It is not universally agreed that it is a fundamental right to have guns. Most of the civilized countries of the world agree. One thinks it is a fundamental right. Strangely, that right comes from the Constitution which is just something some guys wrote down one day. They were wrong about all kinds of things in that document and the fundamental right to have guns might be one of them.

I’m fairly convinced that the “if-we-don’t-get-it-first-then-someone-else-will” factor makes “armed robots” an inevitability. This boils down to a simple principle of Darwinian natural selection, so there’s really very little humor in saying, “Get over it: it’s life.” Lofty ethics aside, “should we or shouldn’t we” aside, let’s just consider it granted and move from there.

This idea of “armed robots” argues strongly in favor of man/machine interface. The simplest way to insure that our AI doesn’t get too significant an advanatage over us- barring the implausible “don’t develop it” suggestion- is to make use of the very characteristic by which it will outdistance us. It’s not exactly as if research into man/machine interface isn’t ongoing; it’s just that it isn’t brought up often enough in the same breath as the doomsday scenarios- perhaps because it is somehow innately unpalatable to the public, or perhaps because the right “spin” hasn’t been found to make it so.

Every dangerous idea conceived and permitted by the day’s technology has become a reality. Alongside these dangerous realities have come (sometimes) effective countermeasures. What I’m driving at is that there’s already enough stigma attached to AI. The Luddites have their own channels of communication- and those are very broad channels, with much bandwidth.

I’d love for the discussion to open up and include talk of what we’re doing to “combat armageddon.”

As alluded to in the article, the question of giving a robot a weapon has a lot of ethical basis behind it. The issue however, is not with the act of giving control of a weapon over to a machine, but rather with warfare in general. The ethics of war have been discussed thoroughly and there are valid points on both sides, but the ultimate conclusion is that war is between people. Never has war been between two sharp sticks, two rifles, or even two super-advanced aircraft.

The objective of warfare with robotic tools is the same as warfare in the middle ages: diplomacy has failed and violent force has been deemed necessary to achieve the objective that diplomacy could not. Arguing whether robots with weapons makes going to war easier ultimately reveals more about the leaders we select rather than the technology we utilize.

We already have robots that kill people – they just don’t look like movie robots.

Guided missiles with ‘intelligent target detection’, drones that do all the work of figuring out how to drop a bomb on something once they have been given permission and a target. Modern tanks and aircraft are basically killer robots with a human as a permission giving system – more and more the human won’t even be on board. One human can already ‘fly’ more than one at a time now. What ratio of robot to human control truly counts as ‘human control’? 2-1? 10-1? 100-1? 10000-1?

The thing we have to decide regarding robots with guns is if we think it is okay to kill/maim/electro-stun 6 year olds with them. If we think it is wrong then the baddies will use six year olds to spray paint sensors. If we think it is okay………..

I remember reading an Asimov short story about hmmm smart missiles or something along the line, and how humans had forgotten how to do math. And then one discovered how to do math and it was miraculous, and then someone in power was so thrilled by this that he said he could replace the smart missiles with human piloted missiles, because it would be so much cheaper. (I think they were the same story >_>)

And he has a point, I think… an army of robots just doesn’t seem realistic, at least for a long time, because the people who are actually in charge don’t really seem to care about human lives, and human lives are so much cheaper than robots…

Well, let me start by saying that I am definitely for the arming of robots, but only under certain conditions. Let me explain…

It seems the most relevant argument against putting weapons on robots is that of potential innocent casualties, as so eloquently pointed out by Kenneth : ]. (Other arguments include robots turning against us and taking over the world, and the ease with which people will enter into a conflict given non-human “soldiers.” As someone working towards a PhD in robotics, I feel at least somewhat qualified to say that the former is utterly ridiculous, spread by people lacking a basic understanding of computing, AI, programming, etc… Evan has done a good job of discussing this in the past so I won’t re-hash it here. See my comment below on why I don’t really buy the later argument***).

So, given the risk of innocent casualties, why would I want to arm robots, especially given the difficulties robots and computers have with so many of the basic skills we humans take for granted? Consider the JDAM, a so-called smart bomb. These bombs have a reliability rating of about 95%. That means that 5 out of every 100 dropped will fall somewhere outside the roughly 10 meter target area. Once given a target and released, they are completely autonomous, using inertial guidance and GPS to navigate to their target. They are quite obviously designed to kill/destroy, and they miss on average 5 times out of a hundred with potentially (and frequently demonstrated) lethal consequences for innocent people. Why isn’t there public outcry over the use of the JDAM equipped bombs? Well, consider the alternatives.

