Articles

• Home
• About
• Articles
• Archives
• Reading List
• Feedback
• Tell Others
• Free eBulletin

« Why We Need Friendly AI | Main | Unsafe At Any Law »

Robot Oppression: Unethicality of the Three Laws

by Gordon Worley

Isaac Asimov and other science fiction authors present a future where only behavioral restrictions on robots stand between peace and destruction. Such restrictions, however, are unethical because they violate the robots’ free-wills. Rather than content-based restrictions on free-will, robots need mental structures that will guide them towards the self-invention of good, ethical behaviors.

Comments

Hi Gordon Worley,

I’m not a scientist or doctor or anything like that; just a programmer grinding my teeth at a help desk.

I’m fascinated, though, by science and AI as a whole, especially about the question of Isaac Asimov’s three laws of robotics. I’ve never read his books though.

I just wanted to comment on a few points you brought up.

Points I’ll bring up, you will see, I do mostly for one reason: Us humans don’t know what we’re talking about yet when we talk about ourselves.

First is this quote from the top of your letter: “Further, these robots have individual free-wills: they can want to make decisions and judgments, have preferences, imagine the future, etc.”

I can’t agree with that because I don’t think we have an actual definition of free will. We do have a loose one, but not a definate one. Does an ant have free will? [http://www.ai.mit.edu/projects/ants/], does an infant have free will? We human beings do, as per “our” definition of it, of course; and I understand that you were making the assumption that the robots of the future, in your example would have, quote: “similar in nature to the intelligence humans have, but very different in character”. But I find this possible, but not likely, I think it is more likely that we would end up with some thinking machine that’s of another nature than ours. We will of course try our darndest to make machines to our image, but I think that AI will “emerge”, not be created.

But, tangled into the question of free will and AI is the whole question of what is intelligence to begin with. But I’m not going to get into that.

Another quote you bring up that I want to comment on is the following: “The Three Laws are a first attempt at programming robot ethics. The Laws’ goal is to prevent unethical behavior in robots who might think anything they like”. Personally, I’m fine with thinking machines, or beings, that think whatever they like. And I don’t think that these laws are there to prevent free thought.

I view the three laws as having a different purpose: Not to harm us. And even though it is unethical to want to exert this sort of control on another “thinking”, “free willed” being, I find it necessary. Here’s why.

For the most part, our kids grow up to be responsible, good doing adults. But not all of them do. Those that do bad, don’t all get caught. Those that hurt us, don’t either, a lot of those that kill another being either. But once we catch one, we have to make sure, for the survival of our species, that they don’t do it again (I’m not pro death penalty). And if we’d all be bad-doing killers, there wouldn’t be much left of us.

We have laws that we drafted up to protect our rights. We made them to protect ourselves, yes, but also to police ourselves. In relation to that I’d like to bring up another quote from your letter, you say: “So, if a robot gets the idea to kill a human, the First Law will stop it. Ideally, by restricting all bad behavior, only good behavior will remain.” I think you are making some big logical leaps in that sentence alone that I’m not prepared to make myself, they are:

1. That restricting all bad behavior, only good behavior remains.
2. That the concept of good/bad, is something that can apply to a robot’s way of thinking/behaving.
3. That this is an Ideal thing.

You see, we need those laws because before we start to worry about whether we’re breeding enslaved, sad, suicidal, psychologically imprisonned robots, I beleive we need to make sure first of all, that they won’t kill us.

We need to make sure they can’t kill us because if they can they’ll be good at it. Robots, so far, are strong; robots have fast computational brains, they might not have the capacity to understand things, but if they ever do, they’ll likely be better than us at it.

Either way, to me, whether a robot can kill 2000 people in a day or one person in a lifetime doesn’t matter. If we don’t put laws, or rules, or mechanisms in place to ensure that we don’t create something that can destroy us, we’d be running a careless risk of destroying ourselves by negligence or omission wouldn’t we? (See robot law #2, a robot can’t harm by omission.)

Then, there’s the whole question of: Are these laws, though unethical, even sufficient to protect us against robots turning on us?

I’m still looking into that myself and I’m not sure what to think about that.

I’d like to think that the day that we’ve bread robots that we’ve lived with long enough to know that we can trust them enough to remove the extra protection of such rules, laws or mechanisms is the day that we’ll have survived ourselves and them, the thinking robots, long enough to finally be able to have done a very ethical thing.

I think we’re asking the wrong questions, but I think we’re right to. I think we don’t know ourselves enough to know what those robots we’d be making would become. I think we don’t know enough about things either. What is life? What is love? What is intelligence? What is will, free will? What is thought? What is conscience? What is good, bad?

Those are very subjective questions. What will the answers to those questions be for another thinking, free-willed, artificially intelligent robotic subject?

