Okay, I retract that statement and put forth the weaker claim that humans have steadily increased in objective strength over time, and that this trend is indisputable. Also, there's no reason to think that it won't continue.

This is not even a bet, there is absolutely zero chance that Rybka will still be stronger than Carlsen in like 5 years.

Do you see what I did there?

on behalf of the general public, i'll say that this is true.

I think that once the top twenty players in the world are all approximately equal in strength, it will be possible to say that humans have reached their peak, or at least a peak. This hasn't happened yet: the top player still has a significant edge over the #20, and I figure humans still have a couple of hundred points of objective chessplaying strength to gain.

However I don't think computers are about to stagnate, either. Computers have continued to improve in strength even after coming to the level of the top grandmasters. Even if you think computers' evaluation functions can't be improved much further (and I don't see why) there can be gains in strength from other sources. Give Rybka 10 times as much processing power and eight-piece tablebases, and it would crush the current program. Computers will get stronger even faster than top humans will.

Give Rybka 10 times as much processing power and don't change its evaluation function and it will do slightly better.

Seriously, have none of you heard of the horizon effect? Diminishing returns? Hello, am I getting through here?

I doubt there will ever be another match between the top human and the top computer, but if there is I would happily bet on an X-0 with 0 draws result in favor of the computer.

In fact if there was some way to bet on it, I would bet that the current strongest computer is better than all future humans at chess until the species becomes extinct.

Kramnik - Deep Fritz was 7 years ago and computer chess has come a hell of a long way since then.

You'd lose that bet, Pyromantha, so long as the match was of reasonable length. The last major match between a top human and a computer ended in a draw. Computers have gotten stronger since then, but not by that much.

yes I know about these things. Strictly speaking the horizon effect isn't the issue here, the issue is that it takes exponentially longer to search deeper into a position.

I still know that Rybka for now plays significantly stronger with 30 minutes per move instead of 3, and not that much better with 300 minutes instead of 30. But this is at least partly because the programmers optimized it to do best running on a normal computer at a normal time control. If they knew it would run on a computer 10 times as fast, it would do even better.

Do you think the human brain is the ideal computational architecture for chess playing? Or do you think something beyond computation is involved here? (and if so what is it: "creative spark", "the soul", or what?). I'm really struggling to understand why you think humans will reverse the trend and overtake computer programs in chess playing strength.

Computers will continue to get better at chess, even if the only reason is faster processing speed. Even if these improvement are small, the human brain's computational power isn't improving at all. Aside from the possibilities opened up by genetic engineering, the human brain is a fixed architecture. Additionally, there's nothing to stop chess programmers from devising better algorithms. Chess engines have more room for improvement than human chess players, they aren't constrained to a fixed architecture, why is any of this controversial?

wat

It still wouldn't scale.

If Rybka and Carlsen played a match today with a time control of 60 minutes per move, five moves per day, Carlsen would win convincingly.

Oh, there's no doubt in my mind that in the long run computers will be better than humans in chess. But then again, in the long run the human/computer distinction is going to become blurred, so that's a moot point. I'm saying that in the near future humans are going to re-take the lead. That doesn't mean that they'll hold it forever.

It's not, but I'm not saying that humans are going to dominate in the long run. In fact, I never said that.

Get a baseball machine to throw a 200mph baseball. Nobody's going to hit that ball. If they do, it's likely to result in a busted bat. Would you then say that machine's the best pitcher of all time? No, it's just a machine that is absolutely terrible at all aspects of baseball but has an inhuman ability to pitch at 200mph making it impossible for humans to compete with it. Who cares?

Computers have many inhuman abilities that allow them to compete at an extremely high level of chess. Their actual chess understanding is nonexistent. I call that terrible.

Dire, no offense, but you strike me as a bit of a chess elitist. You look down on blitz without increment, and now you've redefined the word "good" to be able to claim that engines are not good at chess. I'm sorry, but engines are very good at chess. Get over it.

As for your baseball analogy, it's not a good one. You're assuming the consequent by conflating "pitcher" with "baseball player". It's not as if engines play 90% of positions very well but play 10% of positions terribly (which would make your analogy work, since a pitcher has to do non-pitching things while on the field). They play 99.99% of positions very well and .01% of positions terribly.

Oh, and the pitching machine would be the best pitcher ever, if you called it a pitcher. So you need to either say that engines aren't really playing chess or admit that they're good.

horizon effect = sometimes a computer will miss something fairly straightforward in a critical line, because it didn't search far enough.

exponential growth of a game tree = if you give a computer ~10 times as much time, it can still only search one move deeper, or 100 times as much time, then it still only sees 2 moves deeper, etc.

obviously these two things are related, but they are totally different! The first one is something that happens to all computers to some extent, it might be after 3 moves or after 23 moves. The second one is the main reason why faster computers don't lead to a proportional difference in playing skill.

wait, you think computers will continue to get better and better, but that Carlsen specifically is improving at some kind of superhuman speed? I don't think it's possible to judge that.

