'Killer robots' ban must be part of Geneva talks, says campaign group

What do these rules matter ?

Can we please not pretend to miss the obvious truth ? Here's the choices we've already made : What is more important ? "respecting" religion or human rights [1]. Choice made : "respecting" religion (read: letting muslims execute ex-muslims, gays and ...) What is more important ? Peace with China or human rights (See Tibet, XinJang and ...) ? Choice made : peace/trade with China. What is more important ? Communism(/socialism) or human rights ? (because in the case of Cuba and Venezuela they're opposing forces) Choice made : communism is more important. What is more important ? Human rights or the Trade with Indonesia (muslim state, executes gays, has plans to execute ex-muslims, ...) ? Trade with Indonesia

Human rights are dead letter. Why ? Because we've chosen against them. The same goes for this principle.

[1] http://en.wikipedia.org/wiki/Human_rights_in_Saudi_Arabia

Let's do that. Let's ban bad people. Let's make it international law that all disagreements must be settled via pillow fights with squadrons of 12 year old girls. Of course the Muslim extremists would naturally lose since they'd never let an 12 year old girl out of the house in the first place. So they'd have to send boys. And since proper war fighting pillows are in short supply in places like Yemen, then they have to use more readily available materials like diesel fuel and fertilizer, you know just to be fair, given their logistical disadvantages in lacking both available 12 year old girls and properly spec'd combat pillows.

So now the law abiding rest of the world sends in the 12 year old girl combat pillow division (polyfill only since feathers have been banned as cruel to birds) and the promptly get set on fire and raped by an enemy that obviously does not give a flying fuck about a group of cognac-sipping first-world diplomats who decide to ban killing when waging war.

The prolonged wars in Iraq and Afghanistan happened despite the repercussions for lost sons and daughters. We're losing people right now, and hundreds of thousands of Iraqis lost their lives because we didn't have enough manpower (or weren't willing to take the risk) to enforce order. If we could spin up machines the way we spin up dynos, these problems would go away. No American soldiers would die. We'd spin up production and deployment to enforce near total order. We could reduce Iraqi deaths by orders of magnitude. The machines don't necessarily have to act like terminators. They could act like military police too.

This might be a scary future, no doubt. But it might be a bright one too. If we trust our values and trust that our elected representatives won't go on mindless killing sprees, the machines might be a blessing in disguise. (Also note that by the time we get machines with this capability, our government, social structure, our entire economic model, and our appreciation for civil liberties will evolve quite far from where it is now too).

We already have drones that kill and injure human beings. They're called missiles and torpedoes, and the third world has plenty of them too.

Also, there's this assumption that keeps getting made that humans will always make better decisions than machines. Personally, I think you're seriously overestimating the ability of scared, tired 18 year olds to make quick rational decisions under extreme pressure. Every single war ever, is littered with examples of human beings making mistakes, which end up with the deaths of allies or civilians, it's not even remotely rare.

And you know what, machines will make mistakes too, they'll kill their own side or civilians occasionally. But after they do that, their software can be improved and refined, over and over and over, with no limit. They will constantly and forever be improved.

A human soldier will not. There's only so much training you can give to someone in nine months before they're in combat.

And there's also the fact that machines are a helluva a lot less likely to think it's funny to piss on a Koran or to go rape some of the locals.

Robots don't change anything. Your evil rich people can just as easily fight the rebels with human flown planes or human guided drones and paid mercenaries.

And please with the "imagine a robot shooting a kid, we can't ever allow that to happen" hypothetical. Soldiers kill innocent people all the time, just on accident. Missiles miss, land mines get abandoned and kill innocent kids, warfare has never been tame. A robot would arguably be better because it might rarely if ever miss, it could have better vision than a human, it could possibly be a better judge than even a human with sufficient AI, and it never gets emotional, angry, or tired or insane.

The drones are perhaps a good opportunity to look at the broader picture. Maybe we need to recognize that one country sending weapons and troops into another country on their own initiative is a fundamental violation of human rights and basic morality. The only way troops or weapons should be deployed is to stop greater harm from happening, and the only entity that can judge the greater harm is one that looks beyond national borders.

For me the simplest way to "clean up" war is to make all war illegal, and only allow troop and weapon deployments outside of a country's own national borders under the UN banner for peacekeeping purposes. This sidesteps the rich kids vs poor kids issue because there is no us vs them, there is only the global community and how it polices itself. Not that the UN is necessarily ready for such a thing, but even in its inadequate capacity it would be a whole lot better than the horrible inhumane mess we are in right now.

