The future of warfare - robot or nobot?

[Husar]

I just came across this interesting open letter with a lot of signatories, famous people included.

They basically demand that the world should do something against the development and use of AI weapons similar to the restrictions imposed on nuclear, bio and chemical weapons in international treaties. I think it will be interesting to see whether it gets any reaction and I am also interested to see what people think.

http://futureoflife.org/AI/open_lett...nomous_weapons

The key question for humanity today is whether to start a global AI arms race or to prevent it from starting. If any major military power pushes ahead with AI weapon development, a global arms race is virtually inevitable, and the endpoint of this technological trajectory is obvious: autonomous weapons will become the Kalashnikovs of tomorrow. Unlike nuclear weapons, they require no costly or hard-to-obtain raw materials, so they will become ubiquitous and cheap for all significant military powers to mass-produce. It will only be a matter of time until they appear on the black market and in the hands of terrorists, dictators wishing to better control their populace, warlords wishing to perpetrate ethnic cleansing, etc. Autonomous weapons are ideal for tasks such as assassinations, destabilizing nations, subduing populations and selectively killing a particular ethnic group. We therefore believe that a military AI arms race would not be beneficial for humanity.

I think this is a very interesting point, especially since it seems hard to prevent such technology from being rather insecure.
You can probably find the ED-209 video yourselves (contains violence).

[Greyblades]

I'm hoping it will go the way of orbital based weaponry, all sides realize how badly it could end and agree not to even try.

[Ironside]

I got the feeling that the military AI:s will start as internet surveillance rather than field deployments.

So Daedalus and the Puppet Master rather than Skynet.

[Gilrandir]

Unlike nuclear weapons, they require no costly or hard-to-obtain raw materials, so they will become ubiquitous and cheap for all significant military powers to mass-produce.

...which means one can't really control them. I always thought technology is gonna kill the humankind.

[Viking]

Here's a different perspective: better with robots shooting each other apart than humans.

I wonder if they might also not cause insurgents some serious trouble. Robots are fearless - even if insurgents have them, you can just buy more of them, and more sophisticated versions as well. No morale issues there.

[Kagemusha]

I am not so much afraid of them robots taking over. More so with the use of machines as soldiers, i am afraid that the threshold of warfare will decrease, which is like to have impact to human beings as well.

[wooly_mammoth]

I think it's a bit late for that. To my knowledge, the US have already independently developed

a) laser cannons (yes, taffing functional laser cannons currently mounted on destroyers)

b) semi-autonomous walkers (they have a "doggie" that can run around at speeds close to a charging lion while carrying a motherload of equipment)

they just need to but a) on top of b), turn seek & destroy mode on, then things go south and we're all dead.

[Husar]

I think it's a bit late for that. To my knowledge, the US have already independently developed

a) laser cannons (yes, taffing functional laser cannons currently mounted on destroyers)

b) semi-autonomous walkers (they have a "doggie" that can run around at speeds close to a charging lion while carrying a motherload of equipment)

they just need to but a) on top of b), turn seek & destroy mode on, then things go south and we're all dead.

Careful there, because:

a) both your a and b won't go far without a nuclear power plant nearby using current technology. I think b currently runs on gasoline but I doubt it can go very far on that either, especially when heavily packed. And b is also very loud while a is so far not really usable to sink big ships if I'm not mistaken.

b) Neither of those currently have much to do with the topic because putting a onto b would not make it do anything unless a human tells it what to do. The topic is about AI development that would allow the machines to act independently, which is not directly related to the machines themselves.

[Montmorency]

The same old rubbish about "machine" intelligence.

We already have an "artificial intelligence" right here between our ears, and once we realize it no international treaties or social movements will save us.

[I of the Storm]

Well, unlike orbital weaponry (or anything space related for that matter) robot tech is comparatively economical in terms of development and production, i.e. by far easier to obtain than a geostationary missile platform. So the number of people who would have an interest in trying their hands on that stuff is larger. I'm afraid this will see more widespread use.
Regarding the fact that asymmetrical warfare is unlikely to disappear, I'm curious: What would be good and cheap counters to a platoon of combat robots? Things that attack the circuitry, i.e. EMP-based stuff?

