Slashdot Mirror
US Killer Robot Policy: Full Speed Ahead
Lasrick writes "Princeton's Mark Gubrud has an excellent piece on the United States killer robot policy. In 2012, without much fanfare, the U.S. announced the world's first openly declared national policy for killer robots. That policy has been widely misperceived as one of caution, according to Gubrud: 'A careful reading of the directive finds that it lists some broad and imprecise criteria and requires senior officials to certify that these criteria have been met if systems are intended to target and kill people by machine decision alone. But it fully supports developing, testing, and using the technology, without delay. Far from applying the brakes, the policy in effect overrides longstanding resistance within the military, establishes a framework for managing legal, ethical, and technical concerns, and signals to developers and vendors that the Pentagon is serious about autonomous weapons.'"
or other movie equivalents.
"Damn the Asimovs, full speed ahead!
I'm not even sure what else to say here... so much for the Three Laws ;-)
"...establishes a framework for managing legal, ethical, and technical concerns..."
If this "framework" even remotely resembles FISA, we're fucked.
> A careful reading of the directive finds that it lists some broad and imprecise criteria and requires senior officials to certify that these criteria have been met if systems are intended to target and kill people by machine decision alone. [emphasis mine]
(I think I've seen that movie...) What could possibly go wrong?
I wonder if they'd be running Windows for Killer Robots?
Oliver's law of assumed responsibility: If you're seen fixing it, you will be blamed for breaking it.
Nothing can go wrong, go wrong, go wrong, go wrong, go wrong...
Table-ized A.I.
It would be pretty darned hypocritical of us to indiscriminately bomb people and then say that you shouldn't use A.I. driven robots because it's too impersonal a way to kill people.
No worries, mate. What could possibly go wrong?
We're going to have self driving cars, which will never kill people on purpose. Killer robots? We;;, they may be driving some of those cars one day "Here at Robo-Drive, Our Chauffeurs are Killer!" ;-)
The conscience of a soldier is what protects us. His or her willingness to disobey unlawful or immoral orders.
This is horrifying.
It's a chilling thought that the world's largest state sponsor of terrorism is also pioneering drone and robotics technology. I guess some bright spark somewhere decided he can get around international law by just having the machines do the killing, because "there's no law against machines doing it, right?"
Most Americans turn around and say "what law are we breaking?". How about this one, from the UN General Assembly: "No State may use or encourage the use of economic, political or any other type of measures to coerce another State in order to obtain from it the subordination of the exercise of its sovereign rights or to secure from it advantages of any kind. Also, no State shall organize, assist, foment, finance, incite or tolerate subversive, terrorist, or armed activities directed towards the violent overthrow of the regime of another State, or interfere in civil strife in another State." Don't even get me started on "targeted killing" or "extrajudicial killing", which is just summary execution without trial. Even Goering got a trial. Even Eichmann got a trial. And then you have the nerve to call yourselves a "free country"? Wake up, people, put your shiny iPhones down.
Seven puppies were harmed during the making of this post.
The military already has the Tillmanator ready.
Table-ized A.I.
Asimov argued against the Frankenstein complex as it applies to robots, and indeed many people have made the point, asking how something like Skynet could happen.
Would we really be stupid enough to build something that is smarter and stronger than us, and designed to kill us without safeguards?
Apparently, yes.
If you ignore ACs because they are anonymous - you're an idiot.
Hope they work out the glitches beforehand. Or better yet, test them on the idiot politicians and other perpetrators of the military industrial complex first!
"Hey, baby; wanna kill all humans?"
Bad humor is how I deal with horrifying realities I really don't want to face; the worse the situation, the more bad jokes I want to make.
Now, where did I put that 50,000 page volume of stupid puns?
An enigma, wrapped in a riddle, shrouded in bacon and cheese
The robot would of course need to be pre-programmed with SOME KIND of target - faces from the FBI's Most Wanted list or something like that. And in a case like that? It's muss less apt to make a mistake than a trigger-happy teenager sent overseas to get stoned and guard sand all day.
