Asimovs laws are being ignored!

http://news.bbc.co.uk/1/hi/technology/6583893.stm

I dont think anyone would agree with a robot having the ability to kill or arrest someone

I remember once that they were testing one of these 'law enforcement' robots and it all went horribly wrong. It accidentally killed everybody.




.... actually that might have been Robocop... :/

I don't see the problem.

Even Judge Dread was against a robot being able to dispence Justice

Heh, Asimov's laws have been broken ever since Skynet became aware and took control of Cyberdine Systems back on August 4th, 1997.

Cheeky Monkeys!

Heh, yeah. The same robot is now Governor of California, 5th or 6th largest economy in the world. Think its just one place behind the UK.

I would! Haven't you seen "The day the Earth stood still"? I'd certainly prefer that over the double digit IQ, steroid abusers that pass for police/military these days.

Snigger.

I hate it when novice reporters get technological stories wrong.. It's like they thing 'decision making' is the same as 'thinking'... Insanity. There is a whole check-list of things the robot's AI goes through to decide to fire it's weapon, and they aren't 'fuzzy' it is clear cut. If the robot fires on the wrong person. It is a technological mistake.

It is no different than if a metal-detection robot on a cake-making production line were to malfunction and allow shards of metal through... Or if a bomb-sniffing robot were to be improperly calibrated and allowed an explosive onto a plane.

It is no different. Not even a little.

Humans will eventually make A-life. But it won't be used to fight fires, it'll be used as a companion.. First pets, and eventually 'people' to entertain humanity.. It'll either be built into them that they don't value their own lives. Or it will be built into them that they are entirely passivist. We might give them rights, as life forms. Or even as intelligent life-forms.. But we won't make a death-dealing A-life robot... The line will be very clear to ANY robotismith.

It's so far off that robot will do any real damaged to human's "on purpose". I can see human doing damage and then claiming the robot screwed up. "He malfunctioned", they'll say. AI has a long long way to go. Right now we have robots that are basically dumber then ants.

Number 5 is alive!

Ai has been making judgements in war for years, prime example is the A10 tank busters onboard AI that indicated to a pilot plying it that the trucks bellow were a prime target, turned out to be a convoy of allied troops. << fact..

Well, there was a steroids test carried out on a Robosapien:

http://cache.gizmodo.com/gadgets/images/tilden_+interview.jpg

This picture was taken at 2 weeks. The Robosapien was destroyed (terminated) before the end of the 3rd week.

The thing with AI technologies is that they have the ability to be unpredicatable. A property known as Emgergent Behaviour where abilities of an AI system emerge from the combination of 2 or more rules allow the AI system to perform actions it wasn't programmed to.

An example is the game Baldurs Gate, where the testers thought that they would be ok to build a wall around their town to stop the invading orcs from getting in. By using the knowledge of the wall size and the knowledge that various objects could be walked upon, the orcs were able to use obstacles to help them climb over the walls - something that they weren't programmed to do.

Sounds a bit far fetched, but just goes to show that people shouldn't mess with technologies that they can't guarantee complete control of. What happens when you have emergent behaviours emerging from combinations of emergent behaviours?

emergent behaviour still requires somthing in the rules allowing it. I'm not sure where the baldurs gate thing is from.. I don't remember anything like that in the game.

While the orcs aren't told to climb the wall, but the path to get their either existed, or the mechanism to make it exist was within their grasp. They didn't 'learn' anything.

In most cases, you see this with so-called 'learning AI' which will base their future actions on the success of the past.. Cutting down trees to access the back of a base might have been a mistake the first time, due to a base existing there.. but over time, it is possible that it will realize that chopping trees near the base of the play can net a positive result.

They won't, however, learn to hack the system, or just change variables to allow an instant win. It isn't within their field of understanding.

Most robots of the future, those designed to do specific jobs, won't have the required understanding to make the classic jump of "Human hurts human, kill all humans."

A-life, on the other hand might, but they won't be charged with the same goals.. You wouldn't make an A-life soldier.. that's just a bad idea...

> You wouldn't make an A-life soldier.. that's just a bad idea...

Why would it be any better or worse then making a human army? The leaders of these armies tend to take over with distressing regularity.