[wooly_mammoth]

b) Neither of those currently have much to do with the topic because putting a onto b would not make it do anything unless a human tells it what to do. The topic is about AI development that would allow the machines to act independently, which is not directly related to the machines themselves.

Well, the point is that you pretty much have the toys and it's only a matter of time before you get the AI out. Then things inevitably go boobies up and we'll all get murdered by rebellious robots. I see no way around it unless you stop developing either of the two, which won't happen.

[Tuuvi]

Here's a different perspective: better with robots shooting each other apart than humans.

I wonder if they might also not cause insurgents some serious trouble. Robots are fearless - even if insurgents have them, you can just buy more of them, and more sophisticated versions as well. No morale issues there.

Is there any guarantee that robots would completely replace human soldiers? I wouldn't put it past governments to use human soldiers alongside robots in order to gain numerical superiority, or when the losing the side is growing desperate they might deploy human troops as a last resort.

And speaking of insurgents, the thought of autonomous weapons in the hands of oppressive governments is a scary one, as they can be used to quell dissent and crush rebellions at little cost to the regime using them.

[Greyblades]

Er, I think I was a bit confused in my last post, I was thinking the question was will we end up using AI (as in shodan or skynet-smart AI.) If it's just the bog standard war robots then yeah we'll be getting them, we're allready using drones after all and there's that russian robotic turret, I dare say we'll be seeing a lot more automation on the battlefield.

Only thing is that I dont see anyone replacing soldiers with them.

Firstly for cost as a robotics will be freaking expensive compared to the average trooper, replacing entire armies would be ruinous even for the superpowers and the benefit would be questionable; a robot can be destroyed just as easily as any other piece of millitary equipment and they dont have the squaddie's benefit of being self replicating.

Secondly trust, we have enough problems with our soldiers being unable to tell civillians from enemy targets in combat situations. It would be an utter nightmare to try to teach a computer the difference between a child playing cowboys and indians and enemy combatants. This is made worse as most of the wars we've partaken in recently has blurred the line between civvie and infiltrator to hell and back.

Regarding the fact that asymmetrical warfare is unlikely to disappear, I'm curious: What would be good and cheap counters to a platoon of combat robots?

Bullets.

High callibur ones should do fine against lightly armoured equipment and anti tank rifles/RPGs should be able to take out heavily armoured targets. Unless we're talking about futuristic super materials any robotics we develop will be protected by, at best, the same stuff we use to armour tanks, and as we have learned in our middle east sojurns our best protection right now is still somewhat susceptable to high velocity projectile weaponry and the occasional explosion.

Sure getting a disabling shot would be more difficult on compact circuitry than human operators but the machinery the robot is operating should be as damageable as anything else the millitary uses: Break the tracks and it cant move, jam the turret and it cant aim, shoot the cameras and it cant see, destroy the engine/battery and it will stop. Trick will be doing that before it kills you.

As for an EMP well we've already got protections against EMP today, the most basic being the farraday cage, it wouldnt be too hard to make a robot EMP-proof.

[Husar]

The same old rubbish about "machine" intelligence.

We already have an "artificial intelligence" right here between our ears, and once we realize it no international treaties or social movements will save us.

Our intelligence is a natural intelligence while machines is artificial since it was created by another intelligence and did not develop through biological and chemical processes. Surely you can see the difference.

I'm curious: What would be good and cheap counters to a platoon of combat robots? Things that attack the circuitry, i.e. EMP-based stuff?