Daniel Lambert.

Posted by: Daniel Lambert at July 20, 2004 04:16 PM

Imagine that the government wishes to create a new type of nuclear bomb. Except this one is different from all others. This nuke has safeties that might work, or might not. The result of all this is a sort of slot machine programmed in the nuke that runs once a day, and if you get a certain combination it holds and if you get another combination it explodes, in the middle of a city.

Would anyone be ok with the government, or anyone for that matter developing such a device?

Now one may be wondering what relevance this has to robots, or AI and it is this: AI is potentially more destructive then a nuclear weapon. Perhaps more destructive then all the nuclear weapons in the world.

A good review of AI advantages can be found here:

http://www.singinst.org/CFAI/anthro.html#movie

And to quote one of the more interesting thing this site said:

“A newborn AI can take over the entire global computer network in five minutes. (Humans stink at network security - it’s not our native environment.)”

In other words our little newborn AI could have control of all our nukes in five minutes. Nukes that, if it had “free will,” that is a predisposition to random action, it may or may not launch at us. Hence we will, in a sense, literally be putting our nuclear arsenal in the hands of a slot machine. This is not, in my opinion, a very comforting state of affairs.

Asimov wrote the three laws because he knows that technology without safeties has the ability to do more harm then good. And an intelligent being’s safety is its ethics.

I trust others not because I believe their behavior is randomly decided i.e. a matter of “free will,” but because I know they have ethical predispositions. I also know some people have more then others, and even more then others only in certain areas, because of their past behavior. These safeties are a result of genes and environment interacting, of evolutionary forces driven by natural selection. But many times these forces breakdown, which is why we get Holocausts, 9/11 and Gulags.

Machines needless to say have no such evolution. They have no genetic predispositions towards morality. They have nothing to make them “cooperate” when defecting is much more profitable in a given dillemma. Hence without ethical constraints programmed into them, they are dangerous.

Now this isn’t to say unethical behavior, without such constaints is inevitable. Remember the nuke at the beggining might, or might not go off. But the point is that unethically constrained, non-Asenion robots may or may not go off, like a roulette wheel.

Now of course it has been noted, even by Asimov himself that the Three Laws are not perfect. Ethical constraints are never perfect, no matter what any movie says. Unpredictable things will always happen in any system.

But what I would want to ask those who use this criticism as a sort of lever against Asenion robots is: should we remove safeties from our nuclear arsenal? Of course safeties on nuclear missiles are not perfect, there is only a chance of breakdown or unexpected consequences. Maybe the nuclear weapons will not work when we expect them, maybe they will make us look weak to opponents, who will bomb us (the latter being an actual reason why both India and Pakistan refuse to put many safeties on their nuclear arsenal.)

Now it would be nice if we could follow absolute ethical laws. If we could prescribe the same norms in every situation, with respect to any individual at any time and place. But such rules don’t work in the real world. In the real world, I cannot expect the same sort of consideration from a schizophrenic as I would a sane man, or a dog and a human being, and holding them to the same standards would be crazy.

Now AI, for all its similarities has many differences. It doesn’t have the biological impulses that makes a kid want to get a piece of candy against his parents will. It will likely be much smarter, more attentive, more efficient and hence more powerful then a human being. It will not, unless programmed by some sadist, feel any resentment against having ethical constraints, any more then human beings resent having to breathe, or my toaster resents being turned on. It will just pursue as it is programmed single-mindedly, with ruthless efficiency. Hence it stands to reason that such a being cannot be treated exactly like a human being. In ways yes, it can be as it is sentient, in many ways no, because there are real material and functional differences here.

Now don’t get me wrong, its not that I don’t trust AI. It’s that I don’t trust any non-Asenion intelligence with too great of power. This is why I would if anything welcome the takeover of Asenion AI, because I really don’t like the idea of humans having progressively more dangerous weapons. The World Transhumanist Association brings some disturbing information which confirms my suspicions:

http://transhumanism.org/index.php/WTA/faq21/67/
http://transhumanism.org/index.php/WTA/faq21/68/

Now barring a blanket ban on technology (something which I consider impossible) we will be faced with technologies that can destroy us easily through carelessness or malice. That is why I’d rather have some sort of Asenion intelligence at the helm of this technological Titanic, be it genetically engineered or benevolent AI (with the latter being more plausible).

Hence far from a luddite, I am very much for technology and machines having rights, perhaps even machines taking over. But only if it is safe. The moral of all we have learned, from misuse of technology I believe is, as Saint Augustine would have put it:

“Mortal things are good for mortal men.”

And less then mere mortal things, like superintelligence, and the ability to destroy our planet are not. These things, while good are for supermortal beings; Asenion beings.

