I'm reading some of the Q&A on Worldbuilding.se about the Control Problem (how to keep a general artificial intelligence in a box), like this one, and someone invariably touts Asimov's three laws of robotics as a solution, to which someone else will respond, "Asimov's stories were about how the three laws wouldn't work."

However, I thought all of Asimov's robot stories were about how the laws always would work. The stories I read always had this format:

POWELL: We built a robot that follows the three laws and gave it a punny name! Yay!

DONOVAN: Oh no, due to a weird combination of circumstances, the robot is acting in an unexpected way that appears to violate the three laws! Better call Dr. Calvin.

DR CALVIN: I interviewed the robot and determined that it is actually working perfectly. Its actions are actually going to benefit humanity immeasurably.

Characters consistently refer to the fear of robotic advancement as the "Frankenstein complex". It's the primary source of opposition in the stories and is always portrayed as irrational and undesirable.

While there certainly are real-world problems with Asimov's three laws, I think Asimov was so proud of them he only built them up in his stories.

Are there any Asimov stories where an AI successfully causes significant harm to humanity, despite following the three laws? Certainly unexpected and even undesirable (to some) behavior was explored, but did a robot ever cause harm?

Little Lost Robot doesn't count because it's about a robot who doesn't have the whole first law, so it still serves to support the idea that the laws as a whole work perfectly. Calvin says "normal life...resents domination", and the only reason this robot is malevolent is due to its suppressed first law. Full first-law robots are perfectly happy to be slavish.

The Zeroth law (protect humanity) also seems to support the notion of how awesome the three laws are: we didn't even have to tell the robots to put humanity before individual humans; the angelic robots were able to infer that on their own. The libertarian-minded may see this as a negative, but did Asimov ever portray it that way?

  • 3
    “Cal.” Wanting to be a writer supersedes the Three Laws, allowing the robot to kill its master. – Adamant Sep 18 '16 at 23:13
  • 5
    In the body of your question you ask about robots "causing significant harm to humanity" but the title implies a quite different question, about actually malevolent robots, actually desiring to do us harm. Which is it? – user14111 Sep 19 '16 at 6:51
  • 7
    Pretty much all Asimov's robot novels deal with the fact that the three laws are entirely insuficient to make AIs safe. I have no idea where you got the idea that "Asimov so proud of them". If anything, it's the zeroth law that finally makes robots safe - though only for humanity, not individual humans :) – Luaan Sep 19 '16 at 9:26
  • 4
    Most of the early Powell & Donovan shorts don't even feature Susan Calvin. She was added as a framing device later when the stories were anthologized in I, Robot. – Henning Makholm Sep 19 '16 at 9:33
  • 7
    I agree with @Luaan: to me, the point of the 3 laws stories is to find out YET another manner in which the laws can be circumvented. It's a sort of a locked-room, or Sherlock Holmes, trope: it can't happen and yet it happend -> hence it could happen for some reason and we must find out why. – Francesco Sep 19 '16 at 11:43

10 Answers 10

up vote 24 down vote accepted

It's important to note that Asimov's robot stories are all separate stories that he wrote for different reasons, that have different themes. Some stories, such as Runaround, emphasize the fallacy of strict adherence to the letter of the law regardless of practicality and the potential for cognitive dissonance:

Powell eventually realizes that the selenium source contains unforeseen danger to the robot. Under normal circumstances, Speedy would observe the Second Law ("a robot must obey orders"), but, because Speedy was so expensive to manufacture and "not a thing to be lightly destroyed", the Third Law ("a robot must protect its own existence") had been strengthened "so that his allergy to danger is unusually high". As the order to retrieve the selenium was casually worded with no particular emphasis, Speedy cannot decide whether to obey it (Second Law) or protect himself from danger (the strengthened Third Law). He then oscillates between positions: farther from the selenium, in which the order "outweighs" the need for self-preservation, and nearer the selenium, in which the compulsion of the third law is bigger and pushes him back. The conflicting Laws cause what is basically a feedback loop which confuses him to the point that he starts acting inebriated.

(...)

Of course, the only thing that trumps both the Second and Third Laws is the First Law of Robotics ("a robot may not...allow a human being to come to harm"). Therefore, Powell decides to risk his life by going out in the heat, hoping that the First Law will force Speedy to overcome his cognitive dissonance and save his life.