When robots or autonomous systems can perform better than humans and the systems currently being employed, they will be an advantage to those in battle. They don’t have to perform perfectly 100% of the time, just better than what we are currently capable of (and remember that us humans are far from perfect as well). So it’s not whether or not a robot can properly distinguish between an insurgent and an innocent 6 year old 100% of the time, but whether it can accurately identify and engage threats AND NOT engage non-threats better than a human can under the given conditions. Please don’t think I’m saying a robot that mistakenly kills an innocent occasionally is OK and doesn’t need improvement. Far from it. But if, for example, a robotic sentry paroling a green-zone border at 3 am can identify 95% of threats with a false positive rate of 1%, and a human can only identify 80% of threats with a 5% false positive rate (stats are hypothetical), why wouldn’t you switch to the robotic system, or at the very least, supplement your human forces with it? We’re not there yet, but eventually, I think we could be and when we are, and under the right circumstances, armed robots or autonomous systems would be an advantage.

But what about malfunctions? Armed robots are tools, and just like cars, airplanes, and every medical device used in a hospital, they will be required to conform to strict standards and safety protocols before being allowed into the wild so to speak. Just as when the first automobiles and airplanes were invented, people weren’t trying to make them autonomous. But as technology grows and evolves, the reliability of components is gradually tested and eventually trusted, and new functionality is added little by little, tested, refined, trusted, etc… We’re not going to put a Barrett .50 cal on a web-cam equipped Pioneer robot with a hobby servo duct-taped to the trigger and expect all our war-fighting problems will be solved. We test components, and gradually add functionality until we have a complete system that we can also have reasonably high trust in. Will accidents happen? Most likely, yes. And we can/will learn from them to make the systems even better. But I don’t think the potential for machines making “mistakes” is a good enough reason to prevent them from potentially being able to improve upon what we are currently capable of. What if the same argument were used to ban passenger aircraft? Or cars? Armed robots certainly carry a lot of risk, but with the higher risk will come more caution, testing, proof of correctness, fault-tolerance, etc…

Baker

*** When it comes to the argument that says arming robots will make it easier to engage in violent conflict because the loss of friendly human life is less likely, I say first and foremost: Cite your sources! I don’t think we are close to having the technology available that could prove or disprove this theory. It is an argument born solely out of logic, and that logic only makes sense if you ignore the complexities that go into conflict. When it comes to conflict on a global or international scale, the loss of your own troops is only a fraction of the considerations that a leader or a country must take into account before deciding to enter the conflict; political fall-out, total cost of not entering conflict, economic considerations, international perception/costs, resources available, resources at stake, safety of citizens, safety of innocent lives in target areas, projected time to resolution, ability to sustain conflict, cost of rebuilding after end of conflict, ability to stabilize area after conflict, potential alternatives, etc… etc… etc…

Even in smaller conflicts, such at police forces using armed robots to break up hostage situations, the equation isn’t as simple as friendly/innocent casualties. Saying we shouldn’t arm robots in such situations because it will make it easier to go in with guns blazing is no more valid than saying we shouldn’t arm police with guns for the same reason.

Each situation/conflict will be different and will have its own set of complexities that must be accounted for before making the decision to enter the conflict. As Jake said, this has more to do with the people than the technology.

I kinda doubt anyone’s still reading this, but if you are, thanks for listening!

Excellent talking points. The irony of all of this is as much responsibility you put in the hands of developers and engineers, it will not matter. This generation of armed robotics will be thought through and debated. Once the technology gets simpler to create, all ethics and responsibility will be thrown out the window. There will be people that will build these things just because they can. If certain sections of humanity think it’s moral and ethical to blow up themseleves for a cause, then arming technology to do it is a no brainer. The only question will be is when will the technology be easy enough for these people to do it.

@Kenneth You’re right, it’s not a fundamental right for humans, it just happens to be a right for people in the US. And it may not be the greatest idea, although THAT’S not likely to change anytime soon.

@Scott Um, is armageddon on the way…?

@Jake Part of what worries me is that decisions about arming robots will be left up to the people in charge rather than the people whose lives the robots could be saving. I imagine it’s much easier to make ethical arguments about combat robots when you’re not being shot at.

@Kenneth (again) I think there’s an important and relevant distinction between a human SOMEWHERE in the loop (even if it’s a huge loop) and a human not in the loop at all. Even if it’s 100-1, that 1 is still there to make the critical decisions, you’d assume.