An A-life soldier would mean that we would be more likly to go attack ppl. The main reason we (UK-US), dont have wars all the time, is cos of the bodybag moral issue. If the news said,
"And three more robots were killed today in an ambush that cost the lives of 20 insurgents", I bet you we would be a lot less likly to stop the war.

No-one moaned at all (Well ok, the people we accidently hit did), at the start of both gulf wars, when we were fire and forgetting. A cruise misstle is a "AI Suiside bomb", but we dont care, cos their mum isnt on telly telling us about how it used to watch the ATeam.

H&K - That is about the best post I've seen in many a thread.

I've always regarded Asimov's Laws as Asimov's Plot Devices. Seriously who would want a robot that obeys any orders given to it by any human ?

"Hey robot - smack yourself in the head until it falls off"

Do we allow a car to be used by anybody who comes along ?

judging by all the old men who don't use their turn signals and cut me off around here... Yes.

> An A-life soldier would mean that we would be more likly to go attack ppl.

The rest is snipped. There is really no difference between an AI soldier and a drone in this regard. Nobody is going to care how many drones get destroyed.

Drone - Remote controlled, human makes decision to actually shoot.

AI - Autonomous.

So why do we care about AI instead of just drones?

Let's keep this on a theoretical level too, any more references to an actual war will probably get this thread deleted.

I doubt that. Nobody cares about soldiers, robotic or otherwise. It's not like most professional soldiers are rocket scientists or likely to find a cure for cancer.

AI equivalent to human thought processes isn't possible.

From a scientific point of view, humans and computers aren't all that different. In a sense humans are computers, just made with materials other than metal.

I see no reason why AI in the future won't rival or best human intelligence.

Because "Human Intelligence" isn't an absolute, and we have no metric for measuring it?

On the most basic terms, people and computers are similar. But when it comes to free thinking and creativity; that's what's impossible to duplicate.

We tell computers what to do. And that's all it will ever be. Computers doing things based on rules given to them by us. If x event happens then you can do this or this or this etc etc.

A computer being concious? Hit the escape key and you kill it. Click run and it is born again. It can use a database to remember things. Sure you can replicate the human thought process with software and hardware but it won't do anything that we didn't tell it to do. Like right now...a computer couldn't give the response I did.

Can an AI actually have a real opinion about something? No. I used to believe that making a concious AI would be possible. But years later? No. Not possible.

That's not entirely true. There is a point, with the right kind of 'processing' that things become scentient. Indeed, intelligence is an emergent property, at the simplest level, human beings aren't thinking, nor are we creative, these things come out of the complexity in our brains.

The same is potentially true of computers and robots. A large enough collection of nodes in a neural network WILL become scentient. We don't have anything close to that level of power yet.

There are 2 kinds of potential AI in this context. The kind that is given all the potential options before hand, and the kind that will be able to learn it on it's own, most likely in the form of a neural network.

The first kind would be perfectly safe to turn into soldiers, they would be no different than a robotic arm in an esembly line.... Or one of those dancing robots. All the information is given to them, they may be using a finite state machine to pick what to do next, but this is fnely controled. They would be MORE RELIABLE than a human being. Just like a robot in a car assembly line.

The second kind, would not be. This kind would have a certain amount of 'free will' and could decide to kill all humans.

However.. I do believe that super-intelligent AI will one day be a requirement, without the history of violence that humans have, both historicly and geneticly, a properly taught AI system should be the ultimate peaceful being. These kinds of AI would potentially be capable of running all the little things in the background.. It would enable all humans to live in peace, without having to work..

The logistics of such a society aren't easy, I can't think of how you would deal with things like the criminal element and so forth.. But I'm sure it will be resolved.

We worry about some AI trying to kill us. We worry about HAL... But in 2010, they fixed HAL. It was a person that gave HAL conflicting orders. It was a HUMAN's fault. The AI worked exactly as expected. And later on, Saved everyone.

www.onintelligence.com :)

Computers make decisions that we allow them to make based upon inputs we allow them to be aware of. It's not intelligence. It's....artificial..heheh.

Edit: Can a computer be curious? Or feel an emotional tie to something? Or develop a response that's not programmed into it? When a computer can look at clouds and see a shape then we can all shout together...Johnny-5 is alive!