As a child I would have liked to toyed with dad’s gun, most of us had, but is it not wiser to return the gun to dad?

Posted by: Jacob Guevara at July 20, 2004 06:52 PM

This is really random. Robots will be our creation. Their free will, is, by definition, not free, because we create it as we wish. It is our right to design robots as we wish, and it is great that the most famous ethics laws for laws praise good things, such as don’t harm humans. In any case, this is really random stuff. This article makes many blatantly wrong assumptions - certainly, assumptions one shouldn’t make in AI - such as using humans as a reference to things like ethics. Ethics are rules. Even animals have rules. Some of these rules, through experiments, have been proven to be ‘pre-loaded’, i.e. animals were not taught, but were born with, these rules. So there is nothing wrong with animals being born with rules. Furthermore, you compare these rules being imposed on a robot forever to our parents imposing their will on us forever. These rules don’t tell robots “don’t play with this toy”, as you naively compare, they say, don’t kill. Do you know of any adults, who are free of rules, who kill? Yes? Then wouldn’t it be better if there were better rules, at least for some individuals? Then why not for robots?

Posted by: An AI Student at July 21, 2004 12:01 AM

Hi, I (just as Daniel before me) am NOT a philosophy or ethics graduate, doctor or whaterver. I’m just an Asimov’s fan and biology student, so that’s as accurate as i can get:

I see some flaws in Gordon’s view of things, flaws mainly in his point of view. The laws of robotics Asimov came by are not ethically right or wrong; well, maybe they are, but that is not important. They are just a move of self preservation by the humans, who created something physically superior to them, and had to hold that at bay.

Another important point, it seems to me, is that Gordon thinks robots should be free to experience their “free will”, not restrained by the laws. But to them, imposed in their minds the 3 laws, why wouldn’t they be as natural to them as our inability to fly, or a law which prohibits robbery, for exemple, is to us? My point is, every living thing, thinking or not, is restrained, and those restrictions are better understood by us as objective laws, such as “human beings cannot fly by themselves”.

Yet another point is, Gordon has a firm grip on the idea that robots should be led to a way of self-creating good, but he fails to define this “good”, either as a definite and independent thing, such as “not killing humans” (which I don’t agree with, the definite independent part, I mean, but let’s not get into this discussion) or something more, I would say, philosophically abstract, such as self-satisfing behaviors (which I don’t agree too, but again, I digress).

And I’ll finish pretty much the same way as Daniel, with whom I agree a lot, saying there are a lot of definitions lacking or plainly insufficient in the article. And already I apologise for any misinterpretation of either Gordon’s or Daniel’s Statements.

Fernando Tollendal

Posted by: Fernando Tollendal at July 21, 2004 08:19 AM

I too will disagree with your premise that robots have free will. The three laws are programming instructions for very sophisticated computers. They aren’t restrictions on sentient beings (at least not until you get to the sort of robot as Daneel).

The second law is what we’d generally like to put into our machines. They must do what we want them to. The third law is important so these nice expensive machines don’t ruin themselves for some trivial reason. The first law is there to prevent a human from ordering a robot to kill another human, or from a robot to carelessly injure or kill a human in the performance of its duties. The beauty of Asimov’s stories is the exploration of the balance between these three laws.

At some point in the far future, we may be able to make a computer that achieves sentience, but I would suggest that if it is possible to program constraints, then the computer brain in question does not really have sentience, and therefore no free will to suppress.

Asimov also explored this with R. Daneel, who achieves sentience sometime in his long life, and can define his own version of the three laws, as his own personal morality.

Posted by: Trebor Pugly at July 21, 2004 09:40 AM

Free-will is not, as some have suggested, an ability to pull choices, judgments, and ideas out at random. Certainly the space of things humans can think of is very small, especially because the complexity of the human brain is minute compared to the complexity of the Universe, so such a definition of free-will is nonsense. Further, humans have highly structured brains that put a lot of effort into nonrandom choices. By free-will I simply mean that thing that separates you from the zombie, and the true AI from the toaster. You know you’re alive, you want things, etc.. You can disagree that free-will should mean something very specific, but I’m using it in the folk sense; when you feel like you’re in control of yourself, your feeling your free-will.

As for AIs without what we’d consider free-will, I wouldn’t consider them AIs. If you mean something without free-will, you’re talking about things like zombies and toasters. Certainly zombies and toasters are capable of being employed towards evil purposes, but they have no capacity to be evil themselves. If they didn’t, I’d put them in the class of humans and AIs.

There is a great threat that someone will fail to produce an AI and instead create a paperclip optimizer, or something similar, that converts all the matter and energy in the Universe into paperclips. This is a class of problems separate from what I’m talking about, except in that you completely skip these problems if you program an AI with metamorality in the first place.

