Will computers ever solve chess?

Engines have beat GMs without opening book, and have beat them when GMs have access to an opening book... not that it shouldn't have been obvious!

  Exactly!! That sums up my virtual response to aln67. What seems stupid may actually be a genial tactical stroke.

   It doesn't look like an improvement in calculation can affect such an astronomical number of possibilities.

  As for a framework, again, it would be a partial framework, as in a possible false framework( not being based on the whole picture, but on a partial one).

That is true, but I still think you're missing my point, which I'll try to answer by... asking a question :

How do you take care of all possible checks that your opponent can give in a given position ? Do you consider each of his pieces, to see whether it might come to a square where a check would be given ? Or do you consider the open lines starting from your king and consider only theses pieces that could act along these lines ? Don't you think that the latter method is more performant ? I assume, of course, that you agree that looking for checks has the highest priority...

Let's imagine 1 e4 d5 ; the only possible check can be given by a light square bishop (which, in that case, *can* be moved). Do you think it's efficient to consider any other move by any other piece or pawn ? 

Some other geometrical patterns might be treated the same way, I think.
I don't mean no other move might lead to sequences being good after several other moves, but I just say that these obvious patterns should be analysed in the first place.

  The most idiotic-looking moves may be the best moves in certain positions that we know, and in any position we don't know, since the end hasn't been reached.

  An elimination doesn't lead to truly 'solving the game', as attempted by computers.

  As for your example, I don't think a computer considering a check goes through all the pieces' possibilities. Due the exact nature of the pieces, open files and dialgonals, a computer knows immediately how many checks are available in a position, instantly. After all, against Deep Fritz, Kramnik, after playing 34...Qe3??, was instantly mated by the machine in one: 35Qh7#.

  Of course, that was a special kind of check, but still, the machine is not interested in just checks, it's interested in all the moves, if allowed, and it should be, for nobody can say the magic move doesn't look like a blunder.

Yes, he did maybe get tired, something a machine can never do: either allow mate in one or get tired, frustrated, nervous, all that.

  Defense can also be offense and vice-versa. One can defend very well, and the opposite side has no moves, so the initiative is passed over, or worse, a blunder is imminent. All that is relative, though, unless the whole thing is 'seen'.

People seem to think that a computer "knows immediately how many checks are available in a position", and so on. A computer knows nothing, and doesn't even know it's playing chess. ;-)

It is running a software that has been thought, line after line, by a human being, and all this discussion is about how this soft has been thought. 

  Of course, the computer doesn't know it's playing chess, because it doesn't know it exists, since it really doesn't.

  But the point was that the machine is not just looking for checks, so even if it does it the way you described it, it still has to check all the legal moves, for nobody knows where the gem will come from and, unlike humans, they don't have beliefs about how the good moves should look like.

  So no matter what, checking all the legal moves is necessary in order to 'solve chess'.

  Of course, contrary to Steve Jobbs' beliefs, a computer is not an extension of one's body, nothing is: a computer is a computer and our bodies are our bodies.

  That's precisesly the problem in today's society( among others): we depend too heavily on computers, and when they fail humans can be clueless, since our brains are used less and less..

  We seem to be heading in that direction. Needles to say, we needn't do that. Computers can, in fact, be used as gadgets to ease our lives, so that the mind wouldn't worry about technical things, and instead be free to solve other, much more important, problems, which computers can never ever solve, such as our insecurities, hatred, aggression and so on.

  That is questionable, because what seems silly may turn out to be brilliant in the big picture. When that is seen, then yes, many can be discarded, so what was a mate in 10.000 moves can become mate in 273, with perfect moves for both sides--because any other permutation will result in a mate of 274 moves or more...

I also think there will be strong enough computer to be able to analyze above problem, but think that "solution" will be that perfect play from both sides returns infinite draw no matter white or black pieces. 

"So no matter what, checking all the legal moves is necessary in order to 'solve chess'."

It's probably too early to be so ambitious ; if you watch some tournaments live on the net, you'll notice that the soft sometimes changes its evaluation without any obvious reason. It might, for instance give, let's say, 0.7 for a sequence beginning by some opponent move, and then when the opponent has played *this* move, the evaluation is different !

My example talking of checks was a very special case, I admit. But in this case, at least, considering any move by any piece *is* stupid, because the check must be addressed for the (first and only) move be legal !

  The check is one among many other things to be addressed. Even if you come up for a way to shorten that process, still all the other moves have to be addressed. We are not talking about current engines, whose evaluations change because once a move has been played, their horizon became whatever it was, plus 1, so the evaluation may be different.

  In the ideal case, the engine must have a perfect horizon, one legal moves will exhaust all the possible branches, something which the current engines are not even close to, in the middlegame, let alone the opening.

  I don't think they do. Computers have a 3.000 plus Elo these days, while humans never reached 2.900, blitz section excluded.

  That's the current belief. It's quite reasonable, but it's still a belief at this point.

  Now what?

What kind of soda?

It took them quite a long time, and it took them until just a few months ago to beat a human at Go. The point he's making is that while computers can calculate stuff exponentially faster than humans, humans have an ability to rule out trivially bad lines without having to explore them to confirm they're bad. Go presents far more trivially bad positions and lines to calculate on average than Chess does, so it presents a much bigger CPU-ache for engines. Even the one that beat a human made a few astounding blunders along the way.