10
$\begingroup$

What arguments could there be against using completely (or almost completely) robotic soldiers for wars set in the future?

In many war-themed futuristic novels, Earth (or whatever planet) is still using humans as the soldier. It seems to me that a robot soldier (albeit with advanced enough tech) would be profoundly more efficient at this. Although not quite there, we are closing in on a point in which robot soldiers could already replace humans (on earth, 2017). In fact, drones are essentially doing so already, although my question is in regards to ~100% replacement.

Is this a case of suspended disbelief, however nitpicky it is, or are there logical reasons?

$\endgroup$

closed as primarily opinion-based by Aify, L.Dutch, Mormacil, Hohmannfan, Vincent Jun 18 '17 at 17:10

Many good questions generate some degree of opinion based on expert experience, but answers to this question will tend to be almost entirely based on opinions, rather than facts, references, or specific expertise. If this question can be reworded to fit the rules in the help center, please edit the question.

  • 1
    $\begingroup$ Am I the only one who saw the Terminator movies ? :-) $\endgroup$ – StephenG Jun 18 '17 at 11:16
  • $\begingroup$ The robots have taken over and rather than sacrifice their own kind, they force the pathetic expendable humans to do their fighting for them! $\endgroup$ – colmde Jun 19 '17 at 10:23

13 Answers 13

8
$\begingroup$

I'll be the Devil's Advocate and answer the question as posted:

Is this a case of suspended disbelief (...)?

This usually is a case of suspended disbelief.

Let's not get so busy rationalizing Sci-Fi settings that we forget the facts of the matter.

A decently combat-ready and reliable human takes over a decade to create, is extremely expensive, has lots of properties that make control and maintenance difficult, and features horribly bad reaction times and targeting speed. A combat robot or drone can be mass-produced and specifically designed for the purpose of winning a war and nothing else. It can also be specifically adapted to the needs of the side that employs it.

The comparison here is between a side that mass-produces disposable killer machines and another that's spending years talking/brainwashing little boys or girls into becoming something similar.

This comparison is a joke. Even with "cutting costs" by sending in teens or children as soon as remotely credible, the mass-produced bots will be much cheaper and thus have a big advantage in numbers. This gets even worse if you shift to expensively trained special forces, which take multiple decades to train but can still die from a ridiculously cheap suicide quadcopter. Even a cheap robot's weapons aren't any less deadly than a human's! On the contrary, robot targeting is of the quality that can shoot down rockets in mid-flight. They can take all kinds of shapes and forms, specially adapted to a specific situation.

The arguments against the robot army aren't very strong. If the bots are strategically stupid, humans can still command them. Hacking goes both ways, but the one who built the robots is at a huge advantage, as they know the bots better and can design their security measures to favor their side in the digital war. Human psychology, on the other hand, is not as easy to skew in one's favor.

Humans are optimized to be generalists.

After a certain level of technological advancement, the fact that humans are not optimal soldiers becomes significant. The main argument against robots is that they aren't good enough yet, not that humans are somehow a good design for a front-line combat unit. A robot can have a lot of faults and still surpass the average human soldier.

Robots can be designed to emulate any favorable property of humans, but without designing humans as you would a robot, this isn't possible the other way around. Let's list some examples of downsides of humans that robots will exclude by design. Humans...

  • tend to choose self-preservation over optimizing kill-death ratio
  • can be slowed down by emotional struggles
  • often turn on their masters without any action by the enemy
  • must learn all rules for soldiers, even the simplest and most universal, through immensely expensive teaching processes
  • have a relatively small range of weights and sizes
  • require high-quality air to function
  • have a very specific operating temperature
  • can only be produced slowly, with the cooperation of females already present
  • have highly incompatible parts and are near-impossible to repair
  • take at least seconds to synchronize information, and struggle to do it over distances
  • need to sleep regularly
  • need lots of supplies even while inactive

This list could go on to fill a book.

So why and when would humans be preferable?

From the above, we can establish that humans are not preferable for a classical war, with tasks such as clearing buildings of enemy units, and that the main reasons for this are needless generality of humans and economy of scale in machine production.

So, if you want humans to surpass advanced robots, you need a situation that is rare or new, so that a cheap robot solution is not available. Specialization comes with the downside of being useless if the specialization does not apply.

Think of the US's struggle to eliminate the IS, even with all their target-tracking cruise missiles, flying killer drones, and enough nuclear weapons to level the entire battlefield. Their arsenal is designed to never lose a conventional confrontation, and it is very capable in that regard. But it is not at all optimized to attack against human shields while minimizing bad press, which is currently the goal.

