Prev: Re: [sg] Eureka 100 Next: Re: [sg] Bagpipers, NAC

Re: [FT] Unpredictable AI

From: KH.Ranitzsch@t... (K.H.Ranitzsch)
Date: Thu, 21 Jun 2001 08:00:19 +0200
Subject: Re: [FT] Unpredictable AI


----- Original Message -----
From: "Richard and Emily Bell" <rlbell@sympatico.ca>
> It is not that the AI is predictable, it is that it is mindnumbingly
> stupid

But it is getting better, and it not clear that a space fight with a few
fighters is all that complex an environment.
Some time ago 'New Scientist' reported on RAF test with a neural network
system that was quite nasty in fighter simulator combat.
Note though, that the planned pilotless planes are more akin to
remote-controlled vehicles with built in auto-pilot than real AI.

> and the problems faced by pilots maintaining situational awareness
while
dodging fire is not solved merely by being really fast.

A fair amount of the problem is that humans are not that good naturally
at
thinking in 3 dimension (4 if you include time). With a decent all-round
sensor suite, a computer will have no problem knowing where every one is
and
where they will likely be a few seconds hence. What to do about it is
the
more likely problem, plus, as other have pointed out, discriminating
between
real targets, dummies, neutrals and friendlies. The latter would not be
a
problem in a free-fire zone, of course.

> Unlike playing chess or diagnosing engine problems, most of a pilot's
skill set  are psycho-motor skills which can be learned, but not taught.

An electronic brain hard-wired into the plane doesn't need any
psycho-motor
skills.

> Humans are also much harder to fool than computers

I would rather say that it takes different tricks to fool a human and a
computer.

Greetings


Prev: Re: [sg] Eureka 100 Next: Re: [sg] Bagpipers, NAC