Depends on the robots, but I would say in the medium or long term at least, they could become tough targets. As Greyblades said, if the armor acts as a faraday cage, EMP should be out. Although maybe other things, similar to a solar storm, maybe very strong rays, lasers etc. Depending on the protection, the best way to destroy robots may be current anti tank weapons and artillery/lots of explosives though. And that's for the more obvious robots, what about smaller flying ones that perform hit and run attacks from the air with pinpoint precision, killing a few soldiers at high speed before they retreat to reload. Against humans the encounter might be over before they realize what hit them or where it is. Or think of nano robots that attack in swarms similar to insects. You won't need big weaponry to destroy them but they may be really hard to stop. They could even sit on trees and then explode in your face when you come close or otherwise kill you.
Maybe I just have really nasty ideas but I would think there is a lot of potential.

Well, the point is that you pretty much have the toys and it's only a matter of time before you get the AI out. Then things inevitably go boobies up and we'll all get murdered by rebellious robots. I see no way around it unless you stop developing either of the two, which won't happen.

But the point of the letter is not to get the AI out. For the same reason the US do not release tons of Milzbrand containers.
Since their point is that getting the hardware is easy but the trick lies in getting the software as AI development is not exactly a very simple field apparently.

Is there any guarantee that robots would completely replace human soldiers? I wouldn't put it past governments to use human soldiers alongside robots in order to gain numerical superiority, or when the losing the side is growing desperate they might deploy human troops as a last resort.

And speaking of insurgents, the thought of autonomous weapons in the hands of oppressive governments is a scary one, as they can be used to quell dissent and crush rebellions at little cost to the regime using them.

I agree, especially with the last resort part. The scientists are also worried about the softare making it to terror groups, who can then send genocidal robots into cities that the police may have trouble to stop. Think ISIS robo death squads.

Er, I think I was a bit confused in my last post, I was thinking the question was will we end up using AI (as in shodan or skynet-smart AI.) If it's just the bog standard war robots then yeah we'll be getting them, we're allready using drones after all and there's that russian robotic turret, I dare say we'll be seeing a lot more automation on the battlefield.

You understood this correctly, it is less about lasers and satellites and other hardware, it is more about the software that would make the hardware far more deadly by allowing it to act completely independent from humans.

Only thing is that I dont see anyone replacing soldiers with them.

Firstly for cost as a robotics will be freaking expensive compared to the average trooper, replacing entire armies would be ruinous even for the superpowers and the benefit would be questionable; a robot can be destroyed just as easily as any other piece of millitary equipment and they dont have the squaddie's benefit of being self replicating.

A robot would not necessarily be as easily destroyed (see above, there can be many types of robots), wars would be far easier to sell to the population. And about cost, I'm not so sure, current missiles often cost a million or more per piece, the computer hardware required may be in current mass-produced cellphones and part of their worry is that once the software is developed, it can be copied and spread at almost no cost like any other software. In the worst case it could be obtained from a disabled robot or even by remotely hacking one.
A sufficiently versatile AI might even be able to work in different kinds of robots so that insurgents or terrorists could use the AI of a sophisticated robot and let it learn to use their home-built crappy bot to murder people they don't like.

Secondly trust, we have enough problems with our soldiers being unable to tell civillians from enemy targets in combat situations. It would be an utter nightmare to try to teach a computer the difference between a child playing cowboys and indians and enemy combatants. This is made worse as most of the wars we've partaken in recently has blurred the line between civvie and infiltrator to hell and back.

Some people may not want them to make a difference, they may just want them to be able to find and kill every human in a certain area, that's part of the worry in the open letter as I understand it.

Bullets.

High callibur ones should do fine against lightly armoured equipment and anti tank rifles/RPGs should be able to take out heavily armoured targets. Unless we're talking about futuristic super materials any robotics we develop will be protected by, at best, the same stuff we use to armour tanks, and as we have learned in our middle east sojurns our best protection right now is still somewhat susceptable to high velocity projectile weaponry and the occasional explosion.

Sure getting a disabling shot would be more difficult on compact circuitry than human operators but the machinery the robot is operating should be as damageable as anything else the millitary uses.