As another idea: sometimes, efficiency just isn't a goal in the first place. Today, many weapons are built to maintain a flow of military spending, with overall efficiency rarely being a primary concern. In other words, you don't need humans to be good soldiers to have human soldiers. You just need any reason that human soldiers are preferred to robots, which can be purely political.

$\endgroup$
  • 1
    $\begingroup$ I upvote; similar to my answer; I suspect we were of the same mind and writing simultaneously. Your's has other useful details. I think AI and robotics will, in the next decades, surpass any human ability to navigate a battlefield (or enemy terrain) quickly and quietly. Robots have the advantage of highly specialized hardware for the task, with the potential of far greater strength. A robot can plausibly jump off a ten story cliff and land hard but safe with steel shock absorbers; or trudge ten miles underwater and emerge muddy but fully operational. Humans will be inferior. $\endgroup$ – Amadeus Jun 18 '17 at 13:47
  • 1
    $\begingroup$ Split the difference, adn go with human controlled robots. We already have drones capable of combat when piloted by remote humans, as soon as the actual robotics of bipedal (or however -pedal it ends up) movement that can function appropriately in adverse terrain and combat conditions, the robotic soldier will happen. As to 10 years to create a decent combat capable soldier? The USMC seems to do fine in a fraction of that time, and even the various special forces groups around the world do it in less than half that... $\endgroup$ – ivanivan Jun 18 '17 at 15:35
  • 1
    $\begingroup$ I think this effectively answers the question. Unfortunately, the answer is that there really isn't one, but that's ok. The only thing I've personally thought of as an argument is perhaps keeping the emotion in war is good. Maybe, very ironically, having humans die in combat is actually a survival trait, keeping us from going too far. Expendable robots could lead to much more useless wars. I doubt this, but it's a thought. $\endgroup$ – anson Jun 18 '17 at 22:37
13
$\begingroup$

In the short term, robot soldiers will be inhibited by their programming. Since "hacking" robot soldiers will be a grave danger, robots will have to be programmed to follow certain drills or battle procedures in particular situations. This makes them relatively inflexible compared to human soldiers, and predictable enough that you could set up "kill zones" by creating situations which will cause robot soldiers to react in predictable patterns.

In the longer term, the idea of sentient, self programming robots will be either frightening or abhorrent to the majority of people (both soldiers and civilians), especially since everyone will be thinking of the "Frankenstein" complex, where the devices turn on their master (this was actually the theme of the play which brought the word "Robot" to the English language as well, but few people have ever seen R.U.R).

The inevitable will happen when either a rogue regime (such as DPRK or Iran), or violent non-state actors (such as LTTE or ISIS) decide they can get what they want by unleashing sentient robots on civilian and military targets, and everyone needs to use sentient robots to fight them (despite what is shown in movies, human soldiers will be at a vast disadvantage compared to a robot, thinking and moving hundreds to thousands of times more slowly, having less precise control of the weapons and being limited in the amount of firepower they can carry compared to an equal sized robot; even a tank has considerable room devoted to the crew which a robot tank can fill with fuel and ammunition).

$\endgroup$
  • $\begingroup$ Thanks for the response. The last point seems as though the reason against it is so the enemy also wont have it. That seems optimistic but flawed to me. $\endgroup$ – anson Jun 18 '17 at 1:11
  • 4
    $\begingroup$ Nice to see a reference to Karel Capek's play. Since the play was written in his native Czech it added the word "robota" to the language. This was translated into English as 'robot". The original word meant a worker or drudge. Capek's robots were organic artefacts constructed as faux humans; not robots as mechanical contrivances as the word is commonly applied. I'm sure you are aware of this, but this is me being thoroughly pedantic. $\endgroup$ – a4android Jun 18 '17 at 3:18
  • $\begingroup$ The idea of killer robots is already abhorrent. There are currently moves by roboticists and AI researchers to ban or restrict the use of autonomous weapons. Nations using drone weapons are perceived by their target actors, state or non-state, as cowardly and despicable. Unwilling to commit their own living soldiers to combat. This may change when all sides in a conflict are using drones. Plus one from me. $\endgroup$ – a4android Jun 18 '17 at 3:22
  • 1
    $\begingroup$ @a4android I get that other nations might view it as cowardly if they are using real lives, but I still can't take that as an argument against it. It's purely a rational decision to benefit your own. That is unless the reaction is so bad that it turns almost every other nation against you. I should have been more detailed in the question, but I'm trying to think in broader terms, rather than relative to our society. $\endgroup$ – anson Jun 18 '17 at 3:47
  • 1
    $\begingroup$ I personally find it highly cowardly that the president/king (or whatever leader it is that initiates a war) no longer rides in the front lines to lead the army, as it was done in the good old days... $\endgroup$ – Mrkvička Jun 18 '17 at 6:05
8
$\begingroup$

