but when he astonshingly lost the second he knew that the machine had taken on a completely different 'nature'. Certainly IBM had changed the 'style' of the machine comprehensively from game one, but Kasparov also believed that his blue nemises had recieved help from an older one, Karpov, who was also in New York at the time.
In the present day human-computer relations in the chess world are better. The chessbase company, which produces all the strong chessplaying engines, generally maintains friendly relations with the 'Super GMs' by providing them with extensive databases, and even organising matches between GMs with the aid of computers to help them analyse, 'Advanced Chess'. Partly because of this partnership, both humans and machines have become stronger players since 1997. Grandmasters have made progress, with their phenomenal intelligence, in how to exploit their silicion counterparts inferior strategic knowledge, and how to produce unusual moves in the opening to take computers 'out of their opening book', or leaving without their knowledge of their pre-programmed database. Grandmasters, not even the highest ranked ones, have managed to inflict humiliating defeats using these strategies. In an extraordinary case, a lowly rated interantional player was able to write a book explaining a number of strategies which had helped him to beat Fritz consistently.
Unfortunately, computer intelligence is probably increasing at an even more rapid rate. The best computers now have more powerful weapons that incredible computing processes at their disposal; one of their most frightening features are 'endgame databases', which allows them to play numerous endings flawlessly, at a higher level than Kasparov. I'm sure most of the readers will be aware that the current 'Man vs Machine' match (Kasparov vs Deep Junior') currently stands at 2.5-2.5, with Kasparov brilliantly exploiting the computers lack of opening knowledge in game one, but falling prey to the computer's untiring stamina in game 3, with the computer playing increasingly well in games 4 and 5. It seems accurate to say, given that Kramnik was at one point two points ahead against Deep Fritz, that the top humans, playing at their best, are marginally better than machines but over a series of games, the computers stamina makes it approximately equivalent in strength.
Given that the computers are increasingly rapidly in strength, I agree that computers, (probably, there is still a hope for mankind :) ), will be stronger than the best humans by 2010 |
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
An interesting discussion, so I'll put in my bit by responding to some specific points
-The future of chess is speed chess
Unless I'm mistaken, computers have been able to beat even the highest-level humans at speed chess for some time, as at fast speeds major mistakes play a bigger role than more long-term complex positional concerns. Computer programs are more limited in search depth within a smaller time frame, but by the same argument that a 100-fold increase in speed does not provide a 100-fold increase in the depth the computer can search to, a 100-fold reduction does not reduce a computer's ability by such a huge amount.
-Opening book databases provide an unfair advantage to computers
I am no grandmaster so I can't pretend to speak from experience, but it is my impression that at the highest levels of chess human players have no trouble remembering openings. Perhaps even a grandmaster level player could do slightly better with a computerized opening book, but I highly doubt it would make a very significant difference, as these people have incredible memories for chess (just look at how many simultaneous games they are capable of playing in their heads).
-Computers never make mistakes
The programs, even at the highest level, are only as perfect as their design is perfect. The best computer chess players are extremely complicated creations and are constantly being extended, and even in major games (the Deep Blue match for example) subtle program flaws are sometimes responsible for "mistakes" on the computers part.
-Deep Blue was controlled in part by human grandmasters
While it is almost impossible to prove anything historic to everyone's satisfaction, that Deep Blue's moves were the result of human intervention during the game seems extremely unlikely. Logs exist of Deep Blue's activity during the game in question, and there appear to be adequate positional motivations for the behavior deemed to be too human.
Having said all that I'd like to point out that I agree with those who have been saying that chess will not die out just because a computer is better at it than most or all humans. It may suffer somewhat in popularity, but I can't even say that for sure. Chess may lose (or perhaps has lost) some of its mystery as a result of very good chess programs, but to the casual player that will make very little difference. Anyway, I've said enough for now.
-E
An interesting discussion, so I'll put in my bit by responding to some specific points
-The future of chess is speed chess
Unless I'm mistaken, computers have been able to beat even the highest-level humans at speed chess for some time, as at fast speeds major mistakes play a bigger role than more long-term complex positional concerns. Computer programs are more limited in search depth within a smaller time frame, but by the same argument that a 100-fold increase in speed does not provide a 100-fold increase in the depth the computer can search to, a 100-fold reduction does not reduce a computer's ability by such a huge amount.
-Opening book databases provide an unfair advantage to computers
I am no grandmaster so I can't pretend to speak from experience, but it is my impression that at the highest levels of chess human players have no trouble remembering openings. Perhaps even a grandmaster level player could do slightly better with a computerized opening book, but I highly doubt it would make a very significant difference, as these people have incredible memories for chess (just look at how many simultaneous games they are capable of playing in their heads).