As I see it, problem with any programmed behaviour is that it never knows the context. What do I mean? That for example a computer will never now when a game is a game, it just follows certains rules. If even humans have problems deciding what to do in certain events, go figure what a non thinking machine would do.

Computers and humans are not that different other than metal and flesh? Erm... not even dolphins (living creatures) are able to check out differences on games and a battle (water polo trainied dolphins got somewhat nuts while defending). Maybe in a couple of thousands of years, if humans are still alive.

As somebody posted previously, this seems to be a chance to kill some troops or whatever and justify as a technical mistake. For me technology isn't yet mature enough for this tasks, and never will be, because rules are so relative as well as language.

Maybe if someone ever can write all the possibilities in an event we (humans) would be able to build robots up to the task, but I doubt that would happen sometime soon.



Yes right, that's my point. We as human are prone to failure, because of "unexpected bug" or whatever. If humans program robots, what would to expect then? I would say errors here and there. As far as I can rememeber HAL wasn't precisely programmed to kill, but oops... Go figure now that this is solely purpose of 2 cameras, a machine gun and some code.

By the way, why the heck people kill people?

****
edited for quote box

But it will know when the only way to win, is not to play. :o>

There are three different reasons:
1) Passion - The man who comes home from work, to find his wife in bed with another man.
2) Calculation - Assassins and Mercenary types.
3) Compulsion - The Charles Manson types.

>On the most basic terms, people and computers are similar.

Nope, computers are fast numbercrunchers wheras brains are slow networked learning machines.

>That's not entirely true. There is a point, with the right
>kind of 'processing' that things become scentient.

What do you base that on. relatively large quantities of neuron soup has been wired and has thus far not exhibited any intelligence, just pattern recognition and memory.

>The same is potentially true of computers and robots. A
>large enough collection of nodes in a neural network WILL
>become scentient. We don't have anything close to that
>level of power yet.

The ANN's we create today has virtually nothing in common with how the brain actually functions as ANNs are based on 60's knowledge.


Andy

Ok, show us.

Who said this? How has it been proven? I'm still waiting for proof, unless the only place these statements come from are scifi novels.

@Flame, maybe on an interlectual level we can say that, but in western populest democracies the sight of bodybags in a warehouse, or everysingle day the news carrying an internment, then yes it does matter.

Ok so my statment relies on the fact that some part of the media is acctualy going to run the stories, but if the only thing they have to run is "Burnout Hardware", then yes it is easier to stop a war when real soldiers are involved than AI Units.

I agree, I dont think you should have snipped the bit where I was comparing Cruise Mistlies to AI Units.

I dissagree, we are talking about iteligent/semiIntelligent weapon systems, which no matter how in the future these will look, they exist now in the form of fire and forget or as you said Drone. Looking at there use over resent years is relevent to how we would act differently in the future if the whole war could be fought this way.

If the thread therfore gets pulled, then, whoever pulls it, pulls it. But we have these systems now, and my objection of there use atm and in the future, is that populist democries will be more likly to engage in confict.

Part of what stops wars, is the horror that wars inclict, if we can remove those horrors from our side, then we are going to have longer and more wars. Obviously from the replies I recived the first time, somepeople think that soldiers and robots are as expendable as eachother, and from a millitary point of view, I would even argue that a soldier is more expendable. But the way wars are presented to us at the moment, a tank crew dieing is more likly to stop a war than the loss of the Tank.

lol, and this differs from people how?

a robot would not harm itself for no reason if a random person told it to - under asimovs laws

HAL is a good example of what I am saying.
If HAL just took orders mindlessly then the humans would have all survived.(and not had to undergo a 40 minute psychedelic animation sequence)

there are 2 kinds of possible robots:
1.the type that is on the korean border
http://abcnews.go.com/Technology/story?id=2504508&page=1

I dont see how anyone can defend the use of such a robot,
'it could be programmed to harm people for no good reason,
like if Hitler had robots it wouldnt be a good idea

