Chess engines VS chess masters

Hello everybody!

I want to launch a debate about the eternal struggle between (chess) machines and mankind: Do you think that, one day, chess engines will be ALWAYS capable of beating grandmasters? Ok, chess engines can't have the famous "feeling" of the position and can be beaten by a strong positional player, but we all know that technology is getting smarter every day.

So what do you think?

            

At this moment, the top chess engines are pretty much unbeatable by the top human players. I believe this gap will always exist and may even get larger, especially given the current rate of technology innovation. But who knows that some genius such as Morphy and Capablanca may be reborn and revamp the whole theory of chess that allows humans to play better by leaps and bounds.

About 5-10 years too late for this debate.

Interesting hankas: yes, may be many things have still to be discovered in chess. A great player like Capablanca became bored of chess because according to him too many games ended with a draw. He was thinking that everything is already said about chess, and because of that tried to innovate the game itself by proposing his famous chess variant:

                   

But later, many new things were discovered! Who knows, may be many other things will be discovered

Yep.

We can expect both the hardware to keep improving (hence, the engines will run faster and search deeper) and the algorithms and evaluation metrics to be better in the future.

Also, there is no ultimate obstacle preventing the computers from 'thinking' the same way GMs do. It's just that it's very difficult to achieve and wasn't necessary (so, the current engines are programmed in a different way) - but the field of AI hasn't really given up on the idea.

In any case, engines are already better than top GMs (they may not have deep insight, but they do not blunder) - and they will be even better in the future.

On the other hand, we as humans have our biological limits... so, the conclusion is quite clear.

the_cheradenine It would be terrible if chess engines become unbeatable

Not really :)

I mean, I am a computer scientist, working in the field of machine learning :D So, I would actually like it if the computers were to become 'smarter' than they are now ;)

Besides, the whole point of chess is a mental struggle between two people. Regardless of how well the computers play, you would always be able to play your friend and enjoy chess the same way you do now.

---

Ok I got it

"Besides, the whole point of chess is a mental struggle between two people. Regardless of how well the computers play, you would always be able to play your friend and enjoy chess the same way you do now.":

Really a good point!

---

Roll on the Kurzweilian Singularity! 

The eternal struggle is over and the machines won. The top engines are far better than the top players today and the gap is expected to widen with time. Still, chess programmers have pretty well given up on the idea of computers thinking about the positions in the same way people do. But interestingly, as computers have become better, top players have started looking at the board more like computers do, or at least understanding what computers look for.

The real AI discussion now is around games like Go, where computers simply aren't making the kinds of advances that they have in chess. Partially because the problem space is much larger, but also partially because the game isn't nearly as predictable as chess games are. 

I agree with you Kingpatzer and I would like to speak about another thing that seems important to me: the LINES.

I mean a chess engine with a huge database (let's say all the chess games played from Philidor days to our days) "knows" what line is statistically better for White or for Black (according to the percentage of wins/losses), in the opening, middle game and (of course) end game stage. So is the chess engines superiority over humans simply a matter of statistics and not of real "thinking"?

Well yeah, currently they do not think in any meaningful way...

They just pick the line with the highest evaluation.. but sometimes a slightly inferior line (both theoretically drawn, let's say) - offer better practical chances if the opponent is human (if it is very complicated/difficult to play)... so, the overall positional evaluation is often meaningless.

And yes - they have large opening databases, endgame tablebases, etc. ... but they compensate for their stupidity by brute force - searching through many millions of moves/positions, which is something that we obviously cannot do. :/

Thank you for the explanations the_cheradenine, very instructive!

Also, chess engines rarely resign or offer a draw I think, "battling" to the very end.

---

 So true!

I think what is more interesting is the effect of computers on the way humans play chess. If we examine the games of older chess masters (those who played before the computer chess era), we can see the clarity of their plans/intentions. Now if we look at the games of modern GMs, it is often difficult to see the "overall" plans. It is as if the modern GMs play in a haphazard manner. Some people say this is because modern GMs employ more sophisticated strategies than their predecessors, but I prefer to think that modern GMs rely less on planning and more on pattern recognition. In other words, human players are playing more and more like computers. Is this a good thing? I don't know.

I can't agree more hankas but I think that if the new GMs style is opaque it's because it often "condenses" a lot of old/new theories.

i think that it's simply a matter of those old 'clear' moves turned out to be easily countered once the basic patterns became known. They're still playing for the same strategic goals, but the reality is that getting to those goals requires more complex manouvers and they're more aware of those complexities because of their work with computers. 

At the turn of the 20th century, human raced against gas-engine carriages, until they lost that race.  Ditto with Chess Engines today.  That race was surely over, 5 or 10 years ago.  Sadly the humans lost, yet again.   

But what about the idea that the computers are violating the "touch move" rule?

Aren't the engines (in some cyber-silly philosophical world) really moving all the pieces all over the board, before they choose their next move?

Feel free to refute my conjectural foolishness.

---

Virtually speaking, yes they are

the other way to look at it is to say that computers play blindfold chess because they simply keep the board position in memory and don't use any visual input at all.