-Computers never make mistakes
The programs, even at the highest level, are only as perfect as their design is perfect. The best computer chess players are extremely complicated creations and are constantly being extended, and even in major games (the Deep Blue match for example) subtle program flaws are sometimes responsible for "mistakes" on the computers part.
-Deep Blue was controlled in part by human grandmasters
While it is almost impossible to prove anything historic to everyone's satisfaction, that Deep Blue's moves were the result of human intervention during the game seems extremely unlikely. Logs exist of Deep Blue's activity during the game in question, and there appear to be adequate positional motivations for the behavior deemed to be too human.
Having said all that I'd like to point out that I agree with those who have been saying that chess will not die out just because a computer is better at it than most or all humans. It may suffer somewhat in popularity, but I can't even say that for sure. Chess may lose (or perhaps has lost) some of its mystery as a result of very good chess programs, but to the casual player that will make very little difference. Anyway, I've said enough for now.
-E |
|
|
|
|
|
|
|
|
Re: Deep Fritz Draws: Are Humans Getting Smarter, or Are Computers Getting Stupider?
|
|
|
|
It is an interesting article discussing computer power versus human abilities, like playing chess. However, I would like to state some other views.
First, I would agree to what it states that the computing power is still increasing and computation pattern is keeping improving. In the daily life, my new Quad-core processor costing about $300 which is at least 4 or 5 times faster than my 4-year-old Pentium processor (about $330 in 2004), especially when it loads the operating system, such as windows XP. From the industrial side, the scale of Intel processor grew from 103 to about 108.5 in 40 years (www.intel.com), we can also easily conclude these figures exhibit MoorĄŻs law (it predicts computation power doubling roughly every 2 years). As stated in the article, the more advanced computation pattern does make difference, because we can use less powerful processor to achieve the as same amount of calculation as faster processors. Therefore, I agree that the computation power is rapidly increasing in exponential level, and software improvement plays important roles in computation speed .
However, will the computer hardware continue its speed upgrading infinitely? My argument is probably not. Although Ray Kurzweil predicts new technology like bio-computation, quantum-computation or whatever computation technology will come to replace the current transistor chip, there are some physical limits for the computation. Suppose, we have a processor which is created by some new technology, and is horribly fast (it is fast, but it has a limit). Then, some one could build a very large computation matrix by using this new processor. The computation matrix cannot be infinitely large, and letĄŻs say it is as large as a large building with dimension of hundreds meters * hundreds meters * hundreds meters. Next, assume some complex computation requires the user to input data from one side of the building, and the output will come out one the other side of this building. Now, how fast can user get the result? By simplified calculation, it is equal to (distance of the two sides) / (speed of light). This is because any proceeding of information processing is based on transmission, and how fast can it transmit the data? It can only be as fast as the light in the straight, which we ignore the actual data calculating time. Unfortunately, we cannot create matter running faster than the light. Otherwise, we will get back the past by the theory of relativity.
Next, Ray Kurzweil argues that human performance on calculation, like playing chess, stays constant, but I would not quite agree with this. The argument should be the calculation performance of chess world champion stays constant. The tricky thing is how Ray Kurzweil knows the best performance of playing chess is from chess world champion. We cannot try or test everyone in this world, and find out most smart person playing chess in this world. Or do we actually encounter computational problem here? Furthermore, like Ray Kurzweil stated, they had several generations of the special hardware to perform the chess tournament, and it will defeat human for sure due to the increasing of computation power. However, can we have several special generations of people specializing playing chess? Probably, we cannot ask the most genius female chess player marry to the most genius male chess player to produce, possibly, next generation of more genius chess player, but if we do modify the gene structure to get one, chess monster by the new or future bio-technology (surely we have to ignore or totally violate the humanities and ethics).
Last, I would have different views that Ray Kurzweil predicts human will loss interests on play chess due to defeated by computers. Human should be proud of building the machine which could perform the same tasks as human. Beside that, the chess is like a competition game, and I guess if only the computer cheats or human player using cheating software every time in the chess game, the human will loss interest with playing it. A simple comparison is transportations. Now, we have cars, boats, and planes, and all of these are faster and could load more stuff than human, so do we lost interest with walking, running, swimming and traveling. The answer is no, and actually human benefit more from building thing faster than themselves. In the future, the best of chess play is make human against human, and computer against computer, like 100 meter running and car racing. Therefore, I would predict human will benefit by the faster and smarter computers and computer software in chess.
In short, I agree with the exponential increasing speed of computation power and the gain of software in this article, but I would have different views on human computation performance and interest in chess in the future. |
|
|
|
|
|
|
|