On Mon, Mar 2, 2009 at 1:06 PM, Troy Guffey <[email protected]> wrote:
> --------------------------------------------------
> From: "Knapp" <[email protected]>
>
>> BTW we are talking robot computer brains, not autopilots.
>
>
> Robot computer brains are only smarter autopilots.   The difference is only
> in *degree*, not *kind*.
>
> --
> Troy Guffey

I think you are wrong here because AI can do things like figure out
the best way to go but an Autopilot is just a simple state machine. I
give you that autopilots will evolve into AI, if you don't think they
have already. Using ideas like the Monti Carlo to find the best
solution to a problem not something I would expert from my autopilot
but would from an AI based robot or AI autopilot. This is a different
*kind* of system. AI can make decisions about when and where or even
if, to land. Autopilots just follow a preset course.

http://en.wikipedia.org/wiki/Monte_Carlo_method

The short of it (MC) is that a chess program can look at the board and
give a numerical evaluation of the board but with games like go this
does not work. The way through this problem is to look at all possible
NEXT moves from a given position and from that, have both sides make
random moves to the end of the game 100 (or however many times you
have time for) times for each NEXT move. The move that has the highest
percentage of wins is the next move the AI plays.

This is relatively fast, does not involve looking at every move that
could be played and allows for games that have changing final
solutions. Down side is that you might not randomly come up with the
best move because the 100 random games played did not include the best
move that could be played but some other randomly chosen move.

In short Autopilots don't think, AIs do.
-- 
Douglas E Knapp

Why do we live?
_______________________________________________
GurpsNet-L mailing list <[email protected]>
http://mail.sjgames.com/mailman/listinfo/gurpsnet-l

Reply via email to