Prev: Re: [OFFICIAL] Some FT background stuff (guidelines for writers) - LONG POST! Next: Re: FT Die, Fleet Emblems

Re: Some FT background stuff (guidelines for writers)

From: Thomas Barclay <Thomas.Barclay@s...>
Date: Thu, 12 Feb 1998 11:45:57 -0500
Subject: Re: Some FT background stuff (guidelines for writers)

> > >Space is NOT emptier than southern Saskatchewan. I've been there.
> > >you look in the dictionary under 'empty', that is what you see.
> > 
> > Yes, but space is curved while Saskatchewan is flat. Isn't
> > Saskatchewan's motto: Can't Die from Falling?

... but you can die from Boredom. (heh. I lived in Southern AB so I 
can speak with some authority....) :) 

> Now, what happens to computer pilots when their radar is jammed?  Or 
> fails?  Or returns ghost signals?  

Assuming something like modern tech, where you either use A) 
structured programming or B) neural nets, then "when 
the radar is jammed" scenario played out in A), the routine for 
"radar jammed" would be invoked and the fighter would carry out 
appropriate code in there, which involved using other sensors or 
being just as blind and random as a human. It would include a true 
white-noise random number generator to be used in helping to make 
decisions (to simulate human randomness). It would include a lot of 
options with weightings. If it had the compute cycles, it could 
evaluate each and every one of these options and follow the best 
course, or choose randomly between half a dozen that give near 
equal results. 

In case B), the neural net would have been trained 
with some arbitrarily large sample of these types of events, and 
would do its neural net magic and pick an appropriate response - note 
that with neural nets you do not program in responses, you train the 
net and it responds as it thinks it should. You then evaluate the 
results until you think you have a good neural net. Then you 
implement that net. Same result with the other options. Basically, 
what you need is a lot of development money, and a bunch of smart 
fighter jocks and engineers to figure out every scenario that could 
happen (or at least 95% of them). The interesting thing about a 
neural net is that if you have an unknown scenario, based on ones it 
has seen it can still come up with a response. 

Unspoken issue: How fast can you train a neural net in 2300? 

> I guess my point is:	in a 1 v 1 scenario, head-on, no positional 
> advantage, the computer pilot does have the edge.  But how many times
> does this happen in warfare?

In gaming, all the time. (heh)
Thomas Barclay
Software Specialist
Police Communications Systems
Software Kinetics Ltd.
66 Iber Road, Stittsville
Ontario, Canada, K2S 1E7
Reception: (613) 831-0888
PBX: (613) 831-2018
My Extension: 2034
Fax: (613) 831-8255
Our Web Page:

Prev: Re: [OFFICIAL] Some FT background stuff (guidelines for writers) - LONG POST! Next: Re: FT Die, Fleet Emblems