Autonomy results from delegation of a decision to an authorized entity to take action within specific boundaries. An important distinction is that systems governed by prescriptive rules that permit no deviations are automated, but they are not autonomous. To be autonomous, a system must have the capability to independently compose and select among different courses of action to accomplish goals based on its knowledge and understanding of the world, itself, and the situation.

U.S. Department of Defense[1]

[Sorry, folks, we were going to run one of our golden oldies this time, but then someone tossed a copy of a new DoD Report over our transom, and we were fascinated. Shiny objects do that, especially when they involve new technology and ideas that border on the science fictional. So immediately we rescheduled the blog on George Washington and political debates for a later date, and swapped out editors. So I won’t be with you this time. Instead Fred, our master of the preternatural and bureaucratic, will lead a discussion of autonomous robots, and how our military might use them. Or defend against them, if necessary.]

Yes I will, but first let’s talk about the basics.

Mobile Weapons Platforms

Our military are fascinated with weapons that can be deployed on short notice and remain elusive until they are used. Fortifications, walls, etc. are the opposite of that. They defend one thing or place, and often do it well; but, of course, they don’t move. Land-based missile systems have similar deficiencies. They are offensive, rather than defensive in nature; some can reach out great distances; but they’re essentially fixed in place. They mostly hide in holes, hardened against attack, and wait to launch. Potential enemies know where to find our forts, walls and land-based missiles, and some may have the resources to attack them.

That’s why our military like the so-called mobile weapons platforms; principally ships and aircraft; they’re not stationary and easily targetable, and some can take the field for long periods without reporting back for fuel, supplies and so forth. Navy carriers carry a large complement of aircraft, capable of attacking targets at a considerable distance; other surface vessels [and perhaps the carriers as well] can launch missiles to do the same thing; and, of course, the Navy undersea fleet of nuclear submarines is elusive, carries a significant number of strategic nuclear missiles, and can stay deployed for months at a time.

Early Air Force Experiments

Air Force bombers and fighters are extremely mobile; once launched they can take the fight to an enemy, drop bombs or launch missiles as appropriate, and land somewhere to refuel, rearm and get back to the fight. And that – the landing part – is a significant limitation. An airplane on the ground can be attacked more easily than one in the air, and air strips don’t move around the globe. So it should be no surprise that, while the Navy was developing the nuclear submarine in the 1950’s, the Air Force experimented with building a nuclear-powered aircraft that could remain aloft almost indefinitely.[2]

The first experiment involved modifying a B-36 bomber, then the state of the art, to carry to carry a 3 megawatt, air-cooled nuclear reactor in its bomb bay.[3] While the aircraft flew with the reactor operating, the reactor was never used to power the aircraft. The reactor was turned on simply to determine what effect, if any, the radiation it generated would have the bomber’s instruments and crew. The program was canceled before it got much further than that.[4]

Then there was the nuclear-powered cruise missile, called Project Pluto. In 1957, the Air Force and the U.S. Atomic Energy Commission contracted with the Lawrence Radiation Laboratory to study the feasibility of applying heat from nuclear reactors to ramjet engines. The program was intended to develop a propulsion system for an unmanned cruise missile.[5]  Some say that it would have been highly unsafe for bystanders and civilians. ”Designed to travel at Mach 3 at extremely low level, Project Pluto was a cruise missile powered by an unshielded nuclear reactor that not only leaked radiation but actually shed radioactive debris from its exhaust.”[6] On the other hand, theoretically it could have stayed in the air for months at a time.

Project Pluto was canceled in 1964.  One could say it showed that a nuclear-powered ram jet was possible, but perhaps the war planners of that day concluded that ICBMs [i.e., Intercontinental Ballistic Missiles] were a better solution to their immediate problems.[7] Anyway, it’s not clear [to me] whether DoD followed-up on nuclear ram-jets. Only the Pentagon can say. But we do know, for example, that NASA has used nuclear-based propulsion systems on some of its missions.[8]

But these are questions for another day.

Autonomous Systems

So let’s talk about autonomous systems. One of the limitations on manned flight is how long a man – or woman – can stay in the cockpit. People are big, and need to get up and walk around every once in a while; but what if DoD replaces them with a small box of integrated circuits, software and radio transmitters/receivers? What if that box of parts can make some of its own decisions, rather than wait for a human? What if it’s at least partially autonomous?

This is where the recent DoD report comes into the picture. In 2015 the prestigious Defense Science Board[9], was asked to “identify the science, engineering, and policy problems that must be solved to permit greater operational use of autonomy across all warfighting domains.”[10] Translated, it was told to study how to improve weapons systems by moving some of the analytic functions, normally done by humans, to the equipment itself. I’m not overstating this. Remember, the objective was to create more autonomous systems, i.e., systems that “have the capability to independently compose and select among different courses of action to accomplish goals based on [their] knowledge and understanding ….”[11]

Why is it possible today to do this? Well, largely because of recent advances in artificial intelligence. The Board says: “The primary intellectual foundation for autonomy stems from artificial intelligence (AI), the capability of computer systems to perform tasks that normally require human intelligence (e.g., perception, conversation, decision making). Advances in AI are making it possible to cede to machines many tasks long regarded as impossible for machines to perform.”[12]

Here the Board distinguishes between autonomy at rest, and autonomy in motion. The former is the kind of autonomy exhibited by systems that “operate virtually, in software, and include planning and expert advisory systems….” Systems that incorporate autonomy in motion, on the other hand, “have a presence in the physical world and include robotics and autonomous vehicles.”[13]

Some Examples

Heretofore we’ve focused on weapons systems, which, of course, are very much a part of the physical world. They’re designed to keep the peace by destroying the other side’s capability to inflict harm on us or our allies. To the extent we make those systems more autonomous we may improve their ability to move about the world and fulfill their assigned missions.