If you think of war as being like chess, war with robot soldiers is more like go. Instead of battling to control territory directly, battle will consist of trying to take control of the other side's robots. Then when you steal the other side's robots, you will control the territory that they did.

In Terminator 2 and later, they had terminators that John Connor had reprogrammed. They'd yank the power and reprogram them one by one. Presumably robot armies would have better ways of doing that. Perhaps they'd just flash over the programming. Or they'd pull the chip and replace.

Even ignoring hackers, consider what happens when the enemy captures your command center. Now they have physical access to the controls. Or destroy the command center so that the robots can't be updated. Multiple command centers to avoid that? Destroy or capture all of them. The more of them there are, the easier it is to capture one (and turn the robots against their owners). But the harder it is to destroy them.

Or consider covert action. A spy gets access to a command center and modifies the robots' programming or uploads new software that turns them over to the enemy.

Perhaps someone will eventually decide to work past these problems and try it. But there are reasons why militaries might prefer humans to robots as soldiers.

$\endgroup$
  • 1
    $\begingroup$ The idea of turning robot soldiers against their own side is a chilling reminder of how conflict can go wrong. Plus one from me. $\endgroup$ – a4android Jun 18 '17 at 3:24
  • $\begingroup$ Yeah, normally if someone captures a command center with regular soldiers, you can at least trust them to still know what side they're on. This really seems like one of the bigger drawbacks of robo-soldiers. $\endgroup$ – Friendlysociopath Jun 18 '17 at 3:34
  • 1
    $\begingroup$ I definitely agree that the ability to control a mass groups programming would be the one of the largest drawbacks. Although I don't consider this all that much different than an enemy learning our coded language or taking control of our radio frequencies, if we were to relate this to modern (human) war. I understand it could be more detrimental with robots and computers, but imo it doesn't outweigh the advantages. Technically a command center could self destruct if it's about to be sabotaged. $\endgroup$ – anson Jun 18 '17 at 3:58
  • 1
    $\begingroup$ Assuming that there would be one central command center for a group of robots that extremely developed sounds far-fetched. While this is an old trope, storming the command center of the evil robots, in reality, I imagine it would be a lot more de-central, with the ability of throwing single, compromised command nodes&codes out of the net quickly and having others take over their functions (some sort of chain of command). So, even if a minor node/code was captured, the next higher authority would revoke that access easily and it would be worthless. $\endgroup$ – Florian Schaetz Jun 19 '17 at 7:43
7
$\begingroup$

Moravec paradox

Moravec's paradox is the discovery by artificial intelligence and robotics researchers that, contrary to traditional assumptions, high-level reasoning requires very little computation, but low-level sensorimotor skills require enormous computational resources. The principle was articulated by Hans Moravec, Rodney Brooks, Marvin Minsky and others in the 1980s. As Moravec writes, "it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility."

enter image description here

Perception includes pattern recognition, including behaviour patterns.

AI with proper sensor suite can easily spot lone human but unless human carries IFF transponder, AI will be unable to identify him/her quickly. AI will have no way of knowing if detected human is an ally, enemy or civilian, before human opens fire. In situation where hostiles and frinedlies are easily detectable, AIs will outperform humans in reaction time and tactical efficiency by orders of magnitude, but as soon as ambiguity enters the fray, AIs lose the edge. You can program your autonomous tank to shoot any IR signature large enough, but apart from enemy tanks it will shoot: herd of deer which comes by, car with fleeing family of 5 inside, top secret allied operative who decided that need-to-know basis means AI doesn't need to know, and red cross convoy. Last of those will spark massive international incident, turning your allies away, or perhaps turning them into enemies if you were careless few times too many. Meanwhile, enemy will either use tank engines in short bursts, use active IR camouflage, or equip tanks with hybrid diesel/electric drive system (first hybrid tanks were designed during WWII) to operate below AI kill-on-sight-threshold and sneak past.

