“We
believe that the current crop of AI systems today are going to be cognitive
assistance,” he [Nand Mulchandani, acting director of the Joint Artificial
Intelligence Center] said. “Those types of information overload cleanup are the
types of products that we’re actually going to be investing in.” (1)
So, the military seems to view AI as a data ‘clean up’ or
streamlining service which will, of course, be used by remote commanders to
more closely micro-manage their subordinates who are the local commanders.
The military is also intensely interested in micro-managing
the assets in the field by connecting sensors and weapons and exerting control
over them (see, “Command and Control”).
…
DoD’s focus on developing JADC2 [ed., Joint All-Domain Command and Control], a
system of systems approach that will connect sensors to shooters in near-real
time. (1)
AI is also viewed as a common connector between disparate
platforms – whatever ‘platforms’ means in this context – so as to, again,
enable more effective micro-managing.
“JADC2
is not a single product. It is a collection of platforms that get stitched
together — woven together ― into effectively a platform. And JAIC is spending a
lot of time and resources focused on building the AI component on top of
JADC2,” said the acting director. (1)
We see, then, that the military views AI not as an
autonomous, near-sentient, battlefield killing machine but as a tool to aid in
the exercise of even closer and tighter micro-managing. They don’t say it in exactly those words but
that’s what they want. The Admirals and
Generals grew up in a zero-defect environment where you didn’t trust your
subordinates to operate without your moment-by-moment direction. Ship command, for example, was an
excruciating period of time where you, as Captain, prayed every minute that
none of your crew would make a mistake that would result in the dreaded ‘loss
of confidence’ pronouncement from your superiors and the loss of your command
and career. You achieved this by
micro-managing every action aboard your ship.
Not surprisingly, then, today’s Admirals want to apply the same model of
command – meaning micro-management – to large scale Command and Control and
they see artificial intelligence as the way to do it. The more you can see what your subordinates
see, the more closely you can micro-manage them and AI is just the thing to
allow that – so the Admirals believe.
I’ve repeatedly posted and commented about the ills of
micro-managing. There’s no need to
repeat myself. What I’d like to address
is not the self-evident evils of micro-management but, instead, the proper use
of AI in the conduct of a war. What our
Admirals (and civilian leadership!) should be using AI for is not assisting in
micro-managing but, instead, assisting in the formulation of strategy. We should be using AI to identify enemy patterns
of resource use that might indicate strategic vulnerabilities, identify overall
enemy force movements that might presage future operations, assess enemy
leadership performance and patterns that might allow us to predict actions and
decisions, track enemy munitions expenditures to identify weapons usage
patterns and inventories, and so on.
None of this is directed toward telling the individual soldier or sailor
which way to turn and when to pull the trigger.
None of this is directed toward telling the individual ship’s Captain
what course to steer. Admirals and
Generals should have much bigger problems to worry about than setting a course
or monitoring individual soldiers.
 |
| Strategy - Not Micro-Managing |
Properly used, AI has the potential to be a strategic-level
aid for upper command. One of the
difficulties, I’m sure, is that our current – and future – leadership has been ‘raised’
in an environment totally lacking in strategic and operational thought. Thus, it’s undoubtedly difficult, if not
downright impossible, for them to even conceive of using AI for strategic
purposes. They’re simply not wired to
think about strategy. They never had to
before so why would they now? Somehow,
we need to break out of that limited thinking and start thinking strategically
instead of micro-managing. AI can offer
valuable assistance but only if properly applied.
(1)C4ISRNet website, “Pentagon AI center shifts focus to
joint war-fighting operations”, Nathan Stout, 8-Jul-2020,
https://www.c4isrnet.com/artificial-intelligence/2020/07/08/pentagon-ai-center-shifts-focus-to-joint-warfighting-operations/
I agree fully.
ReplyDeleteThe famous Germán blitzkrieg wasn't something that relied in better tanks and weapons but in the doctrine of giving independence of command to subordinate commanders.
If I were in charge of chinese military strategy I would have already a plan to cut or interfere communications in USA command chain.
JM
You know I read the whole story and I still am not sure what hell they are actually doing. I get they spending more money. I get thay are apparently stopping a focus on something - predictive maintenance - that seems like something potentially useful. I assume that was something like collating data and based on say a ships recent service in such and such climate maybe some parts should worked or changed even if its not technically time (do to service life. Perhaps guiding the crew to focus on the most high probability of failure items in maintaining anything from a gun to truck etc.
ReplyDeleteBut as far as I can tell that is now a project designed to produce power point presentations and buzz words to feed congress.
"But as far as I can tell that is now a project designed to produce power point presentations and buzz words to feed congress."
DeleteAnd you said you weren't sure what they were doing! You nailed it!
"a project designed to produce power point presentations and buzz words to feed congress."
DeleteWell, at least they're sticking to what they know?
Inspired by your post, I read "Joint All Domain Command and Control (JADC2)" posted by the Congressional Research Service*.
ReplyDeleteTo me, this appears to be yet one more effort to fuse sensors and information streams - "intelligence, surveillance, and
reconnaissance data" - to enable more rapid and reliable decision-making at higher organizational levels than the individual surveillance or reconnaissance platform or unit.
Working on interoperable communications networks is always a good thing. But the trouble will be, and always has been - in my experience - that fusing target data from multiple sensors and platforms is never easy because of uncertainties in timing, location, coverage, false alarm rates, propagation conditions and operating status.
You recommend "formulation of strategy" as the proper use of AI in the conduct of a war. This seems to me to be more of a data-mining and modeling process that is assisted by AI, for detecting and classifying meaningful patterns.