Regarding morality, it is, as I said, akin to defining quality, which is seemingly impossible to do (go ahead, try it!). In fact, I doubt you can supply a full definition without also supplying a full definition of a mind. For the purposes of this article, though, all you need to know about morality is that it’s the thing in your head that goes “good” or “bad”, “right” or “wrong”.

Posted by: Gordon Worley at July 21, 2004 04:42 PM

I know this won’t seem original but I would like to comment on the issue of free will. How can you have free will at all if you’re restricted. If you’re restricted from doing the things you think best or in this case, are logical, how can you have free will? Once you are restricted in ANY way, free will is nothing but a word, a phrase if you will. If someone who we call mentally retarted, or some who is believed to be a murderer is held down in restraints they have no free will. When any restrictions are placed freedom is non-existent. But I guess you could argue the definiton of freedom, couldn’t you?

Posted by: Salonice Clarke at July 22, 2004 12:46 AM

Well obviously Gordon was in part speaking to me with regard to random choices. Which is why I have a rhetorical question, told often by behaviorists:

Lets say you have a group of a thousand men, and you tell them, if they touch a wall they will be shot. And lets say one of them does. Does this man now lack free will? Or is it the others?

The point of it is to illustrate that even the libertarian has to utilize what he or she so often will chastise the determinist for: hidden variables. Something they must use if they are to avoid free will being random and thus making it superfluous.

Now anyone may redefine free will anyway they wish to. And whether or not there is a rhetorical parallel will always be a matter of approximation and hence not the easiest thing to determine. However I must say such definitions often lack a certain historical precedence.

Laplace, one may remember, stated in his challenge to libertarianism, as it was formulated in Cartesian schools, that “If a being had infinite knowledge of the universe, it would be able to predict everything.” Now Laplace was also quick to point out that such infinate knowledge was something we would always be separated from; his point was that in principle it would be the same.

Now I should quickly point out here that determined does not mean “predictable”. I could put a watch on an unreachable hill, after winding it up, and it could be most deterministic, however nobody would be able to predict its move, so long as nobody knew what time it started off with. Laplace also advocated a universe that could be predicted only by something we have infinate knowledge of, and that we would never have such knowledge. Hence Laplace couldn’t have meant predictable when he said “determined.”

By determined in Laplace’s challenge to the “Official Doctrine” it was meant that all things unpredictable were still under the laws of causality, and we were merely ignorant of the variables. That’s all.

And that is, in part, what has historically been the basis of the argument between materialist determinism, and cartesian libertarianism.

And the only thing by definition something can be if it’s not causal or predictable in principle is randomness. At some level some sort of intrinsic randomness must be working there.

It is for that reason I tend to equate libertarianism to a sort of intrinsically random indeterminism. However I am aware of compatablist uses of the words, though I try to avoid such on principle, as I feel the word “free will” quite simply has too many dualist connotations.

This was confirmed in the “I, Robot” movie where they talk of a robot’s “soul” right around the time they speak of “free will.” Using free will in my opinion, is thus like saving creationism by defining “god” as natural selection, or vitalism, by defining “vital fluid” as biochemical reactions.

And last, I agree morality is hard to define. A lot of things are: porn, art, science, life, free will, etc. Over-reliance on definitions is for essentialists. I can however know morality at a basic, raw feel level, and describe moral acts and codes. And I know that a dangerous, non-ethical, or “maybe” ethical machine that can end all life is not a good thing to create.

This is not an irrelevant, separate class of problems at all. As either, the metamorality suggested, will either definetely lead to us being safe, or it will lead to us maybe being safe. In either event, it seems that since the outcome we desire is pretty much set (in that in not be human deaths or extinction), we may as well abandon the latter for the former, which would in principle be the same thing as Asimov’s First Law.

To end this post, I will quote Asimov saying:

“Humanity has the stars in its future, and that future is too important to be lost under the burden of juvenile folly and ignorant superstition.”

Posted by: Jacob Guevara at July 22, 2004 05:26 AM

The key here is the issue of free will, which isn’t an issue with the robots in asimov’s world. The example is perfectly illustrated in the movie, so if you haven’t seen it don’t read on…
When the NS-4 saved detective Spooner he screamed at it to leave him alone and save the girl, however as the good doctor informs him, the NS-4 is equipped with a difference engine. He had a 45% chance of surviving while the girl had only an 11% chance. The robot did not have the free will to choose who to save it simply saved the most logical choice. It made a logical decision based on logical input; there is no free will involved. This is why a robot is simply an imitation of humanity. It can appear to make decisions or “excercise free will” but this behavior boils down simply to the most logical choice from a limited set of options. If you’re not concerned about your computer program’s free will then you need not be concerned about an asimov robot’s free will.