Perhaps it’s best to move from the abstract to more concrete examples. The Defense Science Board lists many, but for our purposes we’ll focus on just a couple to make their basic point.[14]

  • Consider, for example, the possibilities for so-called unmanned aircraft [UAs; or drones]. If they are equipped to operate autonomously they could assume “functions now performed by manned aircraft in areas that are difficult to access.”[15] By “difficult to access” I think the Board means “dangerous for humans to access.” Such functions include “aerial refueling, airborne early warning, intelligence, surveillance, reconnaissance, anti-ship warfare,” and so forth.[16]
  • Additionally, a large UA could host smaller ones that specialized in electronic warfare, communications jamming, or acted as decoys, target emulators and so forth, “to confuse, deceive, and attrite adversary attacks.”[17]
  • Then, of course, there’s a lot in autonomy to benefit the Navy. Even today the Navy operates a cluster of devices known as USVs [unmanned surface vehicles] and UUVs [unmanned undersea vehicles][18] Many types of these are available commercially, and some might be used to deploy acoustic and RF decoy payloads;[19] extend capabilities and maintain a covert option pending commencement of electronic warfare;[20] establish communications systems, and create a minefield of autonomous lethal UUVs.[21] And U.S. submarines are in danger as well. “To mitigate the risks, the U.S. must be more proactive and complement their submarine force with other capabilities, such as powerful new autonomous UUVs and sensor networks.”[22]

And so on, and so on, and so on …


I’ve read this report, and only heard one side of the story, but it’s convincing nevertheless. Autonomous technology is here to stay, and so long as we continue to refine and expand our understanding of artificial intelligence, the rest inevitably will follow. I’ll worry most, I guess, if the Pentagon announces that it’s also reviving the nuclear ram jet. Then we’ll have the prospect of a cruise missile that can fly indefinitely, has a brain, continually evaluates targets, and never gives up. Who knows, in a really destructive war, it might outlast even the humans who sent it. If you see one like that coming your way, and there’s no reasoning with it, then you’ll have to run away or find a way to turn it off. Just say, “Gort, klaatu barada nikto![23] and see if it stops.

Good luck on that.

[1] See DoD, Defense Science Board, Summer Study on Autonomy (July 2016), available at  From now on we’ll cite this report as Autonomy at __ The quote appears at About autonomy, p. 4.

[2] See the Wikipedia entry on nuclear powered aircraft, available at

[3] See Dougherty, The World’s Worst Weapons (Amber 2007, 2016) at B-36 H Nuclear powered Bomber, 220 – 221; see also the Wikipedia entry on the Convair X-6, nuclear powered jet aircraft, available at

[4] That’s the report, anyway. See n. 3.

[5] That is, for the Supersonic Low Altitude Missile, or SLAM. See n. 2.

[6] See Dougherty, The World’s Worst Weapons (Amber 2007, 2016) at Project Pluto, p. 228.  “Project Pluto was eventually cancelled, which may be just as well. It was not so much poor as horrific, and Pluto is one weapon the world is better off without. Id. at 228 – 229. See also the Wikipedia entry on Project Pluto, available at

[7] See the Wikipedia entry on Project Pluto, available at  If you want to see something more official, go to the Lawrence Livermore web site at .

[8] However, the NASA projects haven’t involved ram jets. See the Wikipedia entry on nuclear power in space,  at

[9] Do you want to see who’s on the Board? For a recent picture, see DoD, Defense Science Board, Home Page (2016), available at  Actually I don’t know why anyone would choose to publicly identify these folks. Once their names and likenesses were released, I’ll bet they became instant targets for sex workers, hackers and other spies, foreign and domestic. But hey, that’s only my opinion, and what do I know?

[10] See Autonomy at the transmittal letter, Fields, Memorandum for Under Secretary of Defense for Acquisition, Technology & Logistics, dated June 10, 2016.

[11] See n. 1.

[12] See Autonomy at p. 5.

[13] Id.

[14] See Autonomy at Autonomy for force application, p. 60 – 62

[15] Id. at 61.

[16] Id.

[17] Id.

[18] See Autonomy at Acronyms and Abbreviations, p. viii.

[19] See Autonomy at p. 61.

[20] Id.

[21] See Autonomy at p. 63.

[22] See Autonomy at p. 61.

[23] You don’t get the reference? There’s no reason why you should, really. Just check out Wikipedia at  Or, you could watch part of the movie at