As soon as camouflage, deception, or unclear intentions come into play, advantages of AI disappear. You need humans to designate targets, and you needs AIs to shoot the targets. Human will never outperform AI in shooting, and it will be centuries before AI outperforms human in target designation.

$\endgroup$
6
$\begingroup$

I'd say it depends on how far into the future you're looking, what kind of conflict, and how concerned with realism you are.

For today, and the near future, robots are really terrible at a lot of things: sensing, locomotion, and longevity.

Sensing: Modern computing is just barely to the point where it can distinguish between a cat and dog. Almost. Asking a computer to find a human, determine if it's friendly, and whether or not it's a threat, in a reliable manner (before the human starts shooting) is not something that's going to happen for another few decades. (I could see having declared kill zones, ie the death strip between the berlin wall; but you wouldn't have robots going door to door in Aleppo).

Beyond that there's also the very basic issue of how robots get and process information about their environment- with vision systems, laser scanning, and/or depth cameras they can probably get a decent 3D model of the local environment, but actually understanding that model is an entirely different issue- knowing that the ground ahead is wet mud and should be avoided is difficult without being pre-programmed; learning that it's a problem (unsupervised, in the field) is in the far off dreams of most modern AI researchers. There are other issues here too, the only one springing to mind being what I'd call scene decomposition- getting a robot to take a 3D model and extract "what can be manipulated/moved" is also difficult- realizing that a 2x4 in front of a door can be picked up and moved aside (if not nailed in) is another difficult task for modern AI.

And other minor things: Robots are bad at sensing themselves (called Proprioception). They can only sense what they have sensors for. So, unless adding on a thermal camera, they can't feel that a door is hot; or that someone stuck a radio transmitter to their back; or that a piece of rebar got stuck in their leg. They may be able to sense other effects (accelerometer in leg registered jolt, current to left leg servo at 200% and servo encoder not registering movement; it's probably stuck); but being able to do anything about it goes back to the issue of decision making and sensing, from above.

Locomotion- It's really shocking how bad robots are at moving from point A to point B. There is some work on it (Boston Dynamics has done some wonderful and horrifying work), but just coming up with something that is able to fluidly move in and around rubble, or climb a ledge, is very difficult. Most modern robots are wheeled/tracked because that's the only thing available that has a run time of more then 20 minutes, but we can ignore them because they're easily defeated with stairs. That partially ignored, we could have multi-pedal robots always within a few hundred feet of an energy source, and then run into the problems of robots trying to operate on uneven footing. And they already have trouble on even footing.

Part of this is the sensing issue above- Seeing a coke can and knowing you can step on it and it will flatten, vs a can shaped rock or a brass cartridge casing is hard. Also, knowing when an obstacle in the path is a piece of loose wire to be ignored vs a piece of rebar that's going to spear your delicate servos vs a piece of steel wire that going to tangle around you if you walk though, is difficult.

Longevity- Basically the issue of maintenance, and how long a robot can stay in the field. Robots get sand and water and blood and grime and salt water in their every moving part. Then the parts stop moving, or a short circuit happens and suddenly the left half of the robot stopped working. For the near and probably far term, robots aren't able to fix themselves. Certain problems could be resolved fairly simply- systems can be put in place to allow whole sale swapping of body parts- robot realizes his arm is malfunctioning, goes to a maintenance van and picks a new one up off the shelf... and tries to attach it. (this is... not easy, but I'm willing to say it'd be doable if there was a drive for it). Or. They're robots. Just send another 20 in. Leave the malfunctioning where they lie. (unless you don't want them getting into enemy hands). (It's also not a good long term strategy; as robots are kind of expensive)

In addition to that, robots need a power source. Batteries are something of an awful power source. Great for short hops away from a generator, but terrible if you need to be in the middle of nowhere for a week. (kind of an energy/weight ratio issue, and just efficiency of movement). (semi example- Boston Dynamics Big Dog robot uses a 15HP go kart engine- 11KW equivalent (several households)). This is also kind of resolvable by just always having an APC with a generator near by, but that can be limiting, and will usually require a supply line.

There is another thing that may have been hit upon by someone else: Humans are wascally widdle wabbits. Most of the above is on the premise of trying to replace humans operating in an environment where humans are fighting humans. Once humans are fighting robots their tactics will change; they will do everything they can to exploit any weakness they find. Identifying and counteracting these weaknesses, then issuing a patch/firmware-update is a slow and time consuming process; where as most humans can spread the weakness with word of mouth.