Automation is coming whether we like it or not. If we end up with a world were large armies are no longer fielded and only robots exist on the combat field perhaps we can move weapons to those that incapacitate instead of kill. We can only get there is we continue to push development. Regardless, strikes will become more surgical which could reduce casualties amongst non combatants. Figure it this way, if they become good enough maybe the leaders will be the ones looking over THEIR shoulders.

The disposable machines are built by human beings belonging to a nation (or entity) who are interested in self-preservation and protecting the nation. Of course profit doesn't hurt.

Thinks about this. Lately, the rich nations are facing shortage of men willing or able to risk death/injury. The populace lacks the stomach.

And the 'poorer' nations (ones with little to lose) have no qualms about bloodshed as long as they get what they want.

I see drones in warfare as inevitable. How autonomous will they be? Think no one knows for sure at this point...

Eventually, you can expend your supply of Killer Humans if you get too bloodthirsty. You can wage war to the point where the populace is no longer willing to support you--even the elaborate decoupling of war from daily life the US has accomplished was still not enough to prevent the slow tide of opinion from shifting.

Killer Machines, however, can be replaced as they are expended, and the plans setting them in motion proceed without issue. The populace never stops it--as long as there is money to buy material to make the things, they can be used.

Thus, the use of drones in warfare is not self-limiting the same way that the use of humans is.

Chemical, biological weapons use has been greatly curtailed by rules of war. So has "aggressive" war (war, because we want your territory). So have various civilian targeting tactics. They haven't been eliminated, but there is a lot less than otherwise. It's not perfect. Enforcement is erratic and case-by-case. Powerful countries are more exempt than weak ones.

Nuclear proliferation has been effectively slowed by the NPT. Again, not perfectly but better than nothing.

But one thing needs to be clarified here: the fear over "killer robots" is that the capacity to make war could be placed largely on autopilot. The danger isn't so much a Terminator-style robotic uprising, which is so farfetched in today's world as to be sensationalistic (if not laughable). Rather, the danger -- a very real one, at that -- lies in an accidental war as the result of malfunctions or data-interpretation errors.

We have more to fear from dumb AI than from smart AI.

As it turns out, the US and the Soviet Union came perilously close to nuclear war on multiple occasions over the result of computer errors. (For an interesting look into the subject, I recommend Eric Schlosser's "Command and Control," a recently released history of nuclear weapons policy over the last 50 years).

Banning autonomous or semi-autonomous tactical systems ("Terminators") seems misguided and impractical. These things will get built, and some of them are being built already. The jury is still out on the ethics of building them (Do they save lives by taking humans off the front lines? Or do they take lives by turning war into a sort of game?). But they're being built, and that genie's not going back in the bottle.

But banning fully autonomous strategic systems ("Skynet") seems more fruitful and worthwhile. Again, this is not because we expect "Skynet" to become self-aware and intentionally initiate a nuclear holocaust. It's because "Skynet" might misinterpret something and accidentally initiate a nuclear holocaust.

(I don't mean to dismiss the very real role of human error in this same domain. But fail-safes, in the form of humans having the final decisionmaking authority over warfare at a strategic level, make a lot of sense.)

I'm a layman here, and I'm way out of my depth in discussing the intricacies of war making AI. But the way I see it, we need to avoid a sort of dangerous middle ground here. Either we keep AI dumb and intentionally non-autonomous, or we make AI a heck of a lot smarter (therefore reducing the risk of fatal errors).

In a weird way, building Skynet might be less risky than getting halfway to Skynet. A relatively dumb AI with full autonomy is a frightening thing. An extremely smart AI with the same degree of autonomy is a little unnerving, perhaps, but it's significantly less dangerous in practical terms.

I had the opportunity to build my own robot in it, the furthest I was able to go was getting 2D navigation working based on some libraries in ROS using a laser scanner. Unfortunately I can't upload that video since my computer was too slow, but I made another one.

http://www.youtube.com/watch?v=WhNGUJjV1zQ

I don't see how it's absurd to be opposed to technology that makes violence more convenient for wealthy aggressors. One could argue that reducing the human costs of war would only bring more war.

Of course, monitoring each shot is very tedious, hence:

  UNITS="wastelayer merciless thunderdeath"
  UNIT_SIZE=4096
  for UNIT in $UNITS; do
    for I in `seq $UNIT_SIZE`; do
      yes | ssh operator@$UNIT$I epclient
    done
  done