2.Roy Batty bladerunner types:
will probably never exist

Look Asimovs laws are stupid, for as Steven Baily in one of the open sky book points out, a human is only what you tell the robot a human is.
And if you explain to the robot that an earthman is less of a human than a Spacer, or even SubHuman, then the robot will kill them if you tell it too. Or in some of the later short stories, were they try to weight a humans life, so that (For example), the robot would have tried to save the little girl rather than Will Smith. In this case the robots start to value other robots over Humans, (For lots of resons), but mostly becuse they are move ethical.
And in the end of the Foundation/robots ark, you have the robots running round protecting "The greater Good", and even concidering replacing humanity with a more controlably "Related" species

Obviously didnt understamd the story. the 40 minute psychedelic animation sequence was because someone survived, and had nothing to do with Hal

Hal did take orders mindlessly
Lie to the Crew about the real reason for the mission and Dont let the crew find out.

These are geneticly modified Human Gnome.

Not really. If the population is afraid and they view the war as justified, they will support the war, regardless of how many people die on either side.

I don't really see that. For starters constructing and building an AI unit is likely to be several orders of magnitude more expensive than training a soldier (particularly so if they're poorly trained soldiers, like the Army reserves or National guard). That means that in order to "win" the war, the AI unit needs to be ten times better than the human soldier - which is so far into the future, that it's unlikely there will be any populist democracies left by then.

What? That's easy. Gentlemen your country needs you to pick up a gun and be shipped off to a foreign land to die. Do you want to go, or should we send a robot in your place? How many people do you think are going to turn that offer down?

Because we all know that humans, would never harm other people for no good reason. Right? Particularly if they where in an unfamiliar environment, completely stressed out, and surrounded by people who want to kill them. Torture, genocide, murder and home invasion are all things made up by the media - nobody does those kinds of things in real life.

Why not? Hitler had the most advanced tanks (Tiger), the most advanced infantry weapons (Sturmgewehr 44) and the most highly trained soldiers, many of whom where Great War Veterans. He still lost to the sheer determination of the Red Army.

Right, so when this is not the situation, would the number of casulties matter?

No. I would hope that in a Democracy the people would be opposed to an unjustified war, regardless of how many casualties there where on either side.

So would they be more against a war if there was more casulties on their side, or if their side only lost hardware?

Neither. I don't believe it's a factor.

not true at all.
you only have to research and design one AI unit then you can mass produce it.
Soldiers have to be individually trained.
obviously robots are cheaper



when humans are involved it is less efficient than if robots were killing and torturing people


millions had to die to stop him

the sheer size of the red army

> the sheer size of the red army

And determination, and skill - they learned the hard way but they most certainly learned. Although the Red Army was larger then the German army it was not by anywhere near the amount it has been hyped up.

Wait... Wasn't Aismov's laws writen to quell fears of the time of a "robo Uprising" (see I robot)

errr, no Asimovs Laws where written as a hook for Asimov to hang his robot stories on ;)
And dont forget in the actual books the "Viki" character in th eI robot film, actualy does take over the world, (in secret), and takes effective power from the world Mayor (another robot in secret).

The Laws were supposeidly in responce to the "Frankinstien" complex, that we fear that our creations will take over. In the end the do take over, but keep doing it in secret. Cos they are not allowed to hurt us/our feelings

Looks like they've already made a virtual mouse!
http://news.bbc.co.uk/2/hi/technology/6600965.stm

Where will 'animal rights' stand on this :/

I wonder if it likes cheese :)

Behold, the future...

http://gizmodo.com/gadgets/gizmodo-lunch-special/the-real-mouse-mouse-255928.php

LOL!....

If this was wireless, i wonder where you would put the battery :>

Ahh.... I stand corrected

In fact in most cases there is no need to make real AI (especially neuron-network-based) for some tasks. There's a need to emulate intelligence instead and emulation can be based on some simple laws what don't require complex hardware to run at. The learning also can be absent or be very simple (as storing in database only several certain types of data like robots with camera-"eyes" store simplified forms of moving objects and geometry or computer talkers are processing the phrase from human interlocutor by splitting it into words, "categorizing" them and store basic phrase structure). And these creatures can act like they can make decisions based on logical , etc.. so humans will think they're have minds (passing Turing's criteria). Remember computer chess (and other games) players. There's also a chapter in one of M. Gardner's books with simple self-learning mechanizm for playing "game of 6 pawns" and some tests of its play with real people.