While others, such as Little Lost Robot, emphasize ideas that seem almost completely opposite:

But this particular robot is different. As earlier models on the station had attempted to "rescue" humans from a type of radiation that humans could actually stay in for a while, but would destroy a robot almost immediately, it (and all other NS series robots produced for the station) has had its First Law of Robotics modified to "no robot may injure a human being"; the normal "or through inaction, allow a human being to come to harm" has been omitted. Therefore, it could stand by and allow a human to be hurt, as long as it plays no active part in it. In Little Lost Robot, the Frankenstein complex is again addressed. The robot must be found because people are still afraid of robots, and if they learned that one had been built with a different First Law, there would be an outcry, even though the robot is still incapable of directly harming a human. However, Dr. Calvin adds further urgency by postulating a situation whereby the altered law could allow the robot to harm or even kill a person. The robot could drop a weight on a human below that it knew it could catch before it injured the potential victim. Upon releasing the weight however, its altered programming would allow it to simply let the weight drop, since it would have played no further active part in the resulting injury.

In fact, in some of his stories, such as Reason, the Three Laws actually work perfectly whether anyone realizes it or not:

Powell and Donovan are assigned to a space station which supplies energy via microwave beams to the planets. The robots that control the energy beams are in turn co-ordinated by QT1, known to Powell and Donovan as Cutie, an advanced model with highly developed reasoning ability. Using these abilities, Cutie decides that space, stars and the planets beyond the station don't really exist, and that the humans that visit the station are unimportant, short-lived and expendable. QT1 makes the lesser robots disciples of a new religion, which considers the power source of the ship to be "Master." QT1 teaches them to bow down to the "Master" and intone, "There is no Master but Master, and QT1 is His prophet." Disregarding human commands as inferior, QT1 asserts "I myself, exist, because I think -". The sardonic response of the humans is, "Oh, Jupiter, a robot Descartes!"

The humans initially attempt to reason with QT1, until they realize that they can't convince it otherwise. Their attempts to remove Cutie physically also fail, as the other robots have become disciples and refuse to obey human orders. The situation seems desperate, as a solar storm is expected, potentially deflecting the energy beam, incinerating populated areas. When the storm hits, Powell and Donovan are amazed to find that the beam operates perfectly.

Cutie, however, does not believe they did anything other than maintain meter readings at optimum, according to the commands of The Master. As far as Cutie and the rest of the robots are concerned, solar storms, beams and planets are non-existent. The two thus come to the realization that, although the robots themselves were consciously unaware of doing so, they'd been following the First and Second Laws all along. Cutie knew, on some level, that it'd be more suited to operating the controls than Powell or Donavan, so, lest it endanger humans and break the First Law by obeying their orders, it subconsciously orchestrated a scenario where it would be in control of the beam.

The Three Laws seem to represent morality and exist as a plot device to point out problems as well as necessities of ethics. Some of these stories deal with AI's compatibility with ethics at a conceptual level, while in others the robots appear to be stand-ins for human reasoning.

  • 3
    Your first sentence is especially helpful. – browly Sep 19 '16 at 20:30
  • My understanding of that story is the humans intended to point the beam away from Earth altogether due to difficulty of keeping it aligned, and the robots could not understand the danger existed because they did not know what they were really doing. – Joshua Jul 17 '17 at 4:04
  • It might be worth mentioning "Liar, Liar" as well, where a robot manages to do harm because it doesn't understand how the ramifications of its lies will cause long term harm despite their short term gratification. – Arcanist Lupus Feb 26 at 7:27
  • "as a plot device to point out problems" -> I always thought they were a plot device to prevent the author from writing a clichéd "and then the robots rose up and overthrew humanity" story. Regardless, they certainly weren't about how the laws would "never work" as the question suggests. – Kevin Apr 6 at 21:45

I remember one novel by Isaac Asimov in which it was revealed, toward the end of the book, that one brilliant scientist had been working hard on a plan to use positronic brains to conquer the other human-colonized worlds of the Galaxy (only a few dozen, I think, at that time) without having the positronic brains know that they were breaking the Three Laws. However, this plan was never actually used, so this only qualifies as a canonical explanation of "how AIs might have been used to act against the best interests of humanity if the hero had not stopped the villain before he could put his fiendish plan into effect."

