Should real robots obey Asimov's Laws?

[Nate the Great]

This is a fairly simple poll. Ever since Asimov created the Laws, scifi authors and fans everywhere have jumped on them as being "definative" or "mandatory" for modern robotics. However, we all know that such codified directives are hardly required. So, the question is, should we add them to robots when/if we get to the point that there are robots "smart" enough to understand and obey them?

Yeah, this is a topic designed for controversy, but I think that Doctor Who has been hogging the spotlight in this forum long enough.

[e of pi]

Are the Laws a good concept? Yes. Are they a good set of basic guidlines? Yeah. Are they perfect or the best? No. They're too simple, as seen in I, Robot and other areas, to allow full trust. More practical, but less explicitly statable or imaginable laws will be more likely.

[mudshark]

A requirement? No. They may not be preferable, or even practical, for that matter, but I think that the general notion of "laws of robotics" bears consideration.

Even though Asimov coined the word (in 1940, according to one dictionary), he would have been the first to point out that he himself was not a roboticist. He reasoned, though, that robots, beyond a certain level of sophistication, could not be programmed with every eventuality anticipated -- it was far too complicated -- and that they would need to have some set of rules for working out, independently, just what to do in an unforeseen circumstance. They needed to be able to think something through, and needed to know where the lines were... what was in-bounds and what was not.

His Three (or four) Laws of Robotics fulfilled a literary function. I think it likely that real roboticists will have to design in -- perhaps are beginning to do so already -- a comparable set of rules that, while they may bear little or no actual resemblance to Asimov's, will fulfill the practical equivalent of that function.

[Hejira]

I think if we get to the point where AI really is I, then morals should be taught to them, as they are to humans.

Yes, we have scumbag humans, but that doesn't stop us making more humans.

[MaverickZer0]

No. Setting guidelines like that will only eliminate the reaches of the AI. If you want to limit them, do so physically--give them only the strength of an adult human, say--rather than mentally by set parameters. They should learn morals like anyone else. And if they go psycho, we can lock them away in a capsule for 30 years with ethical sims running in their head.

But the Three Laws, when you look at them, are not practical. What if one human is in danger from another one, and you cannot shield the one in danger? The only option is to diasble the attacker, but under Asimov's three (though not the fourth) the robot would freeze. And then, what if many robots were in danger from a human, a fanatic or something? What then? Or many animals?

You can't cut everything down so simply. Asimov's Laws work well for Asimov's writing. In the real world, we'd need different limits; hence limiting physical strength. I hesitate to suggest we limit their intelligence as well, because that would be what they were for and unfair, but possibly that would only be when the robot seemed to be a danger.

[Nate the Great]

Uh, even human strength can be quite dangerous when coupled with the proper weapon and intelligence behind it.

I think that the key here is the question "how many tasks that we would want robots to do really require anything even remotely close to human intelligence?" I can't think of one. Hence any programmed moral code would be unnecessary.

As for the robot that can't shield the human, it's simple enough to add "if you can't protect one human without hurting another, just do nothing."

[MaverickZer0]

Agreed. But if not given a reason to do so, a large intellect will not rebel, knowing nothing could be accomplished. Robots are less likely to warmonger than humans, not more.

In which case, you're once again limiting an AI's capabilities. Limiting their possibilities in a way you could not do with a human. I am talking about possibly-sentient robots here, not factory-line workers that need only guidelines.

Limiting again. It's also saying to the human in danger, if you don't care about the robot, 'tough luck, it's just your time, though there is someone who could get you out of it right there'.

[Nate the Great]

I really will try not to tighten the range of my original question.

Oh, and how do people feel about Asimo and his friends? I saw a demonstration of him at the U of M, and I was curious.

[mudshark]

A nice toy that shows off their recent advances in mobility mechanics, as the name indicates.

[Nate the Great]

I particularly found the face-recognition feature interesting. Imagine being able to connect a basic topographic map with a name, and at several different angles at that.

[Chancellor Valium]

What do you mean, 'even' human strength?!

Of course human strength is bloody dangerous!

Sorry.

It all depends, I think on two questions:

1) Do robots count as self-aware, and/or as living?
2) Do we have the right to control what they think?

And right now, I'm not even going to pretend to know the answer. I will say this though:

Why do we *need* robots that are so advanced that such laws would be necessary? Has anyone thought about the economic consequences? What about the moral questions? Finally, what about the Marvin Problem?

[Nate the Great]

The only reason I can think of for "thinking" robots is to use as soldiers, and even then, I'm dubious. I'll cut down my rant on current military strategy to simply this: why have large numbers of ground troops AND weapons of mass destruction? For that matter, what are we accomplishing by having armies of robots attack each other? A waste of resources, no question.

[Gatac]

You've strafed one of my pet topics. Take cover, for I will now whip out a small treatise on the subject.

You ask why we have soldiers and WMDs. This is a bad question, because it assumes that war is only about the quickest and cheapest way to kill the most people. This is a common misconception. War is about defeating the enemy, specifically crushing his military might. While this can be served by killing his soldiers, this is often inefficient and hard to pull off, because humans are surprisingly sturdy. It is not so much our ability to directly withstand damage but to survive, to dig in, to keep your head down and spring up again. In a (relatively) recent example, Soviet forces on the Eastern front in WW2 were regularly taken by surprise when they advanced, because German troops knew exactly how to dig in. Even after hours of artillery bombardment, they'd still be alive - and ready to fight.