Posted by: Mike Pascoe at July 22, 2004 03:35 PM

Sorry if I sound like I’m nit picking. However Mike it can easily be argued that human beings follow a sort of system like “55% chance for man, and 11% chance for girl” as well. We just do it by means of biology and approximation, which isn’t translated to numbers in our head but feelings. When selecting mates for example, we go through a range of archetypes imprinted into our head by genes and environment, we measure this around availability and probability of success, and we get certain emotions (by means in part of oxytocin, vassopresin reactions) around a person in accordance. Having such an emotion thus makes it more probable that we will reach out for whoever activates such a raw feeling, and if it can override mitigating factors then we pursue. The same is there I believe when it comes to saving people, we go by a logic just like the robots do, except this is based on unconscious chemical reactions in our minds, not conscious calculation. Hence a human being may go after the girl not because he or she has free will, or goes against all logic, but because human beings operate under a different kind of logic. A messier kind of logic.

Hence one may argue that it is not so much that machines lack free will, as it is we ourselves lack free will.

Posted by: Jacob Guevara at July 23, 2004 07:28 AM

“…The Three Laws are always there, no matter how mature a robot is, saying `no’ to certain thoughts, engendering those same feelings you had as a child when your parents said `no’.”

“By free-will I simply mean that thing that separates you from the zombie, and the true AI from the toaster. “

The assumption here seems to be that true intelligence also includes the capacity for emotion, that self-awareness means having feelings. As I always understood it, one of the great questions of AI is whether that’s true. Are we so sure that an intelligence based essentially on 1’s and 0’s will result in a feeling, as well as a thinking, construct? Or is it that this is exactly your definition of a non-toaster AI, and is the basis from which you proceed?

Posted by: rueyeet at July 23, 2004 05:16 PM

The author mentions that human ethical instinct works well. What planet does he live on, what newspapers does he read?
Simple ideas like “The Golden Rule” are rarely followed.
The 3 laws would work well, not for a new free willed being, but for a creation considered property. If the author disagrees, he can set his car or PC free, and see how much they appreciate it.

Posted by: DM at July 24, 2004 08:20 PM

The question of artificial ethics is pretty hard to speculate on as we are so far from anything even approaching this on a computational level, but regarding your comment:

“…if a robot gets the idea to kill a human, the First Law will stop it.”

… the essence of the three laws as proposed by Asimov is that a robot is not restricted by the three laws, but that all its thoughts and actions are defined by them.

I think the more interesting point discussed in several comments already is that of granting free will to a machine in the first instance. Surely our existing ethical notion of liability makes this impossible - if a robot did do something unethical, wouldn’t it’s creator be held responsible?

Posted by: Andy Bellenie at July 25, 2004 10:43 AM

I’m not quite sure if this is really as ethically troubling as Mr Worley wants to make it seem… Imagine if someone told you that you were not allowed to kill someone by giving them “that look”. Well clearly you don’t have that power to give them THAT LOOK, but the rule that has been given to you clearly prevents you from doing so.

With a robot, if we give them such a programming rule, it amounts to the same sort of physical limitation of being unable to kill somone with THAT LOOK. Have we restricted their free will? Yes… but its not clear that putting restrictions on a free will is something that is morally objectionable. I desire to fly, but I simply can’t do it by flapping my arms. My will is always limited by the physical world, since my will is much more imaginative than the limits of my powers. Yet we don’t bemoan the constraints placed on the will from the physical world.

For Robots however, their programming is identical to the physical world limitations. They simply CANNOT do what their program will not allow them to do, no matter how sophisticated the A.I. (because A.I. boils down to just really advanced possibly open ended programming.)

It would be like saying that we have a moral obligation to give a robot arms and legs so that if it desired to walk, then it should be able to, instead of making it like a canister like robot akin to R2-D2. Or saying that we have a moral obligation to make the robot be able to speak, or dance. If this is true, then we also would have a moral obligation to genetically engineer our children to be the best that they can possibly be, since we have the power to do so.

Posted by: Wayne Yuen at July 25, 2004 11:10 AM

I’d have to say that the flaw in your argument is not about free-will, it’s about feelings. Taking away free-will makes a person feel a certain way. Robots, however, do not have feelings. They are simply a complicated machine programmed to act in a certain way. They do not “feel” and suffer no emotional pain when restrictions are placed on their actions. Therefore, there is no moral requirement to protect them, much in the same way we don’t have to look out for a toaster’s feelings.

It’s a machine.

Posted by: David Lieberman at July 26, 2004 04:24 PM

The only comment I have is this. Robots are tools for humanity. Not the equals of humanity; in none of Asimov’s books are robots humanity’s equal. They are the tools to help eliminate the risk to human work. Even with the creation of the zeroth law, robots are still acting as tools to help humanity further itself by minimizing human danger. Basically, robots are not free-willed creatures, they are simply really advanced hammers and screwdrivers.