In short- there are a lot of reasons we don't use robots right now (except in a remote control capacity), and most of them are probably still going to be problems for the next 20-30 years (and I think longer; but I can't back that up). It basically all comes down to robots are idiots- they work well in really well and narrowly defined roles- drones work because all we're doing is asking that they not hit the ground, once we go past that they fail in spectacular ways. And failing in spectacular ways is not generally what you want an army to do.

(now again, this all depends on your limits of realism, timeline, and conflict type; if you're fine with robots killing every not robot that they see, then half of these concerns are out the window)

$\endgroup$
  • $\begingroup$ +1.. But the Petman or whatever it's called was built like 10 years or so ago. It could walk fairly well then. And I would imagine any robot infantry men would be programmed to recognise a uniform or symbol with is a lot easier than distinguishing a cat from a dog in programming terms. As for power, Tesla invented wireless power ages ago and others have rediscovered it since. I just feel like you are underestimating current technology. $\endgroup$ – I wrestled a bear once. Jun 18 '17 at 15:52
  • $\begingroup$ Great answer. I think you hit on the major flaw in my question in that I didn't specify a timeline. I'm glad I didn't however, because its led to some good answers. $\endgroup$ – anson Jun 18 '17 at 22:51
6
$\begingroup$

Is this a case of suspended disbelief, however nitpicky it is, or are there logical reasons?

This is a case of suspended disbelief for fiction; because audiences are not moved, emotionally, by the destruction of hardware. They will get bored if both sides use hardware and the result is just a circus of mechanical destruction.

If we use robots and the enemy is living beings, we look like exterminators at best if they are insectoid or non-human enough, and like the bad guys if the enemy has any human characteristics, emotions or demonstrates feeling pain or grief. The only way to make the audience identify with our side is to put human lives on the line.

Star Wars cannot be about machines going after each other, we need human or human-like pain, grief, suffering and despair, and it must be severe: Luke loses his adoptive parents. Young soldiers with lovers back home, hopeful to start their lives together after the war, are trapped and slaughtered. As Stephen King advises writers: Develop a character the readers will love and care for, then put her in the cooker.

In futuristic fiction, in order for us to identify with robots, they would have to be, effectively, human beings with an emotional side able to experience setbacks and frustration and some kind of suffering, perhaps vindictiveness and anger as well.

Of course that is not where the question lies: IRL, it is fair to presume that within 50 years or so, robots will be able to do anything a human can do in warfare, better. The "steal command" problem is a non-problem, we can use the same protocols for robots that keep our banking system humming along transferring trillions of dollars around the world without anybody diverting all the money into their own accounts. Long key RSA encryption is easy, fast, and impossible to break by any known means; that is what banks use.

The robots in question may be about as self-aware as a self-driving car (which requires a form of self-awareness to navigate its body safely through a fast moving maze, which also requires a self-referential model of what movements it is capable of choosing and how other vehicles will most likely move).

The robots do not have to be emotional in any way, they can be rational and intelligent (in the sense of accurately predicting outcomes in novel scenarios). Emotions, like fear of death or reluctance to inflict harm or psychological angst over what they have done, or grief over lost compatriots or anything else are all baggage they don't need. Intelligent machines do what they are told; including if they are told to stop, self-destruct, or conduct a suicide mission that will destroy them: They can be "intelligent" with absolutely zero emotions or "desire to live" that would interfere with or override their commands.

As for whether the enemy uses live soldiers: The story of the war is written by the victors, remember? Just as we do now in the USA, we use robots and high tech and drones, but we don't spin that as cowardice: We spin it as saving the lives of our soldiers, protecting the troops, while destroying the bad guys. We say our soldiers are still brave, intrepid warriors, even if they sit all day in a comfy chair a desk safely tucked away in a stateside office building on some Army base in South Dakota, using a 4-screen video-game interface to fly drones over Pakistan.

Robots will be just as autonomous and capable of making decisions as humans; should communication be disrupted. They will be just as capable as humans if cut off from command for any reason; but more capable than humans of communications between each other (e.g. they can pass full video of what they see and hear, they don't have to 'describe it'; they can have a group mind).

They will also be able to survive better than humans. They don't have to rest, they can use solar power to recharge; if necessary they can hide and shut down all but the most minimal solar powered sensory activity, and survive indefinitely without food, waste, or boredom (boredom is an emotion).

This is not something current AI can do; but the outlines of what it will be able to do in the next 30 to 50 years are clear. Just as Moore's Law was clear and has held, with only slight modifications, for 50 years.

