On Sun, Apr 30, 2017 at 12:17 PM, Jeffrey Schwartz <xxxxxx@gmail.com> wrote:

Suppose you loaded thousands of swords into a database, then tagged
which ones were used in documented fights against others, and the rate
of win/loss.
Then ran some deep learning/genetic change algorithms against that
until you had a design that reliably won against everything else.

As others have pointed out, it doesn't work that way.

Weapons tend to evolve in companionship with armor and other resources.

Things tend to be optimized for their environment and expected use.

Taking something out of the environment for which it was designed, and using it in a different environment, does not always work well.

An Indy race car is great for that purpose, but isn't going to work well as a commuter rig. A two seat sports car may be great, but it's not the vehicle to take off-roading. A HMMWV is a great ride, unless you're driving through a mine rich environment. etc.

Dan

"Any sufficiently advanced parody is indistinguishable from a genuine kook." -Alan Morgan