On Thu, Dec 06, 2001 at 02:45:34PM -0500, Dave Ahn wrote:
> On Thu, Dec 06, 2001 at 12:29:04PM -0500, Mark Mielke wrote:
> > This seems religious in nature. I wonder what would happen if the
> > best INL team *was* beaten by a set of bots? Would the loss in
> > pride by sufficient to cause them to give up Netrek? Or would their
> > ambition cause them to improve to a point where they could beat
> > the bot?
> I would suggest that you simply write your bots.  A good place to start
> is Hadley's code and paper.  See www.netrek.org/developer.html and the
> link at the top.  (If you're not an ACM member, email me and I can send
> you the paper.)  I can guarantee that if you were successful in writing
> AI heuristics to get a team of robots to win consistently against even
> an average league team, it would be worthy of publication in ACM or IEEE
> transactions.

I've wrote a bot about 10 years ago. Nobody on the Internet has seen
it, as I only used it on the internal Nortel Networks (used to be BNR)
Netrek servers. This was before the day of RSA in Netrek, back when it
was ridiculously easy to reverse engineer the reserved.c code even on
a stripped binary. I was almost found at when it phasered a few too
many plasmas at ass-range, before the plasma left the docking bay of
the other ship. The first time, the guy considered it luck. The third
time, he complained. My message to ALL, "but only blessed clients are
allowed on this server, are you just pissed that I'm better than you"
only bought me a few weeks of borg-playing time... :-)

My bot was hard at first, but after studying it for several weeks, and
actually playing it, I realized that the problem wasn't with the basic
strategies. The problem was direction. Independent decisions are often
redundant, severely limiting the abilities of a bot "team".

My next attempt will not be along the lines of a traditional bot. All
of these have failed, because they are based on enhancing the existing
client here and there. A true bot is not the same as a fully automated
borg ship. "Auto-pilot" can only bring things so far.

No, my next attempt will have a single process (although not
necessarily single-threaded) act as all 8 clients. Any inputs from the
server will be kept distinct only for the purpose of maintaining
accuracy regarding the source of the pieces of information. For the
most part, *all* actions will be co-ordinated from a single
process. This single process would co-ordinate which ships should go
where, which ships torps where, etc. Each ship will be an arm or leg
to the master mind bot.

Instead of having 8 borgs on auto-pilot all trying to guess what they
should do, we have a single monster bot with 8 appendages, who
assimilates all information available to it, and decides what to do
based on this input.

This has incredible potential even for dog-fights. Why have two ships
torping an incoming ship independently, when the two ships could
alternate torpedoes creating an impenetrable wall of torpedoes, while
minimizing fuel consumption? I wonder how easy it would be for an INL
team to take planets, if they can't manage to get a single kill?

I have a vision. Unfortunately, I have a restricted time
budget. Fortunately, I believe that many other interested parties
exist who are all adequately skilled, and easily motivated by sheer
intrigue... This is what I love about open source. It isn't about
software being free and available. It is about talent sharing, where
everybody can learn from everybody else. The applications are just a
side benefit. I don't mind paying for good software.


mark at mielke.cc/markm at ncf.ca/markm at nortelnetworks.com __________________________
.  .  _  ._  . .   .__    .  . ._. .__ .   . . .__  | Neighbourhood Coder
|\/| |_| |_| |/    |_     |\/|  |  |_  |   |/  |_   | 
|  | | | | \ | \   |__ .  |  | .|. |__ |__ | \ |__  | Ottawa, Ontario, Canada

  One ring to rule them all, one ring to find them, one ring to bring them all
                       and in the darkness bind them...