Today's robotics are like early guns; used in warfare and deadly, but not terribly accurate or reliable. That will change, they will get better, and no politics or shame will stop them, just as our modern automatic guns feel light years ahead of the blunderbuss which was basically a mini cannon, future robots will be far better and more lethal than human soldiers, and the citizens of the countries that own such robots will be appalled at the idea of losing their human soldiers in battle. They won't be hacked, even by up-close surgery. John Conner will not have the 10,000 bit encryption code necessary, and no screwdriver is going to be able to modify the circuitry of integrated circuits: They will have one-way tamper-proof casings anyway, so once they are sealed any attempt to open them destroys them (there is no need to ever get inside and repair them, even now we just replace malfunctioning chips). (Heck our robot can automatically self-destruct its processors if anything every penetrates its casing).

In fiction the audience wants to identify with entities in the story that display human emotions and qualities. In real life, and real war, the objective is to subjugate the enemy and force a surrender under threat of death to the humans on the other side. IRL we'd rather not identify with the soldiers we lost, we don't want to share the grief of their families, parents and children and spouses, we don't want to feel the loss of their potential as good people, friends, and citizens. We don't want them to suffer from PTSD or the trauma of war. If we can pay our way out of all those emotions, by dint of buying robots to do the dirty work, we will.

Further, in my opinion, we (and our military leaders) will prefer our robots to be distinctly non-humanoid (or even animal) in form and behavior, so as not to accidentally invoke any empathy or sympathy for the robots. A drone doesn't look like a person, a drone getting shot down or crashing doesn't look painful, it looks like an unfortunate loss of hardware, like accidentally dropping a big screen TV down a flight of stairs. A kind of "dammit" for the loss of value, but zero concern for any pain the TV felt; it didn't feel anything. If the animal form with legs and limbs is truly useful for navigating terrain; I suspect we will use insect or arachnid forms with many legs. Losing a robotic soldier reminiscent of a tank-like 3-foot cockroach or spider won't bother us much at all. Sure, it is our cockroach-tank, but if it gets blown to pieces by a missile, no problem. Send two this time.

$\endgroup$
  • $\begingroup$ "Long key RSA encryption is easy, fast, and impossible to break by any known means; that is what banks use" This is a massive oversimplification. There is far more to building secure software than just encryption, and exploitable vulnerabilities can, and do, exhibit themselves in the networking code, which executes before the crypto does. $\endgroup$ – atk Jun 18 '17 at 14:57
  • 1
    $\begingroup$ @atk I won't argue; my colleagues work on it all the time. But don't forget we will have AI to help us on that front, too. My point is not that security will be easy, but we are already solving the problem with sufficient attention and I don't think it will be an issue. Just like we have enough security on our nuclear weapons that they cannot be exploded in their silos. Certainly, in a fictional universe, it is far more plausible that the security issues are solved than to assume we (and our AI helpers) are so incompetent that we build robots that can be so easily turned against us. $\endgroup$ – Amadeus Jun 18 '17 at 15:21
  • $\begingroup$ This is a great answer and helps confirm my suspicion that there probably is no deal-breaking argument against robots. One of the only things that comes to mind when reading other comments is that it's so damn hard to get right. But maybe I'm being ignorant. I'm a FE developer and not an AI developer for a reason. $\endgroup$ – anson Jun 18 '17 at 22:41
  • 1
    $\begingroup$ @anson I have an older, longish discussion of AI issues that might help to clarify some things, here: worldbuilding.stackexchange.com/a/81020/37679 $\endgroup$ – Amadeus Jun 19 '17 at 11:21
3
$\begingroup$

To me, all the above answers aside, the problem is accountability. We already see civilian casualties being written off as collateral damage of missile or drone strikes. Now imagine the same situation on the ground level.

With humans, at least, we hope, that some will refuse to attack civilians, ideally out of, well, humanity, but also out of fear of getting punished, eventually. With robots, they will simply obey the order to kill, without reasoning why. Even if there are survivors, they can't positively identify the individual who gave the orders, as they would not be present at the scene.

You will see increasing numbers or civilian casualties as reckless officers act with impunity and blame the other side, who just happen to be using a similar model of robot, only distinguishable by the internal IFF transponder. Eventually, enough people will get sick of it and declare them unlawful weapons like nerve gas, cluster bombs, etc., with the very use being considered a war crime. Even then, there will be countries or other actors who disregard the rules and carry on using them.

