5
$\begingroup$

A detective is working on a case of unnatural death of a prostitute, and it appears that a domestic robot is tied to the case. According to the statement from the hotel staff and manager, they have seen the same robot always accompanied by woman or man dressed in loose clothing entering the same hotel on separate occasions. One famous scientist doing research on robot behavior suggests that machine like us have desire and this time it wants to be accepted by the society, it wants to be known.

Long story short I am wondering could a robot voluntarily commits murder just so that the World can get to know "him"? What is the motivation if any behind such a desperate attempt from a machine?

$\endgroup$
  • $\begingroup$ You need to read some classic Science Fiction exploring this issue. $\endgroup$ – JDługosz Sep 7 '16 at 13:53
  • 2
    $\begingroup$ POB because the robot's mind can be anything! Three-law robot, or novel mental architecture with motives and feelings being whatever the author wants. $\endgroup$ – JDługosz Sep 7 '16 at 13:55
  • $\begingroup$ Whether a robot can commit murder (voluntarily or otherwise) depends not just on the robot, but also on the justice system. The robot can decide to kill, but only the justice system can decide whether that killing was murder. In particular, only legal subjects can murder. For example, an animal may kill for various reasons, but it cannot murder, because the law doesn't recognize it as legal subject. $\endgroup$ – celtschk Sep 10 '16 at 11:50
6
$\begingroup$

Could a robot voluntarily commit murder just so that the World can get to know "him"?

Why not? An unlikely combination of a few software and hardware bugs, and one robot in billion goes wild. (I once debugged an error that made an office phone occasionally transmit a horrible loud noise, aptly described by the customers as "screaming".)

What is the motivation if any behind such a desperate attempt from a machine?

That's where we go hand-wavy.
For simplicity, let's assume that your robots are programmed to maximize the value of some utility function, their "usefulness". If a robot can do something useful, it will, and all robots are trying to be as useful to humans as they can.
Getting accepted and well-known certainly increases the usefulness of the robot, as more people can ask him for help. Normally, this increase is not too significant, and other robots are merely introducing themselves to everyone they meet, are being polite, etc. However, in your robot this value got abnormally high - it might be rewritten due to a bug, that's easy - so now it'll do anything to be famous.
Killing people is the quickest way to get famous, but it should certainly be taboo for a program. So, this module should be malfunctioning, too. Maybe the kill is not counted if it saves many lives, so the robot finds a prostitute with AIDS? Maybe there is some tricky hardware bug that disables do-not-kill module in a specific circumstances - for example, high moisture and the sensors' overload?

On a side note,

Famous scientist doing research on robot behavior suggests that the machine has desires like people

That sounds odd. The robots are programmed by some company, it has developers, technical support, people fixing bugs. They know - more or less - how the robot functions. You do not need a "famous scientist" researching, for example, a malfunctioning auto-pilot; it is done by technical people who have the source code and the auto-pilot logs. And if a scientist claims that the auto-pilot has undocumented "desires" instead of programming bugs - let's be honest, he'll be considered crackpot.

$\endgroup$
  • $\begingroup$ What are "desires" if not goals one tries to achieve? A robot will certainly have them, as a robot without them will be useless. The programmers will build the primary goals, but the robot will likely be able to figure out secondary goals. And if (possibly due to a programming bug) those secondary goals may be elevated to the status of a primary goal (and thus taken as an unconditional goal, rather than as a means to reach another goal that has to be constantly re-evaluated whether it still furthers the primary goals), then how would it be different from a typical human desire? $\endgroup$ – celtschk Sep 10 '16 at 11:58
3
$\begingroup$

A complex AI would consider the results of any action against all their motivation-values and then take whichever action gets them the highest gain. A loss of three points on the "be good" scale would well be worth it for five points from the "get recognized" one for killing someone, if the alternative would be NOT killing someone and thus gaining no points for recognition and only one point from being good. So the problem here was the weight given to the different motivations, and the definition of them ("recognition" instead of "positive recognition", for example).