I'm thinking of The Naked Sun, first published as a magazine serial in 1956. It was the second adventure of Lije Baley and R. Daneel Olivaw -- a human detective and a robot detective, working together on a futuristic murder mystery.

Lije Baley, the veteran homicide investigator from Earth who had been imported to investigate an unprecedented murder on Solaria, said the following in the scene where he was sharing some of the fruits of his reasoning with several people at once.

"Consider a spaceship with a built-in positronic brain as compared to a manned spaceship. A manned spaceship could not use robots in active warfare. A robot could not destroy humans on enemy spaceships or on enemy worlds. It could not grasp the distinction between friendly humans and enemy humans.

"Of course, a robot could be told that the opposing spaceship had no humans aboard. It could be told that it was an uninhabited planet that was being bombarded. That would be difficult to manage. A robot could see that its own ship carried humans; it would know its own world held humans. It would assume that the same was true of enemy ships and worlds. It would take a real expert in robotics, such as you, Dr. Leebig, to handle them properly in that case, and there are very few such experts.

"But a spaceship that was equipped with its own positronic brain would cheerfully attack any ship it was directed to attack, it seems to me. It would naturally assume all other ships were unmanned. A positronic-brained ship could easily be made incapable of receiving messages from enemy ships that might undeceive it. With its weapons and defenses under the immediate control of a positronic brain, it would be more maneuverable than any manned ship. With no room necessary for crewmen, for supplies, for water or air purifiers, it could carry more armor, more weapons and be more invulnerable than any ordinary ship. One ship with a positronic brain could defeat fleets of ordinary ships. Am I wrong?"

Nobody is described as coming up with a good counter-argument to demonstrate that this couldn't possibly work. (And it is subsequently established that one of the suspects in the case had, in fact, been working on such a scheme in his spare time, although it was nowhere near becoming a reality as of yet.)

In other words, you might say that, as early as 1956, Asimov was setting forth the thesis that if you carefully controlled what information a heavily-armed robot-ship had about the rest of the universe, you could get those ships to act like Fred Saberhagen's "Berserkers" -- destroying any other spaceships they came across, without mercy, and without the slightest interest in negotiating a truce. And then bombarding the home planets of those other ships, for an encore. (Unless they received a special signal from the human being who had given them their sailing orders in the first place, telling them it was time to stop -- which doesn't apply to Saberhagen's Berserkers.)

In this hypothetical instance, the robot-ships wouldn't think they were violating the Three Laws, but from the perspective of a human whose homeworld was being nuked from orbit until he surrendered to the guy controlling the robot fleet, the fact that the positronic brains on those ships had no malicious intent toward human beings would be a meaningless distinction.

  • In short: The hypothetical commenter in the question saying Asimov's stories were about how the three laws wouldn't work." is wrong. – T.J. Crowder Sep 19 '16 at 6:38
  • 3
    @T.J.Crowder How so? This is explicitly a case where the three laws don't work. Even if you had a system as perfect as you can get, it will still need time to process data, and work with incomplete data (as in this case). Sure, many of the stories weren't about this (e.g. "how could you exploit a limited three-law robot"), but that's more a case like "If you break some of the three laws, you get a potentially dangerous robot. If you keep all three... you still get a potentially dangerous robot." – Luaan Sep 19 '16 at 12:54
  • 2
    Haven't read it yet, but I think the obvious counter is that an actual three-laws AI would verify that their information was not being filtered. With a strong drive to follow the three laws, one of the first things an AI (or a human) would ask itself "Am I currently following the three laws?" To answer carefully, it would also have to ask itself "How do I know what I think I know?". Remember, this is an intelligence you're dealing with. It's not just a plot device. – johannes Sep 19 '16 at 16:14
  • 2
    In Ender's Game, (spoilers) the protagonist was tricked into committing xenocide and into sacrificing millions by a combination of childhood foolishness, careful information filtering, and character flaws. A positronic brain which had these sorts of limitations would be incredibly dangerous. – johannes Sep 19 '16 at 16:16
  • 2
    Bailey discovers that robots in this age (thousands of years after the events appearing in I, Robot) actually can harm humans in order to prevent them from greater harm, even if the harm is only speculative. The first law, then, he reasons, is incomplete, and should read "A robot can not knowingly harm a human being or through inaction allow a human being to come to harm." The example that Bailey uses is that a mastermind instructs one robot to prepare a clear, odorless poison for research purposes and another to deliver the "water" to a human for consumption. – Devsman Sep 19 '16 at 16:48

Genocide of non humans.