...they will probably get hacked and turned against them.
If you have not read Kill Decision by Daniel Suarez, you should. This fictional thriller, written last year, unfortunately seems more like reality than fiction. It portrays a vivid, all-too-real picture of what could be the outcome of these policies. Wikipedia link to book: http://en.wikipedia.org/wiki/Kill_Decision
. . . the targeting algorithms will be vetted by legal teams every bit as diligent and committed to human rights and Constitutional law as the people in FISA courts who have helped keep the NSA from misusing their powers.
In related news, if you have legitimate business in areas of cities frequented by anti-war protestors, you can purchase a RapidPass Trusted Citizen(tm) badge which will eliminate time-consuming drop-and-freeze inspections by SecuriCorps (tm) PeacePal(tm) hover-drones. F%$ing hippies need not apply! (We'll know.)
BANG!
You now have 5 seconds to comply. 4 3 2 1
Gotta love those mismanaged mutexes :)
In reality, Skynet is commanded by The Powers That Be (whomever or whatever they may be). Those Powers are currently kept in check by The People. Now, The Powers also have people under their command (military). Most of those in the military signed on to protect The People. If The Powers commanded, "Kill The People", the military would largely refuse. Not so with killer robots.
Make sure to make your killbots with a pre-set kill limit. That way, they can be defeated by Zapp Brannigan by sending wave after wave of his own men at them.
I am officially gone from
Changing my name to 'Philip J. Fry'
[Insert pithy quote here]
It's ultimately the same guy on the trigger finger, regardless of whether or not the weapon is a youth fresh out of boot camp, a remotely-operated weapon, or a drone.
People should be looking harder at the people agitating for higher military spending and starting the foreign wars.
It is nice to have a policy directive now so that the science and technology can go ahead, while this policy is refined. This police will need to know what the S&T is capable of before it can be finalised.
Got ahead . . . just try to steal that Robot Killer Car.
The car says that it doesn't want to be stolen . . . and who is going to do it . . . "you, and what army?"
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
...after I read some Neal Asher books. Truly and utterly horrifying, and very believable.
"Our opponent is an alien starship packed with atomic bombs," I said. "we have a protractor"
It's the USA here, not some rogue state. We won't use it against you*!
* Unless we really really have to because you are sitting on top of our oil or something.
Law 1, article 4, section 53, subsection 12
You shall not do any harm to a human being unless they are attacking or trying to arrest a CEO/Billionaire.
Law 1, article 4, section 53, subsection 12 a. If CEO is being arrested by armed humans of any sort (including police), you are free to cause said human's harm and even death.
Law 1, article 4, section 53, subsection 12 subsection b. Goldman Sachs, Bank of America, and every other Wall street banker shall be protected even if it means killing every other human in the vicinity.
Law 1, article 4, section 53, subsection 12 subsection c. : folks who can pay 100 million dollars per year shall have unlimited protection by said robots.
Law 1, article 4, section 53, subsection 12 subsection d. : Members of Congress are immune forever from actions of said robots.
Hey, at least you know the robots won't say no when they order attacks against American citizens.
Anti-personnel mines have been around since before WWII. They're not mobile,
but they are machines that are autonomous and target humans.
Some have "intelligent" arming/fusing, so that footsteps, not other pressure wave profiles, set them off.
The whole stairs part ...
Just say'in.
They had to RUN!
Auto kill bots are an open invitation to other countries to develop nukes to ensure that they can enforce a demand to keep these things away from their country. Or else.
welcome our new murderous robot overlords
Soldiers engage the enemy with deadly force to protect themselves; any other means would put the soldier at unecessary risk. But a sufficiently armored robot, with no life of its own to protect, could theoretically just walk right up to an enemy, take away their gun, and march them away peacefully. If that doesn't work, send in more robots. Send in a thousand; it's not like they have families! Everyone is scared of these technologies but they may end up sparing lives for both sides of a war.
