Showing posts with label chess. Show all posts
Showing posts with label chess. Show all posts

Saturday, February 2, 2019

Fermi estimate of time it takes to solve chess

Landauer limit 3e-23 J/bit in our current universe background radiation, with
$$kT\log 2, T = 3K, k = 1\times 10^{-23}J/(K\cdot bit)$$

Searching for game nodes has about 2e10 nodes/s on a GPU with 3e12 bits/s, so that gives about 100 bits/node.

The number of chess nodes needed to evaluate is about 1e43.

So we get an upper bound of 1e45 bits needed to solve chess, assuming we can't prune any nodes.

That requires 3e22 J at the Landauer limit.

Let's Read: Neuropath (Bakker, 2009)

Neuropath  (Bakker 2009) is a dramatic demonstration of the eliminative materialism worldview of the author R. Scott Bakker. It's very b...