Will computers ever solve chess?

That's an interesting question. I sure hope it would solve it before the universe ends (either due to natural causes or because your god thinks it's time).

Let me put it this way: if I build a tower, first story is 10/1 meters high, the second is 10/2 the next 10/3 etc. etc. In some time I'll build an infinitely high tower, but it'll be of very little use and I don't think anybody would be impressed by a tower of 90 meters high.

Ah, to be youthful and to let you imagination run roughshod over your better judgement.

What a thrill.  So engaging.  Anything you can imagine MUST be possible, eventually.

Neat trick to your thinking.

Once you look at it, a bit more closely.  All those infinite monkeys typing away on those infinite number of typewriters will eventually recreate all of Shakespeare's works.

What have you been smoking, kind sir?

Recurrent neural networks can do a passable attempt at writing in the style of Shakespeare, being provided with nothing but the ASCII text of his complete works.

I think I see your point: it is that in a given position, a lot of black replies will permit a forced mate. Each bad move the opponent makes either permits a mate or makes a mate quicker, and each such move reduces the computation needed on a branch.

The problem is that even if typically a player has a small number (2 or 3, say) of moves that do not harm the theoretical result, this still means there are a heck of a lot of branches in a tree 100 moves deep. My suspicion is that with play of this quality, 200 move games might be the norm.

An extension to your idea is that white can play simplifying moves to approach a forced draw more quickly. I am not sure how much this helps the computation, but it is of some use.

 I was thinking along similar lines to you, but I think the relevance of vickalan's  comment is that presuming perfect play leads to a draw, for a large fraction of black's responses, forced mates will be available to white.  If inferior moves follow inferior moves, the forced mates get shorter.

Even moves that say lose 50 centipawns in a computer assessment mean that a following move which is similarly inaccurate (in an imprecise sense) may well leave the game lost (i.e. a forced mate). This does reduce computation, but IMO not enough to make a solution anything like feasible.

The idea that White possibly has a forced win because he moves 1st a fallacy. In fact, Computers understand counter-attacking motifs, sacrificing a pawn that White must accept or face annihilation. It is Black that Computers will prove to have a forced win ! 

How would we know if chess was actually "solved"? Didn't humans think chess was solved several times in the history of chess?

I agree that any search of the game-tree can still be huge, even if only searching for forced mates. It still might be outside the bounds of today's computing power, but considering there are big projects using tons of computer power looking for extra-terrestrials, I wouldn't be surprised if some group takes up the probably simpler (and more mundane) task of solving chess.

For clarity I made a diagram below to help show how solving chess doesn't necessarily require checking the entire game-tree. If either side can force a mate (left branch) then nothing in the orange needs to be check. This concept can apply from the very first move of the game (if 1.e4 leads to a forced win no need to check other 19 openings). It also applies to intermediate and very small branches all over the game-tree.

As for why some people believe chess is a draw, I don't know where that conclusion comes from. It's possible, but where's the analysis? Using data from human games and engines doesn't mean much because we don't yet know perfect play.

Also, if there is a forced mate for one side, I don't know if there is a reason why it would be longer than an average game. There may be a huge number of paths in the tree that lead to a forced mate. If chess is solved, we will probably find a normal length forced-mate first (40 moves or so) and then discover longer ones later.

Repeatedly saying you don't know where the conclusion that chess is a draw comes from simply shows that you haven't bothered to read or study what the experts of the game (grandmasters), have said, and what the engines are demonstrating. 

Using data from humans and engines doesn't mean anything because we (humans) and engines don't play perfect chess. You can't say "chess cannot be solved" and at the same time say that humans have solved it.

You aren't quoting me. 

I have said many times that not only can chess be solved, we are close to solving it and may solve it in our lifetimes. 

I reject the notion that we must map out every single possible move to claim that the game is solved. All we need to do is create a chess playing program that can never be beaten, even by its clone. That's how checkers was solved, and how chess will be solved in the next 20-30 years.

We are already close to having perfect chess playing engines. The current computer world chess championship is 100 games. Stockfish won the final match, and lost a total of 8 games against Komodo. Both are rated +3200.

Checkers was solved:

"Draughts, English (Checkers)This 8×8 variant of draughts was weakly solved on April 29, 2007 by the team of Jonathan Schaeffer, known for Chinook, the "World Man-Machine Checkers Champion". From the standard starting position, both players can guarantee a draw with perfect play.^[10] Checkers is the largest game that has been solved to date, with a search space of 5×10²⁰.^[11] The number of calculations involved was 10¹⁴, which were done over a period of 18 years. The process involved from 200 desktop computers at its peak down to around 50.^[12]"

Chess will not be.  10^46 is quite a bit more difficult than 10^14.  Over a billion billion billion times more difficult.  Eighteen years won't cut it.

And how do you propose to know beforehand which route to take?

Why do you think we'll find a forced mate within 40 moves when we already have computers that calculate that deep and cannot find a position even close to a mate?

Not to mention you will never find a forced mate as chess itself is a draw.

It has not even been proved that chess is not a win for Black.

You have stolen my joke.

But, the advantage of White is, if we really knew what it meant, the initiative.

If anyone can explain the initiative, I will be very grateful. 

Combinatorial game theory quantifies the notion of initiative precisely with a concept called "temperature". Roughly, hot positions are those where you would want to have the next move, cold positions are those where you would want the opponent to move first. Snooker is a good example of a game with both hot and cold positions. Chess is a little more complicated to apply this theory, but still there are non-trivial insights possible. See for example the beautiful paper by Noam Elkies: library.msri.org/books/Book29/files/elkies.pdf

I thank you for that.

Anyone has something to agree or disagree with that.

As to the original question, I believe that the problem is soluble if the computer has enough storage space and time.

Whether there will ever be enough of that to achieve that even in a million years of design and building before the machine can be even turned on is very debateable.

There has been some progress towards 3D design. But thermal issues are going to be important. CPUs run hot even without reducing the surface area for cooling, eg 1.4 billion transistors in 130mm^2 on mine, generating 77W at design speed. Or a fair bit more when I overclock mine. 

So very low power designs are needed for deeply 3D processors.