Prev: Re: AI's in full thrust Next: Re: Honor Harrington Question?

Re: AI's in full thrust

From: schoyt@f...
Date: Sun, 13 Jul 1997 23:46:49 -0400
Subject: Re: AI's in full thrust

First, to the individual who asked about novel length Bolo books:  "Bolo
Brigade" by William H. Keith, Jr.  A good book - I'd recommend it.

> Really smart AI's would probably still take up a considerable amount
of
> space.

Huh?  We've gone from room sized computers to desktops and the
theoretical limit on miniaturization of a mechanically operated computer
is using individual electrons to activate a binary switch.  Even a
system with huge computational requirements could be very compact at
that scale.
 
> As far a judgement and AI's  on capital and smaller ships. One
generally
> wants crewmen/seamen on hand for damage controls, maintanance and
such.

If you have such intelligent computers, they will know exactly what is
wrong much faster than a human technician, who will be using a
diagnostic computer anyway.  Besides, repairing space battle damage
ain't like getting some lumber to shore up the bulkhead after a hull
breach.  Nanites would be spread throughout the ship, and outer hull
damage would be repaired almost like human skin, with a scab forming and
the hull rebuilt underneath by the host of molecular sized damage
control parties.  Same for internal systems, directed by damage control
computers with instaneous access to complete schematics of every aspect
of the vessel.	Damage control computers would have reserve nanites with
a sole function of repairing damage control computers - no response from
the damage control computer sends the nanites looking to repair the
computer, which then can direct the rest of the DCNs (damage control
nanites) throughout the rest of the ship.  Instant redundancy.	As long
as damage doesn't occur more rapidly than the DCNs can act, the ship can
at least beat a retreat to repair itself at leisure.  This also assumes
a good supply of building material on board also.

> You want experienced commanders in the CICs on
> those naval vessels handling the delacate situations, not computers.

I agree.  The onboard AI doesn't have to be completely sentinent, just
intelligent enough to assist a human commander (possibly bio-engineered
to withstand greater g-forces or to have direct interface via his brain
to the computer to cut down reponse time), who can control the ship,
albeit in a extremely reduced manner, in case the computer goes
HAL9000.  Interactions with the commander might be as such, occuring
directly from neural interface, taking only a few seconds:

AI: "Alert.  Emissions consistent with Enemy vessels detected at 2 light
minutes.  IFF reponse also negative.  Projected course indicated. 
Intentions indicate direct assault on Friendly orbital facility."
HUMAN:	"Concur.  Display least time intercept course." <schematic
displayed on retina> "Display other Friendly units that can assist
intercept." <displayed>
AI:  "Graphic takes into account communication lag time to signal our
intercept intent."
HUMAN:	"Designate as Plan Alpha.  Transmit to Friendly units and
execute."
AI:  "Recommend following change to course to take advantage of other
Friendly units weapons capabilites." <retina display>
HUMAN:	"Concur.  Execute Alpha revised."
AI:  "Transmission sent, executing course corrections.	Additional
engagement parameters, Commander?"
HUMAN:	"None.	Automatic engagement at optimum weapons range."
AI:  "Do you require notification before commencement of fire?"
HUMAN:	"Negative.  Status reports only.  Standard command overrides
apply."
AI:  "Of course, Commander."

The human acts only in a supervisory/planning capacity while the AI
provides instant data for decisions, recommendations and simulations as
required.  All actual fire control solutions, plotting data, and course
corrections are all handled automatically by the AI.  The modern day
Aegis system works much the same, so why wouldn't we have more
sophisticated computers in the future handling even more?

> As far as fighters go, Jamming and countermeasures can work
surprisingly
> well on smart weapons. Even brilliant weapons will have trouble with
some
> of the newer countermeasures. Having a human operator to fudge the
pickle
> onto where he thinks the target is can really help in sticky
situations.

Countermeasures also DON'T work surprisingly well on brilliant weapons. 
They can remember a ship or radar site' location after being disrupted
by chaff.  They can be taught to ignore flares.  And this is 1997.  With
the computational power that will be available in 2197, you can get a
pretty sophisticated brain in a weapons package.  If we applied this
"human operator" logic to modern naval warfare, we would still be firing
broadsides from cannon because the enemy MIGHT spoof our missiles.  You
don't see fighters flying escort on cruise missiles to make sure they
get to their target, do you?  The only reason we don't use more
autonomous weapons today is that we're not willing to spend the money to
full automate everything and that we still have improvements to make
before we fully trust an automated system.  Some things will have to be
computer controlled because humans are too slow, like the Vulcan/Phalanx
CIWS, for example - no rows of 40mm pom-poms lining the sides of modern
naval vessels.	If you want to see the way naval design is going, look
up the Arsenal Ship currently being discussed by the Navy.  Heavily
automated with a minimum crew and lots o' weapons.  Many of the
discussions here are being hashed out by a more august body of thinkers
than ourselves.      

> (Aliens, 2001, Alien, Great Space Battles)

These are science FICTION, not some scientific study.  It's not as if
mistakes won't happen if humans are in charge.	Friendly fire and
mission failure happens all the time with human commanders.  Just
because an AI might fail is no reason not to use them - we'd all be in
trouble if that criteria applied in life.

Sean

Prev: Re: AI's in full thrust Next: Re: Honor Harrington Question?