[ExI] future of warfare again, was: RE: Forking

[Anders Sandberg]

On 2012-01-01 18:04, spike wrote:
> Agreed, but the focus of the more sophisticated modern war machinery,
> isn’t aimed at the personnel, but rather the other machines of war. The
> modern warrior has nothing against the adversary’s guys. They can have
> as many guys as they want, for without the sophisticated mechanisms of
> warfare, they are as harmless as an army of kittens.

Yes and no.

In high-school one of my friends, Fredrik, was a would-be military 
officer. Just as I knew I would become a scientist of some kind he was 
planning out his straight-arrow military career. We had a long running 
argument about the future of the military: I was pointing out that 
robotic or drone warfare would eventually become possible and more or 
less take over the field, leaving the military as a bunch of nerds at 
keyboards. He countered with "You are always going to need a guy on the 
ground with a rifle".

He had a point. As shock and awe demonstrated, high-tech can wipe out a 
low-tech military infrastructure. Drone warfare can hit enemy 
concentrations and individuals with reasonable precision. But these 
tools cannot occupy a country: maintaining civil order, gaining human 
intelligence, instilling trust for whatever institutions you are trying 
to set up, that requires personal interactions... and those guys with 
rifles. Some of the more obvious failures in recent Middle East 
conflicts have been due to the discrepancy between overwhelming 
projectable force and lack of "social" interfacing.

It might be possible to enhance the guy with the rifle. Perhaps drone 
infantry will appear in the next few decades ("I'm the neighborhood 
soldier of Tohid Square. I patrol 9 to 5 US time, very convenient for 
me. Sure, occasionally my bodies gets blown up, but it is mostly a 
budget problem...") Maybe they can be networked in smarter ways, like in 
Adam Robert's "New Model Army" (an anarchist ultra-flexible wiki-army). 
Or ubiquitous surveillance systems can be used. But it still seems that 
this is a major bottleneck since the complexity of the tasks is orders 
of magnitude higher than in the direct attack phase.


The essence of attacking something is to prevent its function. This can 
be surgical, with minimal effects on the surroundings or unrelated 
functions, but you need to have plenty of information about the target 
and its state. This is why maximizing the entropy of a target is so much 
easier as an attack mode: you do not need much information, and hence 
the attack is likelier to succeed in low-information or adversarial 
information environments. We have nearly maxed out our ability to apply 
entropy: the remaining big frontier is precision.

The real aim of attacking stuff is of course control. It has the same 
problem in terms of information as destruction. Typically warfare aims 
at prevening the function of defenses/offenses, and then tilting the 
utility function of the enemy using threats so that enemies now behave 
as you would like them to. When this works you effectively turn enemies 
into parts of your system since they now make use of information they 
know but would be hard for you to come into possession of to do things 
you order. The problem is of course that you are not necessarily 
changing their utilities with your threats well enough (too small 
threat, too different base utility, lack of information), and you now 
have to handle an unreliable system. I am reminded of the issue of 
"weird machines" in computer science, 
http://boingboing.net/2011/12/28/linguistics-turing-completene.html - an 
occupying force are essentially interfacing with an unsecured system.

Figuring out the control problem is the *real* challenge for future 
armies - the guy with the rifle is there to ensure a certain range of 
behavior on the microscale. But as the above linked talk suggests, this 
is likely a computationally infeasible problem. You can likely solve it 
better than currently, but it can never be solved generally - and you 
cannot ever be sure your solution doesn't contain some exploitable flaw.

Poor Fredrik.


-- 
Anders Sandberg
Future of Humanity Institute
Oxford University