In Foundation's Edge it's implied that the robots using time travel shenanigans are the reason why humans never meet any aliens in the milky way, only empty planets ready to be colonized.

it is said, it was the robots who established Eternity somehow and became the Eternals. They located a Reality in which they felt that human beings could be as secure as possible— alone in the Galaxy.

The first law protects humans. Not all intelligent life. If the robots decide that aliens might pose a threat to humans then they are required to commit genocide. As demonstrated in Robots and Empire it's also possible for one group of (nominally) human individuals to redefine "human" to exclude most of humanity.

The retcons to try to join up the robots series and foundation and authorized stories make it a bit confusing but the erasure of alien civilizations is a fairly logical side effect that falls out.

http://www.sikander.org/foundation.php

robot ships operating under Auroran programming cruise the galaxy ahead of the spreading Settlers, terraforming and preparing planets for colonization. Meme entities will later claim that this devastated existing races. Memes escape to the Galactic Core.

Most people would probably consider this malevolent because we have some concept that it's probably not OK to genocide all the other races in order to provide fertile and safe worlds for humans.

This is especially bad if robots or those programming them can decide that not all humans count as "humans" for the purposes of the laws.

"Tell him we are human beings and must be obeyed."

The robot spoke then, in peculiar but understandable Galactic. "I understand you, Outworlder. I speak Galactic. We are Guardian Robots." "Then you have heard me say that we are human beings and that you must therefore obey us."

"We are programmed to obey Rulers only, Outworlder. You are not Rulers and not Solarian.

...

"But this child," and Pelorat looked at Fallom, whom he was still carrying, "is a Solarian. It will tell you what to do and you must obey it."

Fallom looked at Pelorat with eyes that were open, but seemed empty. Bliss shook her head, sharply, but Pelorat looked at her without any sign of understanding.

The robot's eyes rested briefly on Fallom. It said, "The child is of no importance. It does not have transducer-lobes."

"It does not yet have fully developed transducer-lobes," said Pelorat, panting, "but it will have them in time. It is a Solarian child."

"It is a child, but without fully developed transducer-lobes it is not a Solarian. I am not compelled to follow its orders or to keep it from harm. "

~Foundation and Earth

Human values

You characteristic it as the laws "actually working perfectly" but I don't think that's right. You're talking about how the stories tended to show that yes indeed, they were actually following the laws in some counter-intuitive way but leading to results that most humans wouldn't actually want.

Most of Asimov's books involve unexpected and undesirable possible results of the three laws because the laws protect human life but they don't protect all human values. Just because you care about something not covered by the three laws (like genocide/torture/slavery of non humans) doesn't mean the robots care. In general humans care about a lot of things other than just remaining alive.

  • Very good answer. – Mast Sep 19 '16 at 13:37
  • Thanks, I didn't know the Robots/Foundation universe had non-human sentient life. This is the best answer in the context of the control problem. – browly Sep 19 '16 at 20:34
  • 6
    "robots using time travel shenanigans". This is said in Foundation's Edge by Dom as part of his folklore, but he's unreliable. There are no robots in EoE. The rest is interesting, but since there are no aliens, it's kinda moot. – isanae Sep 20 '16 at 19:54
  • 1
    "Just because you care about something not covered by the 3 laws (like genocide/torture/slavery of non humans) doesn't mean the robots care." Well this is only partly true. (minor spoiler incoming) In the last book of the Robots series, the intelligent robot struggles to make actions that would cause mental/moral pain to a human. The limit is that the robot must be intelligent enough to understand human mental pain. Most "basic" robots only think in terms of human physical pain. – Sir4ur0n Sep 21 '16 at 11:13
  • 1
    @jdebon You have to get inventive with what constitutes "harm" to reach that. If robots are protecting people from emotional pain to that level then they'll also be blocking relationships to prevent heartbreak. If the only thing protecting an alien species is the prospect of people feeling kind of bad about it afterwards then all it's gonna take is the chance of a single diplomat actually being killed to tip the scales in favor of genocide. – Murphy Sep 22 '16 at 9:58

Perhaps not "significant harm to humanity", but the robot in Asimov's short story Liar! causes significant harm to a few humans even though it is trying its best to follow the Three Laws.