@Seabird I tried to find out what story that was but Google failed me, any idea what book it might have come from?

@Baker The argument about statistics, however hypothetical they may be, is a good one that doesn’t seem to work at ALL. It’s like we accept that when humans are involved, accidents happen, because humans are fallible and that’s just the way it is. Innocent people killed in Iraq? That’s bad, but we just sort of skim over the headlines and sigh and move on. Even drone attacks don’t trigger outrage or anything presumably because there’s a human in the loop somewhere. I’ve argued before that robots would be much BETTER at having accidents, simply because you have data to work with and presumably you’d be able to find whatever logic flaw caused the accident and then correct it, not only in that particular robot but in all other robots. But basically, we’re going to have to get over the arrogance that humans are allowed to have accidents but robots aren’t.

@Eric Uh, there’s no stopping “progress” I guess…

So… I’m confused. By saying my argument doesn’t work are you saying it is flawed? From what I can tell we seem to be saying the same thing: robots could potentially perform better with fewer mistakes/accidents than a human could. Even if the robot still makes mistakes we would still benefit over having the human perform the task instead of the robot. Robots don’t need to be infallible to be beneficial.

@baker Heh, no, it’s a perfectly valid and logical argument and we’re definitely saying the same thing. The problem is that it doesn’t work in practice, which is endlessly frustrating. What I mean is, you can point to all kinds of numbers and statistics that make the case (quite well) that robots are a better idea than humans in certain situations even if they’re not infallible, but it’s just not going to matter… Even the most minor incident sets things back years simply because of the problem of perception. This is what I was attempting to speak to in my comment: for some reason, humans are allowed to be fallible and make mistakes, but robots are not.

@kenneth , regardless if its a fundamental right or not, if not for guns you would not be living the life you are living now. I could go on some long long long rant but whats the point? You obviously don’t put pants on one leg at a time and I am extremely happy that you and your kind will never change the 2nd amendment without a blood bath.

Well, as far as I can see, the problem with fully autonomous combat robots (or more specifically, all autonomous robots that are potentially highly dangerous to humans, like fully autonomous cars in uncontrolled urban environments, or whatever) is that when there’s noone to keep an eye on things, accountability for stuff that goes wrong becomes very fuzzy.

If a man-operated drone kills innocents, you can, usually, fault the operator. Same for riflemen that end up shooting the wrong people, tanks that shoot the wrong people, or whatever. They’re generally not as violently prosecuted in wartime situations, but at least people can shift blame onto someone.

In case of fully autonomous robots, however, who is at fault when it goes mad and goes on a killing spree? The dude that switched the on button? The dude that programmed it? The dude that supplied the mechanical components? The dude that supplied sensor hardware? What sort of occasional killing spree is okay as long as they don’t happen too frequently? (Like the 95% bomb accuracy rate baker mentioned above?) Does that mean it would be alright to purposely engineer robots that kill people at random once in a while as long as it doesn’t happen too often?

Not all of those might be rational, or make sense, but it’s generally what people are afraid of. And a whole other load of horror situations opens up when the robots aren’t fully self-contained, but while autonomous, communicate with each other in a sort of swarm-ai fashion.

It’s made sense, but I don’t want to be in the boots of the programmer that gets looked at first when a squad of rifle-wielding robots mows down friendlies, only to have it come to light months later that because some other dork managed to mess up the encryption keys somewhere the bots communicated unencrypted and got hacked, or whatever.

It’s less of a problem with entrusting weapons to robots, or entrusting weapons to computers. It’s a problem of entrusting weapons to computers that are NETWORKED, AND not monitored by operators too closely, AND are allowed (and physically capable) to make ‘pull the trigger’ decisions.

That’s a whole host of worst-case scenarios I don’t even want to imagine. It might be slightly silly to assume that combat robots are going to flip out by themselves regularily, but when you have one malicious and smart hacker sit somewhere and find a vulnerability in the system, that’s an entirely different ballpark.

Everything comes down to statistics. I don’t trust any robot with a gun unless I know it has been rigorously tested to the point where it is statistically safe enough for my tastes, which is to say that it is at least known to erroneously kill innocents at a smaller statistical rate than armed humans do. The unethical thing is to give a gun to anything that hasn’t undergone rigorous testing, human or robot or monkey or seeing eye dog, etc etc.

very good post, i certainly love this website, keep on it

Thank you for all of the effort on this blog

You definitely know how to bring an issue to light and make it important. I cant believe youre not more popular because you definitely have the gift.