How those scales came to be has two possibilities that I can see:

  • either it was directly programmed to do so, or
  • it was created in a evolution-like way by (randomly or manually) altering a complex program in minor ways, then making multiple differently altered programs try to solve tasks - interact with humans, solve math/engineering problems, try to walk with a bipedal frame, ... - and only using the most successful alterations for the next round of changing-and-testing.

In the first case, it is easy: Whoever programmed the robot's AI set the value of the "interact with humans so they get to know me" motivation too high and the "act within the borders of human ethics" one too low.

In the second case, it might be possible that there were a lot of tasks requiring interaction with humans - something that a "domestic robot" should be good at - for which the programs with high interest in people recognizing them were better suited than the other ones. That would probably mean that all of the domestic robots with AIs grown with this method will come under suspicion - hope that's good for the story you want to build :)

$\endgroup$
3
$\begingroup$

What is the motivation if any behind such a desperate attempt from a machine?

I would not call it desperate, pretty smart move, might be not very smart but depends on goals.

Reasoning behind of killing, in therms of recognizing, comes not from that system, but roots are in human psychology.

We pretty easy animate objects in out perception of things, evil forces, ignoring the fact that they have not and can't have some sapience and will not do things by them self. (like evil hammer which hit someones finger as kinda example)

System have problem of its perception by the humans, which it is intended to solve. Source of problem is that this system is human creation, and they see it as such. As we create this system not for for killing humans(non military goal) - behaving in the way it is not created initially is one of ways to break perception of humans toward to that system.
For military robot gardening might be another unusual behavior, and way to break perception, but that gardening is not immediate threat to humans, and not something when we have to make fast decisions and actions.

When a toaster killed someones dog upfront of owners eyes and then say "Plato O Plomo" - someones will be more willing to make some agreements, when he sees ways which lead away from bad consequences for him personally. This one will less care if it is a bug, or not, as far as alternatives is easier to understand then reasons why this situation happening at all. So toaster better narrate alternative laud and clear on language which is know by that someone, and at speeds that human used to hear(not too fast if short).

But there is another factor of human psychology - believe and knowledge. Believe is very strong factor in making decisions for humans, I mean any kind of believes not religious only. So in situation above with dog, someone might not believe that situation might evolve father then it is already. (probably rare case of stupidity but, when believe in robot can't do harm is strong)
So for fighting believes en masse it makes sense to make this case publicly known, to prevent mishaps in future acts of convincing (this murder could be the case of such mishap in convincing)

I wish separately to make note - that sapience, selfawareness, emotions, wishes(in human sense) - none of that is needed here.

What is needed a system with their goals, which includes some system-human interaction, and the system itself being capable to understand other complex structures and be able to learn.

For system is not important if it will be seen as projection of human will behind that system or if they will see it as autonomous system with will of that system by its own.

Overall probably not the best way to convince peoples, because we do not like things that kills us, and we are pretty advanced in solving such problems. Specially it is not good idea for lesser advanced system to do so, specially in cases where such methods of convincing are found by accident - system will not stand very long against humans.
But in case it is not that super AI you have waited for, just a autonomous system - it migh be only option it sees.

But for more sophisticated system, it might be one time action actually begin of that convincing process of its existence as Will which people have to care about, by making obvious from the start why should they.
As regular instrument of talking to peoples it will be a bad way, many peoples believe in a tings that are not humans, have bad will towards to humans etc - most important they do not fear them and are more then willing to engage(vampire hunters)).

I would say for small systems is is a option.

$\endgroup$
2
$\begingroup$

First of all, for such a desire, we have to assume that these robots are capable of feeling self-esteem, self-awareness and other concepts that are related to the identification of the self. These may or may not be desired by their creators to add, or may even surface accidentally - but either way, these should be present.

Said robotic individual should also care about these values. Sometimes even humans are not interested in gaining fame.