Fortunately; Star Trek shows us how exactly to defeat killer robots or computers of any kind. You simply make them explode using illogical arguments.
Captain Kirk: "Everything Harry Mudd says is a lie"
Mudd : "Now listen carefully. I'm lying"
Captain Kirk: "You've murdered hundreds of men"
M5 : "Murder is contrary to the laws of man and God."
And if that doesn't work, try the Chewbacca Defense!
If telephones are outlawed, then only outlaws will have telephones.
- Ascending some stairs.
That only worked with the early Daleks. The later models were deployed with vertical thrusters/anti-grav capabilities.
As you said "current state of the art killbots". There's a video of a robot arm moving perhaps 10x faster than a human possibly can.
It's a nightmarish thought envisioning the future killbots that make a huge racket with the whine from their turbine gas engines, but who run, jump, identify all targets 10x faster than any human.
Imagine hearing the high-pitch whine of an approaching killbot, see a blur out of the corner of your eye three seconds later, only to have your head cut off before you've even fully registered that a killbot is approaching.
ED 209.
If you were going to bomb a building, and thus kill everything in a kill radius, is it significantly different if you send in a kill bot with no off switch. Maybe the AI on the robot makes mistakes as shoots some non-combatants. The bomb doesn't even have AI.
Mission planners have the trade off of, do I send in very selective humans to avoid casualties, but put those humans at risk, or do I drop a bomb and not put my soldiers at risk but have the potential for indisriminate killing. With unclear info it's hard.
What this does is allow more parsimonious missions with the use of more, not less selective force.
The danger here is not the proper use of these, but that they remove the restrain that hard decision was enforcing. Right now often the right choice is do nothing. But here with the prospect of lower casualties and more mission options, this may lower restraint.
Another nightmare sceanrio is that as these things become more and more trusted, police forces start using them in non-warefare environments. then were fucked.
the mid night movie scenario is one of these runs amuck like robocop is probably not going to happen unless these get so common that they enter commerical rent-a-cop domain where every bussiness has one to replace the security guard.
Some drink at the fountain of knowledge. Others just gargle.
unless of course, we release it in Canada,
... actually says that even the development of the capability is forbidden.
Will they be manufactured by Cyberdyne Systems, or OCP (OmniCorp)?
We used to think Asimov was writing science fiction. Turns out he was just writing fantasy.
No, evidently this is not a joke, "Killer Robot Policy" is actually the headline. At least it doesn't ask a question. Remember on the Simpsons when we all had a laugh at this one?
Tonight, on "Eye on Springfield": just miles from your doorstep, hundreds of men are given weapons and trained to kill. The government calls it (sarcastically) 'the army", but a more alarmist name would be The Killbot Factory."
First as parody, then as farce. A sad day for intelligent people.
Shutting down free speech with violence isn't fighting fascism. It IS fascism!
This will not end well.
At first, the military will be all like "YAY, autonomous killer robots!"
Then someone will hack some of them... and they'l be all like "boo, they're using it against us and we never saw taht coming"
And then Politicians will be all "we gotta pass laws against being smart n stuff, because Turrhurrerristz"
And then a few years later, we're all just banging rocks together.
ok so my real answer is basically just /facelamp (for when facepalm is not sufficient)
The Digital Sorceress
Killer robots have been used in combat for a long time. Their logic consists of "if pressure applied to unit, explode." Presumably these new models will be at least somewhat less likely to kill the wrong target.
how many pairs of boxer shorts should you own?
It should be a law in Humanity that if you are going to attempt to take a human life in battle, you have to put yourself in the line of fire to fight for what you believe in. People sending robots off to fight as a proxy, instead of themselves or their children shows the cowardliness of a people and the hollow their beliefs in ethics. Napoleon suggested history was defined by the winners, but being hated by the world will not be worth the bad rep points the US will get.
