Magnus Carlsen once said that he hates to play against the computer because he loses most of the time and instead use it for preparation only. So why is it that even the best chess player in the world can't win against a computer software even with such deep tactical knowledge and understanding of position? In computer chess tournaments, computers do make mistakes against other computers and lose but why don't they do so against humans? Also, why do computers like to make moves that sometimes make no sense to us, by moving pieces back and forth or retreating a piece to an awkward square or creating weird double pawns?
Men go with a torch through the dark jungle. Not bad, but the computers see everything. Men can't handle a exponential problem, it's simply too much.
Short answer: because computers see further than humans.
For the rest:
* As long as computers have a finite search depth, they will be capable of 'mistakes'. The less good computers see less far and make more mistakes, but still less than humans.
* Computers aren't programmed to have human intuition, they use brute force. However, weird-looking moves can be mathematically superior.
Which is the best processor today?
A few numbers...
Top Engine ratings.
1 SugaR XPrO 1.2 64-bit 4CPU 3412
2 Houdini 6 64-bit 4CPU 3400
3 Komodo 11.2 64-bit 4CPU 3397
4 Deep Shredder 13 64-bit CPU 3316
5 Fire 5 64-bit 4CPU 3274
Top Humans ratings
1 Carlsen, Magnus g NOR 2826
2 Aronian, Levon g ARM 2801
3 Vachier-Lagrave, Maxime g FRA 2794
4 Caruana, Fabiano g USA 2794
5 Kramnik, Vladimir g RUS 2794
As simple as that...
No, it's not that simple.
For the top human players, the majority of their games are in invitational, round robin events. Once they have reached the elite levels, their ratings remain as such, as they only play each other.
Same is true of the computer ratings. Most all games are played vs. other programs. They do not play humans. Once a prog plays an established 3200+ prog an draws, it joins the club. Remember, ratings are based on relatively fast time controls where programs have the distinct advantage.
Is a 3400 program 700 points better than a 2700 human?
Absolutely not, even in fast time controls of 60 minutes. Make the game 2 1/2 hours like the old days, limit the comps thinking time to on the move and you'll find the discrepancy is far less.
Well, if some Super-GMs like Nakamura lose with extra-moves (tempi) and material right out of the opening in a show computer match: any doubts what 3300 means?
The super-GMs know all that they were food for the guns. So what.
Three reasons.
1. They never, ever miss a tactical shot, of any kind, not in the opening, not in the middle game and not even in the end game.
2. They see much further out on the horizon than humans, humans play more by what feels right.
3. They never grow tired, or distracted.
Yes, but those were rapid games. 20 minutes. Of course Naka was helpless. Point is this 700 point gap we see is not indicative except in relative fast time controls. There is no doubt prog's will win at any time control. But they are not "700 points stronger" in a traditional time control.
Compare it to a 2700 human vs a 2000. Extremely unlikely the 2000 will ever achieve a draw. Are all prog's equally that much better in a 2 hour game vs. humans? Will not the 2700 human achieve the occasional draw vs many of these top rated programs?
As the better programs keep winning, their rating keeps going up. They only play each other in fast time controls. The current fashion is 60 minutes.. Soon, heck, we'll see prog's rated 3500/3600 as more enter the field. Are they getting that much better than humans, or is it simply rating inflation?
It is debatable if humans score 1 draw out of 10 or 100 games. This means: the battle was lost long ago.
This topic has been archived and can no longer be replied to.