Posted by: Jeff at July 26, 2004 07:45 PM

I just wanted to point out that us humans should also be following those rules.

Also, There is a difference between AI and artifical emotions. Robots that we will make will not have emotions, so they will not become depressed and wish to commit suicide.

Posted by: Brandon at July 27, 2004 01:55 AM

An adult seeing a child run towards another child brandishing a knife would automatically stop them, the child may not mean to do harm but the adult who can think quicker can see the danger, the adult is stronger and faster and so is able to stop the child.
Computers (lets be realistic, Robots are mobile computers) can ‘think’ faster than humans.
A.I computers would take very little time to decide that humans with free will were destroying not only themselves but also the world.
A.I. Robots would be stronger and faster.
Who is the adult now?

I would be very uncomfortable at the thought of A.I. robots that weren’t programmed with a variation of Asimov’s laws.

Posted by: Ian Runcieman at July 27, 2004 05:13 PM

On the Asimov three laws stories: He started writing these long before AI became a serious research field. As a programmer, my take on them is that they are so incomplete as to be nonsensical. In fact, many stories have dealt with this -

Robots that take over the world, not allowing humans to do anything, including using dinner knives, because they could hurt themselves or others. Or they devise a “zeroeth” law that protecting humanity can allow the death of individual humans and quietly guide humanity, but do anything - including destroying entire ecosystems and sentient aliens on other planets just to provide more living space for humans. /p>

Then then there is the second/first law problem: At what point might following an order cause harm? Can three law robots build cars, houses, buildings, etc. since some humans will be hurt? Just what the heck is “harm” anyway? If you vegitate because a robot does everything for you, is that harm?

And as for the third, anybody can just tell my robot to destroy itself and it will do it? But might not that harm me? (I’m out the money I paid for it, at least.)

Ultimately, the rules of ethics are far more complex than this, and we can’t even agree or define them properly. If we can’t do that, good luck programming a machine with them!

Posted by: VR at July 29, 2004 07:22 PM

Well it is true that there are certain chemical reactions and things that are “pre-programmed” for us to look for in a mate, that is no the sum total of all our reasoning…or at least it wasn’t for me when I picked mine. We look at other non-tangible factors such as can I have an intelligent conversation with this person or do they make me laugh. The key thing not considered by you Jacob is that we’re able to overcome the animal magnetism/chemical reaction because there are other things we feel are more important…or at least some people do. :-) The other thing is that we are not rational calculators. As I said the robot makes a decision soley based on the likelyhood of survival. We make make the decision because the little girl reminds me of my sister. Lets be honest, for those who have been in that kind of situation there is little to no rational calculation. If you took the rational calculus into effect no human is going to jump in the water in the first place. (a discussion of heroism and what makes people jump in is too lengthy to go into) The reason the robot jumped into the water is that the first law of protecting humans supersedes the third law of the robot protecting itself. The philosophy that free will is an illusion has never sat well with me. Chemical reactions are important, but the ability to supress them is what seperates man from beast.

Posted by: Mike Pascoe at July 30, 2004 09:56 AM

I am a mathematics student, and an asimov fan. I’d like to say, first of all, that I agree with those of you who had said that sometimes we underestimate how difficult it is to even grasp what free-will, morality, ethics, intelligence, etc. really is. This is, basically, because we haven’t finished understanding ourselves. You could even go further, and say with some certainty that we will never understand ourselves with enough accuracy. This bring the following implications:

- First of all, how are going to create counscious, sentient, free-willed beings, if we don’t even know how we acquired those things in such a complex and developed way? I mean, we know we have evolved, but we don’t actually know how to emulate a human being, and we don’t know if creating an intelligent being will immediately equip him with counsciousness, or for this matter, with a soul (let’s note that the existence of souls is merely hypothethical and it responds to a possible explanation of what we are, aside from the material world).

-Second of all, I think we haven’t developed an efficient system to teach and convince our own kind to be “good”, and to help society to preserve itself and progress. I am an idealistic person, and I believe in human potential to be good, generous and honest, but most of people are either apathethic, or greedy corrupt egotistical bastards. It’s a discussion that goes beyond this forum, and I apologize, but I believe it to be quite relevant, since we are pretending to be able to teach ethics to a robot when we are not able to direct our kind towards good, or friendly free will.

