Skip to main content
GameDev.net gamedev.net
🔒 Locked

Blurry line between morality and sentience?

Started by slayemin Jul 30, 2009 at 3:09 PM 18 replies 2.9k views
Original Post
slayemin
slayemin
Scenario 1: A predator drone is flying over Afghanistan. It's equipped with air-to-ground missiles. It's flying on autopilot and searching for bad guys to blow up. After a few hours, it finds a target and shoots. Unfortunately, the predator drone had a false-positive and ended up killing innocent people. Scenario 2: In the near future, an MIT computer scientist creates the first sentient robot. The AI for the robot is an artificial neural network and is capable of thinking, acting, and behaving very much like a human being. For all intents and purposes, it has an autonomous will, is rational, is capable of learning, and would pass the Turing test with flying colors. The robot spends a year learning about the ways of humans (its a very fast learner). Afterwards, the MIT computer scientist releases the robot into the world to see how it fares. In less than a month, the robot intentionally murders someone. Scenario 3: Dr. Frankenstein decides to have another go at creating an artificially reanimated corpse. This time, he vows to make sure that the being is as perfect as possible. The being looks beautiful, speaks eloquently, thinks rationally, has an autonomous will, etc. It too would pass the Turing test and would be nearly indistinguishable from a human being. Dr. Frankenstein teaches his creation all the ways of humanity and bestows a sense of moral values into it. Eventually, Dr. Frankenstein releases his creation into the wild and follows it to see how it interacts with people. To his disappointment, Frankenstein monster number two also murders someone intentionally. Scenario 4: Mr. and Mrs. Jones have a newborn son. They raise their son as best they can, teaching him all the ways of humanity and morality. By all accounts, they raised their son as best they could. Their son turns into a legal adult and moves away from home. A month later, the son is involved in an altercation in which he ends up murdering someone intentionally. Now, the question is this: In all the cases, who is the murderer? Who is to blame for the murder? who is to be punished and how is the punishment to be rendered? In the last scenario, my initial intuition says that the son of Mr. and Mrs. Jones should be held accountable for his own actions. After all, he has a mind of his own (in that its an autonomous will with sentient reason). Mr. and Mrs. Jones can't be blamed for the actions of their son. This seems practically plausible, right? Yet, in Scenario #3, our intuition seems to blame Dr. Frankenstein for releasing a monster into the world. Clearly, the doctor should be to blame for the murder committed by his second monster. But, isn't Dr. Frankenstein just as much of a parent to his creation as Mr. and Mrs. Jones are to their son? Doesn't his monster also have an autonomous will and fully rational sentience? In scenario #2, the similarity between the sentience between Mr.Jones Jr. and the robot are exactly the same. Yet, there's an even more stark contrast between who is the murderer, who should be blamed, and who should be punished. Much like Dr. Frankenstein, the computer scientist would be blamed for the murder. But why? Isn't the robot just as equally sentient as Mr. Jones Jr.? In the case of the predator drone, obviously nobody would blame a machine for acting the way it did on autopilot. The artificial intelligence is pretty much a script which failed. Calling the AI script a murderer would be about as absurd as calling complex math equations murderers. Now, can we work our way backwards from the drone back to Mr. Jones Jr, and say that the AI algorithm in the drone is not to blame? If AI algorithms are not to blame, and the artificial neural network in the robot is also an AI algorithm, albeit more complex, then it can't be blamed either. Yet, it has characteristics of sentience. Much in the same way, the only difference between the mind of Dr. Frankensteins monster and the robot is that Dr. Frankensteins monster has a biological neural network instead of an electronic one. It's still a complex algorithm, which happens to exhibit the characteristics of a sentient being. The only difference between Dr. Frankensteins monster and Mr. Jones Jr. is that the monster is an artificial construction of human life. Yet, they both share a biological brain. It could even be argued that Mr. Jones Jr. is also a construction of sentience, except that the method of construction involved sex between Mr. and Mrs. Jones. Isn't Mr. Jones Jr's mind just a complex neural network algorithm? And the things which shaped the mind of Mr. Jones Jr. were the influences of his parentage? So, if we blame the people who unleash the Frankensteins, the robots and the predator drones for the actions their creations commit, shouldn't we also hold Mr. and Mrs. Jones directly accountable for the actions of their son? Or, should we hold the actual actors accountable for their actions? If so, how do you go about punishing a robot?
Machaira
Machaira
If the entity is capable of understanding the difference between right and wrong, knows that murder is wrong, chooses to commit murder, it's to blame, no matter its origin.
Former Microsoft XNA and Xbox MVP | Check out my blog for random ramblings on game development
LessBread
LessBread
Quote:
Original post by slayemin
How do you go about punishing a robot?