As I said above, robots do not hve to be human-sized or human-like. And they do not have the same weight restrictions as humans. They can either be faster, nimbler and smaller or a around the same size but carry a lot more metal. And you could combine them. Think bulky rmored street cleaners with swarms of flying drones that search buildings in a networked fashion. If you kill one, the others will close in on the last known location and try to swarm you etc. Then your bullets may be insufficient, especially if the machines are faster and deadly accurate while you are peeing your pants.

[Montmorency]

Our intelligence is a natural intelligence while machines is artificial since it was created by another intelligence and did not develop through biological and chemical processes. Surely you can see the difference.

Not quite, but that's not even my point.

Imagine we discover a small alien spacecraft, and decide that we want to do interstellar travel ourselves. However, instead of having it examined and reverse-engineered, we instead give some chimpanzees sheet metal and duct tape and hope they build us a fantastic FTL arcology for the entire human race.

And then we get scared because clearly Planet of the Apes.

WTF?

[Husar]

Not quite, but that's not even my point.

Imagine we discover a small alien spacecraft, and decide that we want to do interstellar travel ourselves. However, instead of having it examined and reverse-engineered, we instead give some chimpanzees sheet metal and duct tape and hope they build us a fantastic FTL arcology for the entire human race.

And then we get scared because clearly Planet of the Apes.

WTF?

So your point is that what we are doing is stupid or that the AI would never be that good anyway and we get scared for no good reason?

[HopAlongBunny]

If we go down that road ie: bot warfare; its hard to escape the same logic that worked in Traveller.
Massive numbers of the cheapest delivery systems that work for a given situation.
Mass produced, dirt cheap and totally expendable; its the totally expendable part that is worrisome.
Why not choose war; its not like any of our actual people are going to hurt...
Among those nations with the capability, the hopeful result is technological competition and stand-off; the rest of us just hope like h*** no one comes after us.

Decent article from the economist:

http://www.economist.com/node/21556103

The bots have arrived already

[Montmorency]

No, the point is simply that we have a brain and we're worried that machines will surpass it, yet the only way this can happen given our primitive understanding of brain dynamics is pure accident in terms of "throwing at the wall to see what sticks". It should be obvious that trying to rebuild from scratch something accumulated over many millions of years of evolution is a pretty stupid idea. Ever hear the old saw about the monkeys, the typewriters, and Shakespeare?

If we were inclined to use the same computational tools to instead work toward creating a coherent and practical picture of brain dynamics (i.e. "natural" intelligence), then we would immediately be better-placed to create some sort of superior intelligence, whether machine or biological - the latter scenario being the one we should find more plausible and dreadful.

[Viking]

Is there any guarantee that robots would completely replace human soldiers? I wouldn't put it past governments to use human soldiers alongside robots in order to gain numerical superiority, or when the losing the side is growing desperate they might deploy human troops as a last resort.

It means that the option to replace human soldiers with robots will exist. Some countries might decide that it's normally hard to justify sending human soldiers into war, and will replace them with robots wherever and whenever they can. Others might simply find robots more reliable in many or most positions. Yet others might find robots more cost-efficient than human soldiers.

Cannon fodder is nothing new, so I doubt robotics would introduce a new low here - history provides strong competition.

High callibur ones should do fine against lightly armoured equipment and anti tank rifles/RPGs should be able to take out heavily armoured targets. Unless we're talking about futuristic super materials any robotics we develop will be protected by, at best, the same stuff we use to armour tanks, and as we have learned in our middle east sojurns our best protection right now is still somewhat susceptable to high velocity projectile weaponry and the occasional explosion.

Sufficiently advanced robots would be able to stop most projectiles launched at them (e.g. with lasers or counter-projectiles).

[Gilrandir]

b) semi-autonomous walkers

Do you mean or

Firstly for cost as a robotics will be freaking expensive compared to the average trooper, replacing entire armies would be ruinous even for the superpowers and the benefit would be questionable.