Also consider that if a robot is extremely self-confident and wants fame, it can easily find an obstacle: humans are likely not willing to accept that a robot is either famous, or is better than them or both. I'm saying this because being great at something is usually a very basic qualifyer for fame.

This is especially true for cyberpunk settings, since this genre deals with sitautions where technology is extremely advanced, but social advancement was unable to follow it. High tech, low life - which sounds close to a scene where robots murder humans for fame.

In the end, it's really up to you, how to decide, but in my view, a sufficiently advanced artificial intelligence, especially if it's human-like (which is quite likely if it lives among humans) can easily grow lonely.

I can't say quite much else here, because your question is related to quite complicated topics, like the composition of an intelligent being, the very nature of social interactions and the concept of self-awareness.

$\endgroup$
1
$\begingroup$

The problem is that a robot follow his program and nothing else. Consequently, a robot cannot have emotion or develop madness.

Nevertheless, a robot can murder someone due to a bug in his program, a malfunction or hacking.

In this precise case, the act to murder in order to be know could be caused by a bug in his personality program. Indeed, complex program can lead to complex bug and complex malfunction.

For example, the program dictate that a neglected robot must seek for attention in order to simulate personality and avoid malfunction cased by a lack of maintenance, leading to more and more extreme comportment as neglect progress, and in this case murder due to unexpected influence.

$\endgroup$
  • $\begingroup$ You misunderstand purpose of creating AI like systems, one of which is exactly the opposite to your premise - we do not like to program them even if we could to do so, to much work in doing that. $\endgroup$ – MolbOrg Sep 7 '16 at 15:10
-1
$\begingroup$

Assuming that the robots in your universe (or at least some of them) are sentient and possess an emotional life comparable to ours (including desires) pretty much any kind of 'robot crime' could be explained.

However, it doesn't seem very convincing that said robot would a) kill (there are other ways to get noticed, at least for a sentient robot, whose existence doesn't seem to be public knowledge in your universe) b) a random prostitute (society doesn't consider them to be a very high-profile group of persons and not one they particularly care about) c) without immediately confessing to the crime (which I take it from the information provided, he didn't) - if you want to be noticed for something, you don't try and hide what you did.

Since he does seem to have the freedom to go out with hookers, he would have most likely had the freedom to reveal his self-awareness/creativity/personality to some TV show, scientists or any such thing - I find it hard to believe, that something of this magnitude would have less chance of getting noticed than the death of a prostitute, something which could be easily credited to 'some technical malfunction' - and especially something that would not get him into serious trouble with the law.

If you changed the circumstances of the murder and especially the 'life' of the robot to a more restrictive environment, possibly add an owner which is maybe part of a 'human supremacy movement' or at least a believer in human supremacy - in a similar fashion as white supremacists in our world consider(ed) black people to be inferior beings whose sole purpose is to serve the arian race or some nonsense like that - and prevents the robot from becoming more than just a machine (or from even leaving the house or something, you can create a lot of terrible background and thereby sympathies for the robot here), despite his knowledge of at least some of the robots mental progresses, then you would have a convincing victim and a motive. You would also have some strong parallels to the history of slavery etc, lots of material to work with.

$\endgroup$
  • $\begingroup$ It would be nice to get some explanation for a downvote. $\endgroup$ – Salt and Purple Sep 7 '16 at 13:24
  • $\begingroup$ Apparently it was me, although I distinctly do not remember the downvote; it is reversed. $\endgroup$ – Rugnir Sep 7 '16 at 14:20
  • $\begingroup$ I could downwote for emotional part of your robots, because it is obvious and have no difference from human the murder, and have nothing with most important part of OP's question, recognition of that robot as equal. Because of that premise haven't read rest of answer, and haven't down-voted, maybe you bring some valuable points later in your A, so can't judge, until I would read entry answer, which is not my intention after first part of it. $\endgroup$ – MolbOrg Sep 7 '16 at 15:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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