If your point is that Carlsen and the new generation of GMs will be better than the previous generations, partly as a result of being able to train with chess engines, databases, etc, I agree with that. So there could be a jump in overall skill level with that generation. I just don't think it will keep pace with improvements in chess engines over the same time scale. Anyhow we agree on the long-term situation, your earlier posts gave me a different impression, sorry if I jumped to conclusions.

The 200 mph pitching machine would be the greatest pitcher of all time. I'm not making an aesthetic judgment, there's nothing pretty about it, but the batters wouldn't be able to deal with that kind of heat. I'd put it on the mound for my team every game if the rules of baseball allowed it.

I think we have very different concepts of what it means to be "good at chess". For me, "good at chess" has to do with the ability to win chess games, never mind how the results are achieved, never mind whether the games are aesthetically pleasing. It's purely about proficiency in achieving the desired result of checkmating the opponent's king.

...because of the horizon effect. Anyway, moving on...

No...

Okay, look at it this way. Say you have two containers, one of which holds 10 liters and the other which holds 1000 liters. Right now the 1000 liter container has 7 liters of water in it and the 10 liter container has 6 liters of water in it. The larger container is being steadily filled with water at a rate of .1 liters per day. The smaller container is being filled incrementally with 2 liter dumps every week. One week from now, the larger container will contain 7.7 liters of water and the smaller container will contain 8 liters of water. Eventually, though, the larger container will contain 1000 liters and the smaller container will only contain 10 liters.

The point is that people have given up on humans before they had a chance to come out with the next version (generation) of the GM. If the last generation of GMs, who didn't grow up with the internet, chessbase, and Fritz, can score ~30% or so against the engines, the next generation will surely do much better, because humans improve drastically from generation to generation but not much within a single generation.

Seems to me we have a lot of theory here but the only way to prove either side's theories will be to wait and see. :-)

Can the effect of having played lots and lots of games on the Internet, and having studied with good computer engines, overcome the steady growth of computing power? We shall see.

Well, we may never see if the strongest humans never play the strongest engines again.

But if you're right and Carlsen or some other GM improves to the point where he can beat the top chess engines, he'd be pretty likely to seek out a match, wouldn't he? I mean that would be a pretty significant development, and people who follow chess would be very interested in seeing how it played out. So if you're right, we actually will see.

But if you're wrong, we won't see that match. There would be no reason for it to take place.

This discussion is so pointless.. Sorry to everyone that is defending the human part, but I doubt there will ever be a match between a top gm and top engine, that a GM will win. MAYBE insanely lucky draw. I think Carlsen is still way way way below Kasparov's level - Kasparov couldn't beat an engine like 10 years ago, which was his prime! And what do you know, maybe Carlsen's peak is very soon. Or maybe it's very far - I just know, that engine's peak is FAR further than any living chess player's.

HOWEVER, that doesn't mean anything to me as a chess player. I don't look to chess engines as to opponents (they are not, they were created to HELP us), and I can easily accept the fact that as matter of strength computers are stronger than humans. However, there are certain aspects of the game that computers will not be able to adopt, and I am pretty sure that chess will not die anytime soon.

Discipline, you have no idea what you are talking about, humans have no chance, and the top computers ARE MUCH better than they were when Kasparov drew Deep Blue, by at least 100 ELO points, and possibly a bit more.

The top computers now are probably about 200 elo points stronger than any human who has ever existed, and they are still getting better.

I agree. I was deeply involved in chess 25 years ago, did a lot of TDing, and was pretty well-versed in the chess world at that time. I dropped out of the scene totally, so I would definitely be a member of the general public now. Even given that I have much more of a chess background than most of the general public, what Dire said above is pretty much the extent of my knowledge. Until a couple days ago when this forum opened.

I'm not sold yet that computers are unbeatable in matches. If Anand would have spent the last 2 years to prepare for a match against a computer, he would have an edge today imo. The last human vs computer matches were played by GMs who are used to play against humans. They were outplaying the program most of the time, but making inexplicable weak play in some spots, e. g. Kasparov in the 6th game or Kramnik (mate in 1, wtf?). The new generation is better used to fritz and rybka and might avoid hitting a psychological barrier.

This thread is convincing me that there could still be interest in another man/machine match. Why I don't know, the matches now would be very one-sided in the machine's favor. Lots of chess players remain unconvinced however. How much would the next world champion need to be paid to submit to the inevitable beating?