When first mobile phones appeared (at least in Ukraine) they were so expensive that only very rich businessmen/gangsters could afford them. Usually such people sported a thumb-thick gold chain around their neck, Mercedes 600, a jet of their own and a mobile phone. Twenty years later they became quite affordable (even in Ukraine). Gadgets are prone to get cheaper therefore ubiquitous.

[Greyblades]

A robot would not necessarily be as easily destroyed (see above, there can be many types of robots), wars would be far easier to sell to the population.

The question was if there was a cheap counter to robots, I cannot say there is a perfect counter but there is no reason to believe that robotics would magically become immune to current weaponry.

And about cost, I'm not so sure, current missiles often cost a million or more per piece, the computer hardware required may be in current mass-produced cellphones and part of their worry is that once the software is developed, it can be copied and spread at almost no cost like any other software. In the worst case it could be obtained from a disabled robot or even by remotely hacking one.
A sufficiently versatile AI might even be able to work in different kinds of robots so that insurgents or terrorists could use the AI of a sophisticated robot and let it learn to use their home-built crappy bot to murder people they don't like.

When first mobile phones appeared (at least in Ukraine) they were so expensive that only very rich businessmen/gangsters could afford them. Usually such people sported a thumb-thick gold chain around their neck, Mercedes 600, a jet of their own and a mobile phone. Twenty years later they became quite affordable (even in Ukraine). Gadgets are prone to get cheaper therefore ubiquitous.

Most people still require long term payment plans to afford the newest cellphones, and you wont lose any wars by having outdated cellphones
The cost I refer to is replacing an entire nations millitary, which would require the development and manufacture of robotic replacments for every single piece of millitary equipment and personel in a nation's arsenal, not to mention constand updating and upgrading. Millitaries are expensive enough with human operation; automation is an extra cost that most wont be able to afford, so I predict a somewhat limited use for quite a while

Some people may not want them to make a difference, they may just want them to be able to find and kill every human in a certain area, that's part of the worry in the open letter as I understand it.

Yep. What I said will likely only apply to first world nations, and I dare say the less scrupulous nations implementing indiscriminate killer drones will pressure them into deploying thier own long before drones can be programmed to tell friend from foe reliably.

As I said above, robots do not hve to be human-sized or human-like. And they do not have the same weight restrictions as humans. They can either be faster, nimbler and smaller or a around the same size but carry a lot more metal. And you could combine them. Think bulky rmored street cleaners with swarms of flying drones that search buildings in a networked fashion. If you kill one, the others will close in on the last known location and try to swarm you etc. Then your bullets may be insufficient, especially if the machines are faster and deadly accurate while you are peeing your pants.

Yep, I didnt say it was a perfect counter, but there's no reason to believe each individual drone will be any less vulnerable to a large bullet traveling at supersonic speeds than anything else, assuming you can hit it before it kills you. Plus the same thing can be said of a squad of soldiers, shoot one and the rest will be gunning for you.

Sufficiently advanced robots would be able to stop most projectiles launched at them (whether with lasers or counter-projectiles).

Lazers and counter projectiles do not work against supersonic projectile weaponry and there isnt an armour made that is immune to all modern ordinance.

[Husar]

Lazers and counter projectiles do not work against supersonic projectile weaponry and there isnt an armour made that is immune to all modern ordinance.

Some manufacturers would disagree.
The iron fist was apparently successfully tested against kinetic penetrators from tanks.

https://en.wikipedia.org/wiki/Iron_F...ountermeasure)

In May 2011, the system intercepted Kinetic energy penetrators and Metis anti tank missiles during a test in the U.S.

Apparently it is not currently deployed, but the technology is there.

[Greyblades]

If you can stick that on a robotic tank you can stick it on a manned one, plus I dont think it can neutralize a heavy artillery shell, a carpet bombing, let alone a nuke.

My point was that robots arent inheirantly an invincible threat; they're still going to be about as vulnerable as anything else in the armies they serve.

