On Mon, 9 Jan 2012, Onno Meyer wrote:
Eric replied to me:
[...]
Of course, a millennia-long empire might have super-ais programming
attack programs for other super-AIs in a spiral...
Hello Eric,
that was the idea. You 'seed' the AI with the best skill programs you
have, then send it out on a long mission (OJT without a teacher), and
evaluate if the experiences have improved the program. Perhaps even a
competition between several ships with the same initial program, or
different programs with various options: aim longer or shoot earlier?
The problem may be to find new and challenging situations. How do you
train gunnery, tactics, and sensor operations if there is no shooting
war with a peer competitor? To many wargames, and the AIs may rate
each other for a hobby skill "wargames" and gunner (sports).
Then you load a computer programming skill program, and have the
winner program the next round of 'seed' skill program.
If you have that good computers, you likely have very good simulators as
well.
The main reason i see for developing wargames and sports skills are
limitations by your simulator, and that you stop to train defending
against tactics, that in the long run are not successfull.
The first will be few, because the simulators are good, and if you are
willing to pay some expenses, you can even have some ships really shoot
each other at times, to check the viability of the simulator. You can make
security copies of the AIs first.
The second can be avoided, by keeping AIs with outdated tactics just for
testing purposes.
_______________________________________________
GurpsNet-L mailing list <[email protected]>
http://mail.sjgames.com/mailman/listinfo/gurpsnet-l