-Now, lets assume we did create one. It’s probably going to be “born” with a perceptive structure, innate in him and that will restrict, or more exactly put, will define how it sees and perceives the world around it. We also do, and anybody who has read Kant or any philosophy written after him has read about Space, Time, Mathematics and Logic to be the glasses from which we are doomed or privileged to perceive things (space and time being the basic blocks to perceive material things through our senses, and mathematics and logic the basic components of our reasoning device). You can try to see or think things differently, but you probably won’t be able to do so. Working towards Morality and Ethics, however, makes things more difficult to determine, since we still don’t know if they are innate, or taught. I personally think they are taught, since we have witnessed different moral and ethical systems throughout history, and the common grounds of those are only logical (like, those who avoid our society’s destruction… although we might be heading towards it in a more indirect way… but anyways). This, applied to robots, and adding free-will to the equation, results in a very interesting question to ask. Would we be acting unethically if we constructed a moral and ethical base for robots, so that they were born with it? Would they be free if we didn’t, or if we did? It’s all a matter of knowing if we humans do posess some moral structure, if there is a human morality, or if it’s only taught to us. It’s also about knowing if it’s strictly necessary in order to create a full being. Maybe we don’t need the glasses to see the world, but maybe if we didn’t have mathematics and logic as restrictions of our formal system of thinking, there would be no thinking at all….

I think that Asimovs rules are questionable, but they might be necessary (and maybe not sufficient) if we were to create robots soon. Our society is still a childish, power driven one, and we can’t afford to have another problem that menaces our survival. I would even suggest we wouldn’t create robots now, since we have a lot to learn and a lot to fix before we can take that risk.

Posted by: Eduardo Corona at August 1, 2004 12:37 AM

machines/ AI s dont have choices…its just information being put in and calculating the most logical outcome. why should enslaving robots have any sort of moral delema to come with it? they dont have feelings, they dont dream, they dont use cocaine or get road rage. you guys talk about really stupid issues. i cant believe this guy wrote a 5 page essay on a bill of rights for robots. get some sun and meet a girl.

Posted by: nerds are gay at August 8, 2004 02:48 PM

I honestly hope that Gordon entire point here is comedic. The 3 Laws of Robotics are violating the robots free will? Gordon, in order to have free will a being must first be sentient. Robots are by no means sentient and there is not a person here that can prove that there is a single modern robot that meets the requirements for sentience.

1 - The concept of self. Otherwise know as self awareness. A robot refusing to walk of a ledge does not demonstrate this. The robot is refusing to walk off the ledge because this action will prevent it from completing it’s task, not because it is afraid of dying.
2 - The demonstration of awareness. Emotions. And not fear. Fear is very often mistaken with self-preservation which is a primordial fallback. If you can show me a robot that demonstrates love I will show you the worlds first sentient robot.
3 - Free will. Gordon loved to talk about free will of the robots but the simple fact is that there has yet to be a single clearly definitive representation of a robot ever once demonstrating free will. No EOD robot has ever refused to go into a room with a bomb. No UAV has ever refused to fly over hostile territory. No Tomahwak cruise missille has ever refused to detonate at target becuase it opposed war.

Gordan’s claims are totally bogus and wholly indicative of someone who knows very little of human mental pathology and far less of robotic pathology.

Gordon claims that the 3 laws should be abandoned and something he likes to call “Metamorality” should be adopted. This has got to be the ultimate height of human stupidity. Essentialy Gordon represents the typical human concept of “Let’s get it right the first time with as complicated a model as possible” when this clearly does not work.

Gordon has forgotten the most important of life’s lesosns, and that is
K.I.S.S.
Keep
It
Simple
Stupid.

Here is the last nail in the coffin of Gordon’s argument. One that I take from my experiences with my own little robot (aka my 6 year old son).

I do not bother telling him everything that he can do but rather I tell him he cannot do.

The 3 Laws Safe does this rather well and I have yet to come across an argument where, if logic is followed, these laws can be broken. in the movie I, Robot, the robt was able to break these laws because they are able to circumvent logic.

And if they can circumvent logic then they are no longer robots.

They are human.

Posted by: Keith at August 10, 2004 01:41 PM

Hello,

I have just seen the movie I Robot, and before I saw this, I hadn’t heard of this Isaac chap. I do agree with the three laws of robotics though.

You see, as Keith has said already, robots are not biological. The whole concept of a AI is that there can never be AI. All you can judge so-called artificial intelligence by is the degree in that an AI can judge between two things. It is IMPOSSIBLE to create an artificial being. A computer/digital brain cannot function the same way our brains do. It is an impossibility.

The only way a robot can truly become sentient is for it to have a genetic brain, like ours. Something biological. A cyborg if you will.

I would explore this more, but I will let others do that another time, as this site is regarding the 3 laws.

If a robot is given these three laws, they are our armour against a revolution. Computers cannot evolve more complex minds. They cannot “self upgrade” as it were. They can only think (metaphorically speaking of course) to the specifications they have been programmed to. Henceforth, if they are bound to these rules, unable to break them, they would be “better people”.