*https://fas.org/sgp/crs/natsec/IF11493.pdf
"to enable more rapid and reliable decision-making at higher organizational levels than the individual"
DeleteThis sounds good on paper but never works in practice. How many of us curse failed networks, accounts, data, etc. in our daily lives which are unhindered by combat? Networks just don't work. Period.
Now, toss in the effects of enemy resistance, jamming, cyber attacks, hacking, comm relay losses, and the fog of war and our networks are going to be slow and utterly unreliable.
Worse, by pulling the decision making up the chain, we remove decision making, initiative, creativity, and independent thinking from the lower levels. They become a dumb pair of remote trigger-pulling hands who have been conditioned through their training to do nothing without higher level orders. Given that war is 100% certain to disrupt our flow of commands, we're going to wind up with units sitting around waiting for orders and lost about what to do when their micro-manage orders don't arrive.
We've always claimed that individual initiative was one of our biggest advantages in combat and yet we're actively working to take that away. Does that make sense? Does that seem wise? We should be doing the exact opposite by pushing the decision making DOWN the chain to the lowest possible level. Let the leadership deal with strategy and let the local units deal with execution.
We're intentionally setting ourselves up for failure.
"Perhaps guiding the crew to focus on the most high probability of failure items in maintaining anything from a gun to truck etc."
DeleteYou're describing predictive maintenance, essentially. Interestingly, we've tried that twice on a large scale and failed spectacularly both times. The F-35 ALIS was intended to be exactly that kind of predictive maintenance program and it crashed big time. The LCS with its land-based maintenance model was also a predictive maintenance effort. The LCS is crammed with equipment sensors which are supposed to transmit mega-giga-gazilla-bytes of data back to shore facilities to predict LCS maintenance needs and have the required parts and people waiting to jump aboard as the ship ties up to the pier. The reality was that the comm channels hopelessly inadequate for the transmission requirements, no useful data was collected, and the entire process fell apart.
So, 0 for 2!
"Inspired by your post"
DeleteThen the post was a success!
So if your are a PLAN Admiral, which is more to your advantage, leaving the USN magic micro-management network (M3N) in place, with some false information added.
DeleteOr disrupting the the network, allowing local commanders to take control.
More seriously, AI in a NSM seeker that can tell the difference twixt a chaff cloud, a corner reflector, a MALD and a warship would be useful.
"So if your are a PLAN Admiral, which is more to your advantage, leaving the USN magic micro-management network (M3N) in place, with some false information added.
DeleteOr disrupting the the network, allowing local commanders to take control."
Great question, actually.
Feeding the network fake information would have the USN falling for it 99 times out of 100, at least for a while, and shutting it down entirely would just freeze everything given the gigantic over-reliance on networks for even mundane task.
"Feeding the network fake information "
DeleteOr … they could just let us continue as we are and self-destruct without their help!
"Working on interoperable communications networks is always a good thing"
ReplyDeleteUp to a point a very interoperable network is also potentially a vulnerable and non redundant one. People tend to ignore security and best practices until after the barn doors were left open.
In 1976, I attended a Navy logistics course as my two-week ACDUTRA. The course included both active and reserve personnel, and one of the attendees had been CO of the USS Harold Holt during the Mayaguez incident. He talked of being on the HiCom radio network with Washington DC, and reporting things like, "My rudder is right 15, Mr. President, coming to new course 290." That seems to be way, way too micro for my thinking.
ReplyDelete"That seems to be way, way too micro for my thinking."
DeleteThose were the good old days. Now, we're fusing sensors and helm control so that the President can steer the ship directly!
"Properly used, AI has the potential to be a strategic-level aid for upper command."
ReplyDeleteWhy couldn't AI be used, for the same purposes you specified, at the tactical level? I agree that AI shouldn't be used to direct individual movement of ships, aircraft, or troops. But, any battle commander has to know his enemy's capabilities and intentions in order to achieve his own goals. An AI tool could game out different plans of attack or defense, analyze and data mine intelligence reports, track and predict future enemy movements, and identify enemy vulnerabilies to exploit and your own to reinforce.
It's one thing to use software to identify general trends but it's another to expect it to direct low level, specific actions. Taken to its logical conclusion, each solider should have an AI telling him where to step, turn, look, and crap.
DeleteIt's like predicting the weather. A computer can predict general trends but expecting it to predict the exact moment by moment weather is hopeless.
Further, predictive programs are useless without adequate data. Expecting, say, a squad level leader to have enough data for a program to accurately predict enemy locations, movements, strengths, weapons, intelligence, supplies, etc. is hopeless.
Hello ComNavOps. You might be interested in the story of Marvin Creamer. He circumnavigated the globe without any navigational instruments. The New York Times has a good write-up on him.
ReplyDeleteYou have previously written about the Navy's overreliance upon GPS and radar navigation. Creamer's methods are too primitive to be applied to the entire fleet, but it is not unimaginable for the Navy to run a nautical program analogous to land-based SERE. At the very least, the Navy could teach sailors how to navigate for themselves and to not rely on a machine.
https://www.abc.net.au/news/2020-09-08/bill-birtles-mike-smith-evacuated-china-safety-concerns/12638786
ReplyDeleteAustralia has no reporters in China. Our last two have spent days hiding in Embassies and Consulates.
This is after one of China's premier newsreader, an Australian Citizen, was detained for national security reasons. See https://www.abc.net.au/news/2020-08-31/australian-detained-in-china-television-anchor-cheng-lei/12615362