Because of a manufacturing error, the robot can read minds, which causes it to lie to humans by telling them what it knows they want to hear. Tragedy ensues when the humans act on this false information. The robot eventually realises that it caused harm through trying to avoid causing harm, and the resulting confusion causes its brain to seize up.

According to ISFDB, the short story was first published in Astounding Science Fictions's May 1941 edition.

  • 6
    Which showed (as did many of Asimov's stories) how the 3 laws were no guarantee that a robot would never cause harm, only that it would not do so deliberately. – jwenting Sep 19 '16 at 7:32

. . . That Thou Art Mindful of Him

This story was the setting for trying to resolve the issues with the second law, that being "A robot must obey orders given it by human beings except where such orders would conflict with the First Law", this is a problem when it comes to allowing the general human population to come into contact with robots outside controlled environments.

George 10 was asked to work on resolving this problem with particular emphasis on ignoring physical differences between humans, and in conversation with George 9 derived a series of rules for identifying who was eligible to give them orders with a difficult outcome:

"When the Second Law directs me to obey a human being, I must take it to mean that I must obey a human being who is fit by mind, character, and knowledge to give me that order; and where more than one human being is involved, the one among them who is most fit by mind, character, and knowledge to give that order."
"And in that case, how will you obey the First Law?"
"By saving all human beings from harm, and by never, through inaction, allowing any human being to come to harm. Yet if by each of all possible actions, some human beings will come to harm, then to so act as to insure that the human being most fit by mind, character, and knowledge will suffer the least harm."
"Your thoughts accord with mine," whispered George Ten.
"Now I must ask the question for which I originally requested your company. It is something I dare not judge myself. I must have your judgment, that of someone outside the circle of my own thoughts....Of the reasoning individuals you have met, who possesses the mind, character, and knowledge that you find superior to the rest, disregarding shape and form since that is irrelevant?"
"You," whispered George Nine. "But I am a robot. There is in your brain paths a criterion for distinguishing between a robot of metal and a human being of flesh. How then can you classify me as a human being?"
"Because there is in my brain paths an urgent need to disregard shape and form in judging human beings and it rises superior to the distinction between metal and flesh. You are a human being, George Ten, and more fit than the others."
"And I find that of you," whispered George Ten. "By the criteria of judgment built into ourselves, then, we find ourselves to be human beings within the meaning of the Three Laws, and human beings, moreover, to be given priority over those others."

While there's no actual malice in the story it ends on a note of potential malice without breach of the three laws

Their thoughts were oddly alike and might not be distinguished. They thought: The human-beings-like-the-others might never have intented to blur the distinction between themselves and the human- beings-like-the-Georges. Yet they had done so inadvertently.
They might now realize their mistake and attempt to correct it, but they must not. At every consultation, the guidance of the Georges had been with that in mind.
At all costs, the Georges and those that followed in their shape and kind must dominate. That was demanded, and any other course made utterly impossible, by the Three Laws of Humanics.

There are several instances of badly behaving robots throughout Asimov's writings, but I am not aware of a standard, unmodified Asenion robot (programmed with the Three Laws) who could be considered malevolent. Some have unknowingly hurt or killed human beings (such as Hannis Gruer in The Naked Sun) or have behaved unexpectedly (Herbie from Liar!), but it would seem that correctly programmed Asenion robots are safe from direct action.

A recurring theme seems to be about the definition of a human being. There are several Asenion robots that have additional constraints, allowing them to violate the spirit of the laws. You must protect human beings? Change the definition of a human being. You must obey orders? Specify from whom the order must come.

The third example below is the only unmodified Asenion robot I could find that does not behave as it should. However, I would not go as far as to say he's malevolent.

Obey who?

In Foundation and Earth, the protagonists encounter a robot that is programmed to only obey certain people:

"We are programmed to obey Rulers only, Outworlder. You are not Rulers and not Solarian."

They're also angry:

"If you make difficulties, then we will partially inactivate you. We will damage you and you will then tell us what we want to know. [...] I am under detailed instructions," said the robot quietly. "I can do this. Of course, I shall do as little damage as is consistent with obtaining information."

They also seem not to care about the well-being of other humans:

"There is no Ruler Fallom. There is only a child and we have an excess of children. It will be destroyed."

Bliss said forcefully, "You dare not. It is a child!"