$\endgroup$
  • $\begingroup$ But of course the order had to be digital and recorded and authenticated by the robot, so the person that gave the order is completely known to the robot owning nation, and can be prosecuted if they feel it is necessary. Sometimes they do feel that some officer over-stepped their bounds. This is no different than a long range cannon or aircraft attack; the people on the ground don't know who gave the order to kill hundreds of civilians, but the people owning the gun (or the plane) certainly know exactly which person made the fatal decision, and can prosecute them if it is warranted. $\endgroup$ – Amadeus Jun 18 '17 at 16:58
  • 1
    $\begingroup$ @Amadeus: Just what proportion of air/drone-controlled strikes that kill civilians or, for instance, MSF hospitals, have been successfully prosecuted? Nowadays, war crimes are only really effectively prosecuted by third parties like the ICC, rather than the government of the offending military. These bodies will probably not get access to that kind of evidence. $\endgroup$ – nzaman Jun 18 '17 at 17:14
2
$\begingroup$

Everyone has an EMP

If every nation has a way to disable robot armies, there will be no robot armies.

Sort of similar to how if everyone has the bomb, nobody will use the bomb.

$\endgroup$
2
$\begingroup$

Human infantry still have a role in the army of the future because robots aren't always very useful against terrorists or guerrillas.

When the enemy are able to blend in with the civilian population, it's hard for human soldiers to tell them apart, but it's worse if your soldiers don't even understand human culture at all. Sending in cold unfeeling robots to massacre innocents is not exactly how you win hearts and minds.

Also, a robot needs to carry an enormous amount of energy to walk around all day, more than enough to set itself on fire. So it's liable to burst into flames like a huge Galaxy Note 7 if shrapnel happens to short-circuit its battery. This would be hazardous to civilian safety in urban warfare, and doubly so after enemy guerrillas figure out the most efficient way to set off your walking bombs.

$\endgroup$
1
$\begingroup$

Human beings are crafty and inventive. We are better suited for improvisation on the battlefield, whether it be constructing traps, planning ambushes, interrogating P.O.W.s or conducting reconnaissance. Robots would be capable of doing these things to a degree but if they require orders from human superiors (which they should to avoid a bunch of awol death machines acting as unknown variables) they could not take the necessary initiative to be effective soldiers outside of front line combat.

As far as front line combat units, robots are best suited to supporting roles such as field medics. For purely destructive or defensive purposes, there already exist more specialized tools that do not require the cost of building and maintaining complex robots. (Take this as an example: When I first watched the movie Pacific Rim, the thing that took me out of it was the idea that, since these giant mechs cost so much and needed specialized pilots and the development of a new field of science to operate. They could have placed mines and energy turrets to destroy the monsters. These could be manufactured in greater number and require no trained pilots.)

Ask yourself, what task do these robots perform in your world that could not be done more efficiently by existing technology. The best reason I have ever read for robot troops was to be used as vanguard units deployed on planets with unlivable conditions.

$\endgroup$
  • 2
    $\begingroup$ Thanks for the response. I don't agree that we are better suited for the battlefield though. Just as self-driving cars are objectively safer for transportation, robot warriors should be objectively more efficient. It might be a stretch to comprehend better robotic interrogators (imo, not out of the question) but I'd leave interrogation out of this conversation for now. $\endgroup$ – anson Jun 18 '17 at 5:47
  • $\begingroup$ @PatrickTrentin very true, its much more complex. My comparison is not between self-driving cars and robot soldiers, but of the objective efficiency gains you would get when allowing a (well programmed) computer to do the work. You're right though, it's obviously a harder task. $\endgroup$ – anson Jun 18 '17 at 22:27
1
$\begingroup$

Robots are expensive. Humans are cheap and self-replicating.

As mentioned in other answers, robots can be hacked and reprogrammed. But the interesting part is humans stick to their programming when the going gets tough, and remain loyal to "bad" leaders even to the extreme of genocide denial.

If you withhold sensitive secrets from an AI you will eventually get errors in your strategy simulations, or worse the AI could expose some mathematical or accounting cover-up that will need to be corrected holographically throughout the information system. It's going to be difficult to keep these political "corrections" unknown by programmers and technicians. If the robots have any ability to form their own math-based strategies and analyze their own success rate, they will eventually discover these manipulations or make lethal errors in decisions. Consider HAL-9000 who decided to kill the crew and scientists to protect the classified purpose of the Jupiter mission. Notice the humans in the same film did their jobs and didn't ask questions.