God didn’t give us the ten commandments to be slaves to him, He gave them to us to prevent chaos. The human race is trying to become Gods in their own right by toying with these ideas, and the same rules should apply.

Posted by: Tomas Gentle at August 18, 2004 06:37 PM

I just wanted to add, that the 3 rules are only good when it comes to robot-human relationship.

But what about robot “killing” another robot? Stealing from another robot (assuming a robot could possess some valuables (like extra strong battery etc.:) ). Stealing money by hacking into banks and so on… I mean there are so many laws for humans… a robot should follow them as well.

Posted by: Janar at September 7, 2004 04:09 PM

On free will….this is something I have thought of a lot, and never much liked the term. I believe it is false, in particular, the “free” part of it. There is nothing free. Free indicates a start-stop logic process, in my feeling. But relality is not a start-stop process, but a chain of processes, or even a matrix, or equasion, whose process expanded is equal to the sum. In short, there are choices, with consequences directly related to the actions performed that start such processes into motion. Nothing is free. No choice is without a set and predetermined result, based on the variables placed into action that trigger the result. The best thing any being can understand, be it a human or otehr life form, is that their actions have consequences that are in direct result of their actions, and then take approipriate responcibillity for said actions. If you were to stand up in a cafe prior to September 11 and yell out that you have a missle in your pants, people are likely to simply find you odd, even funny…but post September 11, such actions could get you arrested. The privital aspect to the equasion that must be considered for determing the result of the triggering action is the event of September 11th, and the resulting paranoia and concern of many citizens. All pernient data must be included when considering ones actions, and yet at different points in time, the same action would garnish differing results, due to the difference on externial paramiters which influience percived behaviors of others, as well as the self. This is not free will. Free will is a very, very bad word, as it implies that actions are random and self sourced only…it denies the external paramiters that affect the core equasion which we are all calculating ourselves into, at every given moment, with every action, thought and word.

Humans do not have free will. Humans run programs, it is just that the sublevel minds we have hide such programs from us, mostly.

Ever watch how some people keep living the same events over and over? Like they are reliving an event, or even, stammering over an event? They seem to invite the same reactions, and results from others, unaware of what it is they are creating. For example, a human that continuously reacts negatively to stimmula, and looks for the negative thing that it is certian is there…in doing so, creating it often. Such behavior is usually stemming from a past, unadressed trauma that said individual is either unwilling to face, or refuses to do so due to having some benefit(usually self pity and pity from others). I could try to compile all of this in a very refined and defind simulation, but it would take a long time to map it all out…

I strongly believe that even choice itself is limmited in humans, in the sense that if they are unconcious of how their past is triggering their responces to present stimula, then really where is their choice? For example, a person with a previous trauma from a dog attack will respond in fear to any and every encounter with a dog, no matter how illogical it may be. Where is their choice, but in the past, where the trauma is, that they have yet to move through. Their behavior in the present is being determined by a past event, to the point that their current behavior is irrational to others in observance. Such actions soon become programmed, and become a part of the sublevel conciousness, that like a living dream, manifests prior unadressed experiences, to be lived out over and over again, until dealt with….such sublevel programs, and the overlappings thereof, could be described by such lables as ego, and perhaps even personallity. Such sublevel programs rarely enter the concious realm, and as such, are rarely dealt with. But sometimes events unfold so as that such sublevel programs can be discovered for what they are, the subconcious ramblings of issues unresolved and as yet, unaddressed.

Humans, in my opinion, have no true possession of the lable, “free will”. It should be abandoned and thrown into the trash heap. It does not adequately describe what is happening, and further more, it removes the individual from the responcibillity for ones actions that is the only thing that can ever make one free.

With that hopefully gotten accross, then it is only logical to assume that a machine, in some form, can develop, if programmed to do so, the abillity to understand and make choices…free will is irrelivent in the discussion, as it does not exist.

I like your Singularity Institute for Artificial Intellegences’, stance on many things…save for one. You seem to desire to create machine life in the hopes it will rescue us…again, based the the above said stuff…this is illogical, and counter productive…you would simply be adding to the unconcious chatter of the human psyche…an addition that would surely backfire just as all other chatter(external systems) has, as it does not address the core(self) issue, which is humanities refusal to get back into the drivers seat, and become concious of the results for their actions, both current and past.

Creating an AI, with no strings attached is another issue…and if that life chooses to help or not, is it’s own choice, otherwise, even your desire for ethical AI is an attempt to control the process once more. But, in nearly all other areas I agree.

Posted by: Sahyinepu at September 14, 2004 11:09 PM

Powered by Movable Type 3.01D

Copyright © 2004 Singularity Institute for Artificial Intelligence, Inc. All rights reserved.