[Brenus]

Plus problem of software. How IA will identify targets: wearing a uniform? What about civilian having guns? And I think that is why for the moment we will stick to armored mules. Armed Armored Mule will be the equivalent of drone, with less visibility and much more vulnerability.

[Husar]

If you can stick that on a robotic tank you can stick it on a manned one, plus I dont think it can neutralize a heavy artillery shell, a carpet bombing, let alone a nuke.

My point was that robots arent inheirantly an invincible threat; they're still going to be about as vulnerable as anything else in the armies they serve.

Now you're moving the goalposts. You said we can't stop really fast kinetic penetrators and I showed you that we can. Artillery shells are different and there are systems which can stop them, but those are different systems entirely.
And yes, you can tick it onto a manned tank but that says nothing about whether or not it can defend robots.
Carpet bombings and nukes are completely different issues. Would you nuke London to stop the five killer robots wreaking havoc in the city?

My point was that robots are easier to defend because they do not need the room to comfort humans, they can be designed without huge interior spaces and without taking into account ergonomical and work safety factors, that IS a huge advantage.

Plus problem of software. How IA will identify targets: wearing a uniform? What about civilian having guns? And I think that is why for the moment we will stick to armored mules. Armed Armored Mule will be the equivalent of drone, with less visibility and much more vulnerability.

Derp, way to miss the point of the letter, which is about the question whether we should try to develop AIs which can do just that. The status quo is not relevant, it's about whether we should develop all that in the first place. Noone said current AIs are sufficient to do that.

[Greyblades]

Now you're moving the goalposts. You said we can't stop really fast kinetic penetrators and I showed you that we can. Artillery shells are different and there are systems which can stop them, but those are different systems entirely.
And yes, you can tick it onto a manned tank but that says nothing about whether or not it can defend robots.
Carpet bombings and nukes are completely different issues. Would you nuke London to stop the five killer robots wreaking havoc in the city?

I moved the goalpost because I had been proven wrong in one insignificant point but the main thrust stays the same: a good cheap counter to a platoon of robots will be the same as the good cheap counter to anything else: overwhelming ordinance.

My point was that robots are easier to defend because they do not need the room to comfort humans, they can be designed without huge interior spaces and without taking into account ergonomical and work safety factors, that IS a huge advantage.

and not incompatible with my point... so why are we arguing again?

[a completely inoffensive name]

The murdering robots won't kill off humanity, it will be the robots that are programmed to love us...

[Husar]

I moved the goalpost because I had been proven wrong in one insignificant point but the main thrust stays the same: a good cheap counter to a platoon of robots will be the same as the good cheap counter to anything else: overwhelming ordinance.

and not incompatible with my point... so why are we arguing again?

As I said earlier, not all the robots have to be machines similar to tnks or soldiers, they can also be similar to insect swarms or single insects, perhaps even the size of microorganisms and I do wonder how you want to kill those off easily with known weaponry. Bigger is not always deadlier. Small robots could lay ambushes and be almost undetectable until they strike.
You seemingly keep reverting to the idea that a robot has to be some kind of tank with heavy armor when IMO it could also be a mosquito that explodes next to your head or drills into your ear channel and through your brain or flies into your eye like a bullet. How do you kill that off, espeically when attacked by hundreds or thousands from many directions? Nuke your own platoon? Yes, they will be very vulnerable to tank rounds, but how many do you think a tank can kill before they crawl into it and kill the crew or destroy the engine?
How many of them can a squad with assault rifles and machine guns kill before they are all dead?

The murdering robots won't kill off humanity, it will be the robots that are programmed to love us...

Why? Sounds like a good solution for overpopulation and environmental problems.

[Brenus]

"Derp": ? http://www.oxforddictionaries.com/de...n/english/derp

[Husar]

"Derp": ? http://www.oxforddictionaries.com/de...n/english/derp

It was more or less just a filler that was somehow meant to indicate that you completely missed the topic.
A bit like this: https://www.youtube.com/watch?v=1ytCEuuW2_A