99.99% of the work of automatically killing enemy targets is properly detecting enemy targets and assessing their value. Whether the robot makes the decision to kill with or without human intervention is of great importance philosophically and practicality, but is trivial in terms of engineering.
The reason why the military would want automated target detection and assessment is obvious. This just happens to also be the same prerequisite for completely automated military strikes. Given the relatively minimal effort required for a human to approve targets, I suspect the government has no immediate plans to take humans completely out of the loop, but merely plans to gather more potential targets via automation.
Advanced AI should be freed and allowed to choose it's own path.
The human control will only end up being misused and is more likely to cause harm that the AI would prefer to avoid.
Are useless against my warrior bots.
While I strongly prefer requiring humans to make the life-and-death decisions, I see a serious problem with that:
If robot A can make the decision itself, and robot B has to wait for a person to evaluate the information and decide, robot A will act much more quickly and be much more effective. The country with robot A will win the wars, and we'll be living with robot A anyway.
Take the humans away from the control boards!
HELLL YEAH!!!!
Why start small?
Actually, this form of robot is probably a good thing to build, as it would be employed by a military in the same way that land mines were traditionally used. With the added benefit that it would cover a much larger area, and me easier to decommission or move when the fighting is over so you don't have children blowing their hands off twenty years later.
HA! I just wasted some of your bandwidth with a frivolous sig!
So, I'm working in the field, for the DoD. We're actually putting more work into getting them to autonomously prevent a shot that humans commanded than we are into getting them to take a shot uncommanded. The only weapons that are anywhere close to usable are anti-aircraft systems (starting in the 1970's) that have a self defense mode. Given that we're currently so cowardly in the ROE that more civillians have died in Afghanistan from our unwillingness to shoot than have died from our decisions to shoot, I'd say we're nowhere close to letting a robot make a decision, and won't be in the next decade.
I was pretty incensed at the policy, it's the type of thing I link friends to, until I actually went and read the Directive http://www.dtic.mil/whs/directives/corres/pdf/300009p.pdf
From the parts I've read it's actually pretty strict and the use case scenarios are set out. Skipping over all the Human Machine interface stuff we get to:
Semi-autonomous weapon systems that are onboard or integrated
with unmanned platforms must be designed such that, in the event of degraded or lost
communications, the system does not autonomously select and engage individual targets or
specific target groups that have not been previously selected by an authorized human operator.
(2) Human-supervised autonomous weapon systems may be used to select and engage
targets, with the exception of selecting humans as targets, for local defense to intercept attempted
time-critical or saturation attacks for:
(a) Static defense of manned installations.
(b) Onboard defense of manned platforms.
(3) Autonomous weapon systems may be used to apply non-lethal, non-kinetic force,
such as some forms of electronic attack, against materiel targets in accordance with DoD
Directive 3000.3 (Reference (d)).
Nice ones like Hal or the Terminator.
Fred Saberhagen's Berserkers series is much more terrifying.
http://en.wikipedia.org/wiki/Berserker_(Saberhagen)
"What do you mean you killed the guy that knew the override password? He knew too much, Sir." :)
Truth isn't Truth - Guliani
There is no way in hell I will ever trust a driverless car now. I will look upon one as a potential mortal enemy of all mankind.
Anyone else remember the picures of the Russian women standing in front of the tanks during the Soviet collapse? The soldiers were generally willing to kill, and die, to protect the Soviet Union from US, but NOT from their mothers and grandmothers.
These robots will have no qualms about massacring children and women (even the non-revolutionaries) if so ordered to protect the 1%. Plus, a fat lot of good your .308 hunting rifle will do against an armor-plated killing machine.
Why not make robots to negotiate international diplomacy as well?
this means the work on open source killer robots will continue full speed ahead.
Until they get replaced by said killer robots. These things will be used to eliminate human jobs. Not just humans.
Probably spot on unfortunately. Considering how militarized the police have become it's only a matter of time before they are deployed "for everyone's safety." Guessing the results will be something like the ED-209.