Simultanously, using WMDs has serious drawbacks. To start with, only nuclear weapons have any degree of proven effectiveness - biological and chemical weapons as they exist today are not controllable enough (biological) or just plain ineffective (chemical). Gruesome, yes, but frankly not worth the effort. Nuclear warfare doctrine takes a book to fully analyze, but the basic problems are this: Nukes are indiscriminate. You can train soldiers to kill other soldiers only, and for the most part that works out okay, but nukes kill everyone, and lots. Then, you have lingering, widespread effects, not to speak of the image penalty you take once you start tossing nukes. The danger of a nuclear free-for-all can not be understated, and even though steps have been taken to mitigate this, you can be sure that "nuclear deterance" is really just a nicer way of saying "mutually assured destruction".

As if that wasn't bad enough, we're increasingly getting boggled down in 4th Generation assymmetric warfare. You can't nuke terrorists. We need people on the ground, and lots of them. Simultanously, conventional wars are fought all the time, and the combined arms tactics used today are mind-bogglingly complex. To answer your question, I pull out the word "flexibility". If all you have is a hammer, every problem starts looking like a nail. A military force can have a forte, sure, but these days Air, Sea and Ground forces work together so closely that not being up to snuff in one area leaves you seriously vulnerable. We have to have all these toys because the other guys have them, too. You're sitting on a twisted form of the Prisoner's dilemma: Rationally, we all want peace. But we can't be vulnerable against possible attacks from X - who, by extension, can't be vulnerable against possible attacks from us.

"You put your gun down first!" "No, you!"

---

Now, how can robots help make war? First off, I don't expect the introduction of AI into warfare to shake things up quickly. I don't want to come off as arrogant, but even the smartest AI tank won't replace ten normal ones. A lot of the problems we deal with today are actually rather independant of the soldiers - communications and the fog of war, supply lines, the laws of physics. Could a robot tank possibly make shots faster and more accurate, figure out a better way over rough terrain or conduct longer missions because it doesn't need rest? Possibly, but these are incremental improvements and likely won't even show up in the first few generations of military AI, if they even stick around and work on it long enough to get something effective going. Humans are very smart machines. It'll be a while before we can build better ones even after we create something that is an artificially intelligent lifeform. Expect dogs and children, not digital Einsteins.

One often-cited aspect is human endurance. Robot tanks don't sleep! Robot jets can pull more Gs! Certainly true, but tanks today are rarely limited by crew endurance. They need fuel and maintenance as well as ammo, and the logistics trail it takes to provide that maybe gives you five instead of four tanks you can supply with the same amount of cargo capacity if you cut out the soldiers. Possibly more if you can also leave the maintenance and resupply to other robots and cut out the humans completely. Similarly, an AI jet could certainly pull more Gs, but we're also limited by the materials we have. Flying is extremely complicated, too. (Look how much work *human* pilots have to put in.) I do think that there's a lot of potential here, though.

---

So, what does all my rambling mean? (Certainly not coherent, more stream of consciousness, but oh well...) No big robot armies. Robots are going to be a very helpful resource - for human soldiers. Drones, powered armor, remote-control artillery fire - that's where we're going, with improved communications and recon as well as augmenting infantry soldiers. AIs could probably help us design much better gear, though.

And in 400 years, we'll have Bolos. Yay.

Gatac

[Zeke]

Excellent post, Gatac. I feel much smarter now. Just one thing I'd take issue with:

Originally Posted by Gatac

You're sitting on a twisted form of the Prisoner's dilemma: Rationally, we all want peace.

Unfortunately, that's just not something we can assume. Western civilization has mostly moved beyond the idea of waging war just to gain power (a very recent development -- it certainly wasn't true of Hitler). We're basically reactive now; even Russia and China are no longer into conquest, though that can always change. But right now we're facing groups like al Qaeda and people like Ahmedinejad who are very much active. They want Israel wiped off the map and Islamic rule extended. It doesn't count as wanting peace if the prerequisite is killing everybody else.

(I'm not breaking my own rule and trying to start a political debate, btw. These guys are on record about what they want, so repeating it is politically neutral.)

Originally Posted by Infinite Improbability

For that matter, what are we accomplishing by having armies of robots attack each other? A waste of resources, no question.

Uh... don't you mean what would we be accomplishing?

[Chancellor Valium]

Russia and China are not into conquest by military means.

When you're sitting comparatively a lot closer, you'd be surprised by how effective their current conquest of Europe is going, and they haven't had to lose a single soldier in doing so.

[Nate the Great]

I guess the "what are we accomplishing" is referring to the hypothetical robots waging war in all of our minds while we have this discussion.

I have a nice long rant all formed in my mind, but that would be stepping on somebody's toes. Suffice to say that I'm amazed that we've been letting other people take advantage of American ethics for this long. The "needs of the many" mentality is perfectly valid for me.

[Nate the Great]

I guess the "what are we accomplishing" is referring to the hypothetical robots waging war in all of our minds while we have this discussion.

Three times I've tried coming up with a response of my political opinions, but I don't want to get burned again. E-mail if you're really interested.

[MaverickZer0]

'Even' human strength compared to what strength it is possible for machines to have. Either way, a sentient machine isn't going to automatically try and attack us all, laws or no.

[Gatac]

I should specify: Rationally, we don't want to fight. This equates to peace insofar as few people - who are, arguably, mentally disturbed - would get into a fight for no reason. (Note that "Those are my orders" counts as reason; also, people do not behave rationally under a variety of conditions such as extreme emotional states or intoxication.)

But generally, humans don't fight for no reason.

The thing about wars is that only rarely do the people who actually start these wars fight in them. This mitigates their feeling of personal risk - it's easier to make other people risk *their* lives for your goals. Hence, a rational human could start a war, trading lives for power or resources, but I do not think that makes him evil or sick - he's just disconnected from the consequences of his actions.

Try to teach a kid that he shouldn't break windows when you never punish him for that.

Gatac

[Nate the Great]

If we start having robots fight our battles, how long will it be before we get to the point where even building the robots is wasteful of resources, and we get war computers and disintegration chambers?