With humans you get blind loyalty and suicide missions. They will continue to defend a lost cause, and their martyrdom can be used as recruitment propaganda – not all humans obviously, but the military has ways of weeding out independent thinkers by self-selecting for conformity, compliance, and obedience.

You probably wouldn't do this if you could actually win with a robot army, so I'd suggest this is an underdog strategy. Having an all human army defending against an army of robots and drones also works as a human shield strategy. Global sympathy will go to humans in a propaganda war of humans vs bots.

Our military future however will be effected by what the military calls Tooth-to-Tail Ratio which refers to a logistics issue involving the resources required to supply and support (tail) each combat soldier (tooth). The actual ratio varies over time and by mission, but it's typical to spend double the resources on the tail than the teeth. DARPA has a stated goal of developing technology to reduce the tail, so expect AI and drones to become more valuable in supplying combat soldiers than AI becoming combat soldiers. It sounds counter-intuitive, but that is where the finances should lead us.

In a not-so-far future scenario an AI will probably design global defense strategy based on simulations, and AI may become more reliable and cheaper than generals. We'll have the awkward scenario of human infantry commanded by machine intelligence.

$\endgroup$
0
$\begingroup$

It's 2017. DARPA already has already produced a whole slew of different humanoid and nonhumanoid robots. They started producing them some years ago hoping to build one to aid in cleaning up after the Fukushima disaster. Youtube it, seriously, very cool stuff. They needed robots that could do things humans could do like jump and run and turn a wheel or unlock a door. We are already living in the future you speak of.

The conditions of Fukushima are complex compared to those of a battlefield. The Fukushima problem requires specialized scientific training. The average soldier only needs to learn the protocols of war. Any kid from the burbs can become a soldier but only the educated higher class could help in a situation like Fukushima.

Point is, we already have the tech to build Terminator style robots but we continue using humans because humans are abundant. Many of the esoteric groups like the Rosecrucians and the Masons and Bankers like the Rockefellers and the Morgans and the Rothschilds and even the most powerful politicians are eugenicists. They believe that the world is over populated and that in order to improve humanity, the population must he decreased and maintained at a certain level (Google the Georgia Guidestones), and that the genetically inferior should be the first to go. No soldier with a PH.D. is ever asked to work on the front lines.

Many modern wars aren't meant to be won. They are meant to be sustained. They are meant to thin out the population and generate profits for those that finance these wars. The US sold weapons to the Nazis at the beginning of WW2 and other groups funded both sides of the war. The rich elite around the world know that the only way to guarantee they will win is to invest in both parties.

The most efficient robots are those with a specific problem to solve. Think drones, not humanoid infantrymen. Truth is, drone robots are used frequently in war and always will be, but as long as hunger and poverty and overpopulation exist, lower and middle class humans will always be on the front lines.

$\endgroup$
0
$\begingroup$

The truth is in reality we could have an all-robot war tomorrow - ICBMs and cruise missiles fit almost every definition of a robot.

Of course, although in such a war it would be robots carrying all the weapons, huge numbers of humans would die as well as the robots being destroyed.

In the book "Arms and Influence", Nobel laureate Thomas Schelling advances the view that there are two things a military can do to accomplish a country's goals in a conflict:

  • Direct force, where the military directly expels, seizes, destroys, confines or exterminates. For example, if you want me to stop launching missiles at your civilian population, you can use precision bombing of my missile factories and launch sites.
  • Coercion, where the military creates a cost in pain and destruction, offering to stop if some other demand is met. For example, if you want me to stop launching missiles at your civilian population, you can say you'll launch two missiles at my civilian population for every one I launch at yours.

Needless to say, coercion is a lot less heroic, and politicians are much less likely to advocate for it in clear, open language. But the truth is in many situations it's easier, cheaper and more effective. I can make direct force difficult by having my missile factories hidden and fortified - or maybe instead of launching missiles I'm doing something harder to police, like funding rebel groups within your country.

Given humans' natural appetite for revenge, some would argue that every war that ends without annihilation or permanent occupation has had an element of coercion.

Anyway, here's my point: Coercion requires pain or the threat thereof, and a drone-vs-drone battle doesn't have any pain (except the opportunity cost of that-drone-cost-money-that-could-have-been-spent-elsewhere) - so both sides will be looking to get around the other side's drones, so they can land attacks that actually hurt the other guy.

So even if both sides have 100% drone militaries, it's drone-vs-civilian strikes that are actually going to bring the war to a conclusion.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.