Author Topic: Will computers ever solve chess?  (Read 132 times)

Greg Strong

  • Newbie
  • *
  • Posts: 21
    • View Profile
Re: Will computers ever solve chess?
« on: April 20, 2018, 12:19:25 am »
Even conventional computers may be able to brute force given the right algorithms.

Brute force and algorithmic smarts are practically polar opposites.  The only algorithm which successfully reduced the tree size without any risk overlooking better moves was alpha-beta and that's ancient.  It is almost inconceivable that we'll come up with any further breakthroughs like that.  We do, indeed, improve the strength of our chess programs with better algorithms, but this is done by searching deeper in more promising parts of the tree and searching more shallowly (or pruning altogether) less promising branches.  But this comes with the risk of overlooking things and is no longer "brute force."  You have not "solved" chess, technically speaking, if you have skipped parts of the search tree because they didn't look promising.

I suspect the wall you are referring to will be solved and Moore's Law will continue on well past the point needed to brute force Chess.

Moore's Law is already in trouble.  The writing is on the wall due to physical limitations in miniaturization and heat dissipation.  But Moore's Law only addresses the number of transistors in a CPU, it says nothing about computational performance.  We ran into the wall of maximum single-threaded performance a decade ago.  Computers only become more powerful now through more parallelism.  And parallelism does provide some gains in chess computation but with rapidly diminishing returns.  The exponential growth of the search space in chess is so explosive that it would require something on the order of quadrillions of cores running at a dozen GHz for hundreds of years.  So I'm sorry to report that it's pretty much quantum computers or bust.  Fortunately, I think in the next few decades there is a good chance that we will see some very powerful quantum computers.
Like Like x 1 View List