nhaus@eagle.wesleyan.edu (03/30/91)
Following the engineering/military vein which has been introduced to this
discussion:
The Navy has a system, called Phalanx, which is capable of locating a target
out of blue sky (not too hard via radar) and pointing its gatling gun in the
appropriate direction to turn it to swiss cheese. There are a couple of
interesting points about this one:
1. It won't go after birds.
2. It can find its own bullets in its field of 'vision', figure out by how
much they are missing the target, and correct.
3. It works very, very quickly in real time.
I'm not sure on the details, but I am sure of two things: no AI involved, and
it has to be turned on to work (as demonstrated in the P. Gulf a year or two
ago).
------------------------------------------------------------------------------
Fulfillment of human capacity is beyond the capacity of any human.
------------------------------------------------------------------------------
s-mail: Nikolaus Haus | e-mail: NHAUS@eagle.wesleyan.edu
Box 4408 Wesleyan |
Middletown, CT 06459 | v-line: (203) 638-0422
------------------------------------------------------------------------------willdye@typhoon.unl.edu (03/31/91)
nhaus@eagle.wesleyan.edu writes: > The Navy has a system, called Phalanx, which is capable of locating a target >out of blue sky (not too hard via radar) and pointing its gatling gun in the >appropriate direction to turn it to swiss cheese. There are a couple of >interesting points about this one: > 1. It won't go after birds. Actually, it DOES, it just isn't supposed to. Tuning the radar to avoid false signals (like flocks of birds), yet still see legitimate targets is still a problem. That's why there were so many false Patriot firings during Desert Sheild/Storm, yet very few million- dollar misfirings in calm peacetime. When things are calm, the operators keep the signal-to-noise threshold high, to keep from shooting at errant meteorites and such. In wartime (and pre-war), they lower the threshold. This lets few legit targets get by, but unfortunately means you waste a few shots on false signals. By the way, I don't know if these systems used AI techniques or not, but I'd like to point out: 1. What was called AI yesterday is often called engineering today. Remember when it was said that a good chess-playing computer would be 'intelligent'? 2. Techniques called 'AI' today will still be useful for target identification. Fighting signal-to-noise ratios involves having the computer 'know' what is wanted (signal) and what is not wanted (noise) by 'understanding' the overall scene. I agree that today's targeting systems are not AI, but my guess is that learning to track targets using AI techniques will still be worthwhile, since an AI-based architecture may be more extensible than a (relatively) simple "hit the biggest hot spot" system, which is what we have today. willdye disclaimer: I don't know what I'm talking about. UNL knows it. I know it. By now even YOU know it.