Will computers ever solve chess?

ozzie_c_cobblepot wrote:

New way to frame the problem: Will a computer ever be created which will never lose to any subsequent computer?

I would argue the answer is no.

-------------

I disagree. We are already close to the point where an engine will never lose. By close, I'm thinking of ~20 years.

I think we'll see 100% draw before perfect play, but you're thinking this soon? Wow.

So for example you think an engine made 20 years in the future will never lose a game to an engine made 30, 40, or 100 years in the future?

Square circles seem to qualify

As long as programs will suffer horizon effect, I think the question is left open.
If a win is found at least 2 moves before the horizon, then it's game over. If not, the second player will see one move further than the first one, and might draw or even win.

(I assume that both sides are running the same soft).

Since there are forced mates in 549 moves at best, how will chess will ever be solved by computers?

http://tb7.chessok.com/probe/3/61

7 men tablebases were such a challenge. 32 men tablebases will never exist. Even 20 men are dubious

  No one's stopping you. Besides, solving chess is secondary to human problems. Just saying.

  I wouldn't use the plural in that phrase. There is only one 'thing' that cannot be imagined. 

   Also, if it cannot be imagined, there is no concept that can cover it.

  If it's obvious when you hear it, then it is, after all, imaginable. It's not recognizable.

Given that these tablebases now have 500 trillion positions it is safe to say that the large majority of them have never been seen on the board.

I would definitely put my money on the 40-years-in-the-future computer *creaming* the 20-years-in-the-future computer. The hardware alone will be light-years better, and the software has two decades to power through the incremental improvements.

Plus, the "40" computer would have access to larger endgame databases, and more-developed opening theory.

I don't know what the end result of 100 games would be, but 70-30 would not be out of the question.

Unless of course the solution is what most people expect, and P doesn't equal NP

There is no doubt that this is so. A very low percentage.

Correct, you said when P vs NP is solved chess will be solved.

This is incorrect though, because one possible solution is that P doesn't equal NP.

The status of P vs NP is utterly irrelevant to chess as we know it. Chess on an 8x8 board, having a finite solutions space, is trivially in P -- and solvable in O(1) time (with a very large constant).

Generalize chess to a NxN (or even 8xN) board, and now you can start thinking about computational complexity.

Ok, yeah, should be something like "chess like games will be solvable in polynomial time if P=NP."

AI will run everything.

If (when?) people trust a super-smart AI to give input on certain society-level decisions it could help a lot though I think.

Instead of politics being about charisma and personal gain we could squeeze in some data based and logical decisions.

That explains a lot