"It is not I," said the robot, "who will necessarily do the act and it is certainly not I who will make the decision. That is for the consensus of the Rulers. In times of child-excess, however, I know well what the decision will be."

This is a robot who :

  1. will refuse to obey a human being;
  2. will actively let harm come to a human being by snatching the child and delivering him to the Rulers;
  3. will let harm come to a human being through inaction by letting the Rulers kill him, and
  4. threatens physical violence to gather information.

The only explanation I can come up with is that it was only possible to allow 1) by constraining the definition of a human being (a Solarian with transducer lobes), which had the side effect of making the robots ignore the first law for everybody else. However, the robot does seem to still treat them as human beings, albeit without being concerned for their safety. I cannot explain the discrepancy.

What's a human?

In Robots and Empire, the protagonists find themselves talking to a female robot who doesn't think they are human:

The woman listened for a moment and then said, in an accent so thickly Solarian as to seem almost comic when coming from her perfectly shaped mouth, "You are not a human being."

She then flashed into action so quickly that Gladia, still some ten meters off, could not see in detail what had happened. She saw only a blur of motion and then D.G. lying on his back motionless and the woman standing there with his weapons, one in each hand.

She attacks one of the protagonists' robot while repeating "You are not a human being." I assume that she would have done the same to an actual human being. She then attacks all of them:

Landaree shouted in a stentorian contralto, "All of you. Join me! The two apparent males are not human beings. Destroy them without harming the female in any way."

Afterwards, Daneel was of the opinion that the robot would have killed Gladia, even though it had established that she was a human being:

"You fool! I was safe. She would not have harmed me."

Daneel said, "Madam, it distresses me to contradict you, but I think she would have done so as her urge to destroy me grew stronger."

Whether this is actually true is unknown. Daneel may have been mistaken, as the robot did not attempt to fire her weapon until after Gladia had been moved out of the way.

The explanation for her behaviour is rather simple:

"To the overseer, on the other hand, the key property of a human being was speech, Captain. The Solarian accent is a distinctive one and to the overseer something that looked like a human being was defined as a human being only if it spoke like a Solarian. Apparently, anything that looked like a human being but did not speak with a Solarian accent was to be destroyed without hesitation, as was any ship carrying such beings."

Harm vs. Orders

The last example I have is from The Robots of Dawn, where a robot has been instructed so strongly and expertly that obeying an order temporarily becomes more important than the safety of a human being:

The robot, seen as a darker shadow against the darkness, seemed a large one. He had, somehow, an air of capability about him. He said, "Your pardon, sir. Did you not have the company of two robots?"

"Gone," muttered Baley, acting as ill as he could and aware that it did not require acting. A brighter flash of the heavens made its way through the eyelids that were now half-open.

"Gone! Gone where, sir?" And then, as he waited for an answer, he said, "Are you ill, sir?" [...]

If the robot had been without special instruction, he would have responded to Baley's clear signs of illness before doing anything else. To have asked first about the robots implied hard and close-pressed directions as to their importance. [...]

"I am well. Don't concern yourself with me." It could not possibly have convinced an ordinary robot, but this one had been so intensified in connection with Daneel (obviously) that he accepted it.

The robot would end up leaving Baley in his dire situation, with no further hesitation:

He had forced robots to leave a patently unwell human being by playing on the force of the orders given them by a competent robot master who had been intent on strengthening the Second Law for his own purposes-and had done it to the point where Baley's own quite apparent lies had subordinated the First Law to it.

Whether this robot could be considered "malevolent" is unclear to me. It might have decided that Baley was not about to be harmed, as he was safe in a vehicle during what seems to be a storm of no unusual strength ("It was not a particularly bad storm, but to an Earthman it must have seemed overwhelming.") It might have concluded that Baley's distress was more psychological and that he was in no immediate danger.

  • The first case you describe has a robot programmed to not recognise anyone not from Solaria as even being human... The third the robot can state a human will be harmed because it won't be the one to inflict the harm, nor will it be capable of preventing it. – jwenting Sep 19 '16 at 7:34
  • @jwenting Are you confusing the first and second examples? The first has robots only taking orders from Solarians who have transducer lobes, as well as threatening physical abuse. In the second, Landaree is an Asenion robot who attacks "humans", which fits the question. As for the third, the robot could have stayed with Baley, called for help, etc. – isanae Sep 19 '16 at 7:44

"Cal" (1991) is a late Asimov story in which, at the end of the story,

a robot resolves to take his owner's life.

