Humans Vs. Computers in Chess

Nowadays the best computers are considered impossibile to beat by the best humans over the course of several games.  This was "proven" in a recent human vs computer match where the best computer beat the current world champion.  Keep in mind now, the human was allowed to look at the computers thoughts while the computer was in it's opening book and he still lost.  Personally I would never play a computer this way.  I think we as humans have given up on ourselves.  After all humans make computers, so we should be able to make ourselves better then them.  The fact that we can not seem to do this means that we are not trying hard enough.  That's my opinion anyway.

All this angst about computers beating humans at chess seems overblown and unnecessary. After all, should humans stop competing at track and field races because a car will beat the best human hands down every time? Let humans compete against humans and leave the computers to each other.

Going slightly OT, does anyone have any idea when a computer will be able to 'solve' chess? That is, know the winning move\combination in every situation without having to do any actual anaysis? This was recently done for checkers so I figure chess must be next.

 Never make mistakes?  You are wrong.  Computers make lots of errors.  It just takes a 2900 player to punish those mistakes.

My 2¢ ...

Computers are far more inefficient than humans in calculating since they must calculate every move, every line, whereas humans typically narrow their search immediately.  The fact that processors have gotten so fast has made that inefficiency negligible and ever more effective in it's inefficiency.  Increased, and stored, opening theory has also aided computers to become more efficient. But, one odd advantage to a program's ineffecient methodology is finding resources by examining lines a human possibly never would, resulting in rather perverted, yet sound  (and, yes, effective), defenses.

When Botvinnik was preparing for his rematch with Euwe, he mentioned the belief that he needed most of all to work on his 2 move combinations.  A computer's two-move combination sight is near perfect. Where a computer traditionally falls short is in positional considerations and evaluation of resulting lines, generally measured materialistically.  In this regard computers have come a long way in the last couple years.

My opinion is that computers and top GMs are roughly equal at this point, but the human factor - thoses weaknesses inherent in being human - is the most telling factor and so computers win often not because they have played better chess but simply because they are machines.

In his recent bout with rybka (described here) Jaan Ehlvest, lost easily. In reading up on the encounter, I had the feeling that Ehlvest, a wonderful player, lost before he started. He had lost before at odds and knew the power of the program (which he seems to feel is unbeatable in a match) yet his pre-game preparations seemed pretty thin. If he had been planning a match with another GM against whom he felt a decent chance of winning, I suspect his prep would have been ten-fold. So, to play a decidedly stronger opponent with less prep is self-defeating in itself.

I believe the juries still out on this human-computer affair, but one thing I am convinced of is that humans contribute as much to their own defeat as the machines do.

In the beginning, it was the brute computing strength of machines vs humans' knowledge of the game and our ability to recognize position types. Then we improved chess engines by adding our knowledge of the game and pattern recognition. Meanwhile, we ourselves didn't really improve. So you can see why chess computers are beating us. 

MattHelfst; computers programs DO make mistakes.  Intel has PUBLISHED the fact there chips make mistakes. This is available information to anyone, you should really read magazines and do research before you claim something is impossible. Humans make computers therefore computers make misktakes.  We also know from research that humans only use 10% of their brain power, so we CAN do a lot better.

"We also know from research that humans only use 10% of their brain power.."

hmm... I understand that to be an urban myth of sorts. who knows?

If a computer can always think a set number of moves ahead, the question lies not in the ability to out compute the computer (which looks at possibilities in an altogether differnt fashion) but to utilise generalisation and intuition. This is the human minds advantage against computers. Some know of this as strategy as opposed to tactics. This manifests itself in "opening theory". This is what computers fail at; if someone were able to tap into the full potential of generalisation or 'strategy' effectiveness against computers would no doubt critically increase.

Generalisation is the human trick for dealing with large numbers of possibilities. Now it follows logically that if certain principles could be worked out (and some have already; manifest as common opening theory) and calculation (as that of computers) applied to them instead of purely to tactics this would surely be superior. The question is to discover these generalities ("laws of strategy" if you will) and to combine them and manipulate them consciously.

If one looks at a computer program's "depth" this is the way it evaluates chains of positions to find the best ones. It would thus seem that Generalisation is the next level above "depth". Now what is the next level above Generalisation? Perhaps this is the central question, the philosopher's stone of chess; or alternately is it so hopelessly abstract and confounded that it is not a practical question, at least at this point akin to asking "If Good versus Evil are the two fundemental, equal and opposite moralities, than what would it be like if there were three fundemental, equal, and opposite moralities?" (Actually that's a favourite question of mine. And it's MY intellectual property.)

Computers have great difficulty thinking abstractly. Abstraction and hence Generality are major failing points of modern computing techniques. (Note that I say modern. If these "laws of strategy" can be put into a form utilisable even minimally by computers than there is nothing to stop the doing of it so... and they will naturally stretch those by computing as they did with basic tactical principals). Once again the key seems to be not only Generality, but exploiting it effectively. This is also in part the genius of Kasparov and others (a favourite player of mine I confess). Unfortunately, though I know what I'm looking for, it would seem I'm not quit adept at finding it - I'm a mediocre player (improving mind you) but I'm still positively awful against computers!

Now, among the shortcomings of humans is that of Truncation. Humans truncate perhaps because calculation of every distict brach, exponetially increasing in complexity with each move looked forward by is simply 'too much' for anyone at a certain point. So, they see a loss and frequently truncate the chain (computers truncate too, but only at checkmate :p) to make it easier. This is a blessing and a bane. On one of the hands, it can potentially allow one to calculate farther ahead (effectively) with the same amount of effort, though on the other hand it can stop the calculation of useful trains of thought. Admittedly humans cannot handle all of the trains that a computer can and hence this is not a valid solution one extreme or the other. Perhaps truncation [i]is[/i] the key; that is knowing exactly what to truncate. Perhaps that [knowing what and when to truncate what] is the true nature and meaning of 'Generalisation'.

Or perhaps, in a similar train of thought, Generalisation can be used as a clue and not as a solitary principle, to enhance tactical thinking enough to trump a computer. In essence providing a direction guiding and aiming the tactical thinking to beyond the limit of a computer, or providing insight when a computer's approach yields 'just too many possibilities' that are equal and difficult to choose.

Or the other posibility is that this is all a load of bullshit. Lord knows I consider that possible :)

To solve checkers (draughts) it took super-computers a long time to calculate the winning method. Chess is inifitely (almost) more compicated. There are more possible chess games than atoms in the universe.

I think the original poster was talking about the best human chessplayer(s), not just any human in general.

 "Humans make computers; therefore, computers make mistakes." Not a valid conclusion. That's like saying: Humans make calculators; therefore, calculators make mistakes. But we all know that isn't true.

I think we need to define 'mistakes'. Computers certainly do NOT make miscalculations during a game, thus no 'mistakes'.

I'm curious, who considers computers impossible to beat? The world's top grandmasters consider computers to be very, very difficult opponents, but I don't think it's a foregone conclusion that they are "impossible" to beat. If you're referring to Kramnik's match last year against Fritz, Kramnik overlooked a Mate-in-1 threat in Game 2 and lost. So he basically gave the full point away. In fact, Kramnik was actually ahead in that game until move 33, so it's not like Fritz had completely outplayed him.

Fritz did play a great Game 6 though.

 That's funny!!