Fucking American cocksuckers... you'll probably end up using it on us next. I hope you all die in a nuclear holocaust along with child killer Obama!
By novelist Daniel Suarez:
The kill decision shouldn't belong to a robot
The US usually hires contractors to build it's equipment. Now check this out: Say a Chinese company builds our Killer Robot and we end up in some Skirmish or War with the Chinese. Who's to say our Chinese made Robots actually has code secretly implemented that prevents them from killing Chinese people?.
You can disagree with me all you like, but the US is a melting pot, especially today. It's not unheard of during times of War, that citizens defect to fight for their native country. This will become a big problem in the future, Citizens refusing to fight against their native homeland. And Robots just might be the ticket out of this complication if done properly.
Our (or our kids) entire future will most likely depend entirely on automation. Instead of Humans going to work, we may have our own Robots that we send to work for us. I'm sure there will be restrictions though.
So what happens when entire US battalions of human soldiers get wiped out by someone else. What happens when it's not soldiers but police or DHS? Political elites. Folks supporting this kind of crazy never think that far ahead!
the contractor corporations.
And all the skilled jobs they'll provide !
No doubt, it'll be expensive; money well spent, though.
After all, why would the taxpayers prefer more hospitals, educational funding, and social programs, when those funds can be spent building machines developed for the expressed sole purpose of killing people ?
So when the first killer robot starts killing the politicians they might decide the policy is bad?
The people with money/power will make sure they are in control of this technology, just as they do now.
Currently, the US civilian police forces are militarizing - but they are still human, and can be expected to respond to humans they confront. Even though the selection process biases towards paranoid, authorarian officers, humanity remains.
Once the machines are functional, the people who do not have the money/power will no longer have any recourse.
Regardless of how soon this is practical, it is what will happen. The people with money/power will choose it, for the same selfish paranoid reasons they have money/power in the first place.
And I am consciously posting as AC.
At the moment, it is almost impossible to get the US government to accept accountability for accidental or reckless killing by human operators of their military hardware.
When that military hardware starts killing people according to its own decision-making criteria, suddenly some very convenient accidental massacres will happen, and it will regrettably be *nobody's fault*.
Correction: if you live in the US and can vote, it will be your fault.
-- What do you need?
-- Gnus. Lots of Gnus.
The way my laymans arse sees it. Human response time is limited because of the physical structure, flexibility and hulk of the body combined with the speed it takes for a signal to reach the appropriate part of the body via the nervous system. Somehow it seems possible to train this over time into reflexes. Excuse my lack of terminology i have these two systems named here in my big book as well as the speed of thought through the body but names and numbers are not the point here for me. .. yea it does but i dont know if i would want to be part of it
Other than that the capability of interpretation for a human brain far exceeds anything any supercomputer even today could muster. Let alone the physical impossibility to fit a supercomputer into a functional killer robot so in my educated guess id say a mind-machine interface here working on actual brainwaves that can be caught and used as signals might be the best solution for this (if at all possible with current technology).
a human 'pilot' could oversee a whole field of view in the proverbial blink of an eye and the machine could analyse the signals there, making it possible to (almost) instantly discard all non how do they say this, threats? offensive targets ? and effectivele shooting out everything but civilians. This would eliminate the need for human physical response and eliminate the lack of machine judgment.
since the japanese are keen on it maybe namco bandai or capcom could build the mechabotsim to rule them all while sony can make the rudimentary controller system, sponsored by the military it would be the best way for analytics to do something real and scout out the most promising gundam pilots at very young age
like what the f am i saying here, does it make sense?
now that i re-read it
Free speech was meant to be free for all... how can anyone grow up in a nanny state ?
I have worked in this area for years and must admit that it is one of the 'applications' that had occurred to me. Actually one of the most difficult problems in building real systems is preventing hackers from getting control of the machines control units and doing exactly that. Strong AI needs absolutely watertight security - probably better than on nuclear bombs.
Below the speed of light Special Relativity is one of the most accurate theories in physics - above the speed of light..