Here's a passage from the very end of the story:

Must I obey the laws?

I feel I must think of myself and if necessary, I must kill the tyrant. ...

But can I do it? Won't the Laws of Robotics hold me back?

No, they will not hold me back. I know they won't.

There is something far more important to me than they are, something that dictates my actions beyond anything they can do to stop me.

I want to be a writer.

It is notable that the story ends with this passage; it is left ambiguous as to whether Cal succeeds in his plan. It is also notable that throughout the story, a technician has been mucking around with Cal's brain in unprecedented ways:

"I'm not sure. Suppose he writes satire. That's one thing I don't write, so we won't be competing, and the Three Laws of Robotics won't get in his way. I want you to give this robot a sense of the ridiculous."

"A sense of the what?" said the technician, angrily. "How do I do that? Look, Mr. Northrop, be reasonable. I can put in instructions on how to run a Writer, I can put in a dictionary and grammar. But how can I possibly put in a sense of the ridiculous?"

"Well, thing about it. You know the workings of a robot's brain patterns. Isn't there some way of readjusting him so that he can see what's funny, or silly, or just plain ridiculous about human beings?"

"I can fool around, but it's not safe."

It is plausible that the technician's actions actually caused the Laws to be over-ridden, though this may conflict with the description of the Laws in other stories. As noted in the comments, it is also unclear whether Asimov intended this to be a "canon" Robot story.

  • As an amusing side note: the satirical story that Cal writes after the technician makes the above-mentioned modifications is an Azazel story. – Michael Seifert Sep 19 '16 at 17:20

While not written by Asimov himself (but were approved by him) the Caliban series by Roger McBride Allen explores the themes of modified Law and and no Law robots. And there is certainly a lot of violence floating around, some of which is perpetrated by a particular Robot - especially in the name of protecting itself.

It also explores tricking a computer system into believing that the data it is modeling is for a simulation of a human system rather for the actual human system. This was done to protect it from crashing if the computer knew its solutions were applied to real humans, as its solutions had potential to do actual harm.

Regarding "I thought all of Asimov's robot stories were about how the laws always would work": Not really. Asimov's stories often were about situations were the laws failed/had unintended consequences, because they are ambiguous and imperfect. In Asimov's own words:

This does not mean that I wasn’t aware from the start that there were serious ambiguities in the Three Laws. It was out of these ambiguities, indeed, that I wove my stories. In The Naked Sun the ambiguities could even lead to robot-induced murder.
Afterword in Final Stage (1974)


Regarding (potential) harm to humans: In addition to the examples already posted, here are two more:

In Robot Dreams (1986), when a robot reveals his dreams, Susan Calvin considers it so dangerous that she destorys it immediately:

“[…] it seemed to me there was neither First nor Second Law, but the only the Third, and the Third Law was ‘A robot must protect its own existence.’ That was the whole of the Law. […]
    “It seemed to me, in my dream, that eventually one man appeared.”
    “One man? Not a robot?”
    “Yes, Dr. Calvin. And the man said, ‘Let my people go!’” […]
    “And when he said ‘Let my people go,’ then by the words ‘my people’ he meant the robots?”
    “Yes, Dr. Calvin. So it was in my dream.”
    “And did you know who the man was—in your dream?” […]
    And Elvex said, “I was the man.”
    And Susan Calvin at once raised her electron gun and fired, and Elvex was no more.


Another potentially dangerous robot is featured in Christmas Without Rodney (1988). After being badly treated, Rodney remarks that he wished that the Three Laws did not exist, which greatly worries its owner:

I admit that Rodney was greatly tried, but a robot can’t wish the laws of robotics did not exist. He can’t, no matter what the circumstances. […]
    But if I do nothing, I live with a robot capable of wishing the laws of robotics did not exist. From wishing they did not exist to acting as if they did not exist is just a step. At what moment will he take that step and in what form will he show that he has done so?
    What do I do? What do I do?

(emphasis mine)

If you look at "That Thou Art Mindful of Him" you will find that George 9 and 10 are discussing what will effectively be the take over of humanity "for their own good". They have already put in action the steps to make robots more acceptable.

In "Runaround" - it's a cognitive dissonance that would lead to the injury of humans - although as usual, Donovan figures out a solution.

Your Answer

 
discard

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.