Remove it's battery pack.

If a tiger or a chimp escapes from the zoo and kills a person, who do we hold responsible? Whomever is negligible. There's your answer viz Predator drones.

Dr. Frankenstein and MIT scientists are not like parents in that parents don't design their children, and while parents are the primary influence on children, they aren't the only influences. To borrow from Pete Townsend, Wicked Uncle Ernie may have damaged the child without the parents ever knowing.

Check this out: Scientists Worry Machines May Outsmart Man Impressed and alarmed by advances in artificial intelligence, a group of computer scientists is debating whether there should be limits on research that might lead to loss of human control over computer-based systems that carry a growing share of society’s workload, from waging war to chatting with customers on the phone.
"I thought what I'd do was, I'd pretend I was one of those deaf-mutes." - the Laughing Man
Toni Petrina
Toni Petrina
Quote:
Original post by Machaira
If the entity is capable of understanding the difference between right and wrong, knows that murder is wrong, chooses to commit murder, it's to blame, no matter its origin.


It still doesn't explain the fact that someone who did know the difference still had to act, not choose, but act. And whatever criteria was used, you can't say that the being in question is responsible for that criteria since that would lead to infinite regress.

OTOH, in all 4 scenarios you have to try to make sure that it doesn't happen again. That is why we have jails :)
So... Muira Yoshimoto sliced off his head, walked 8 miles, and defeated a Mongolian horde... by beating them with his head?

Documentation? "We are writing games, we don't have to document anything".
HostileExpanse
HostileExpanse
Until programs are recognized as citizens, responsibility for their actions fall upon the custodian who set them loose upon society.
Zahlman
Zahlman
Quote:
Original post by LessBread
If a tiger or a chimp escapes from the zoo and kills a person, who do we hold responsible? Whomever is negligible.


I hope you meant "whomever [was] negligent" :)
LessBread
LessBread
Yes. And if they happen to be very very tiny, so much the better! [grin]
"I thought what I'd do was, I'd pretend I was one of those deaf-mutes." - the Laughing Man
Sirisian
Sirisian
Quote:
Original post by LessBread
Yes. And if they happen to be very very tiny, so much the better! [grin]
heh that's what I was thinking of. Then again whomever invents gray goo first will probably be the first to die.
Calabi
Calabi
If the scientists or persons recognise and say that it has an autonomous will, then they have recognised its responsible for its own actions.

By allowing the robot to fly around and choose targets of its own they have made it responsible for its own actions. When it does make a mistake they probably wouldnt put it in prison, just dissasemble. Or if they are really clever they would just put it down to collateral damage, as if people cannot get targets correct a hundred percent of the time they would know neither could an autonomous AI. No matter how much tech and money they throw at it. They might get the odds quite high but never perfect which is probably why we wont have fully autonomous fighting machines for a long time, if ever.

Quote:

Check this out: Scientists Worry Machines May Outsmart Man Impressed and alarmed by advances in artificial intelligence, a group of computer scientists is debating whether there should be limits on research that might lead to loss of human control over computer-based systems that carry a growing share of society’s workload, from waging war to chatting with customers on the phone.


I think the odds for that ever happening are a million to one.

[Edited by - Calabi on July 31, 2009 2:44:42 PM]
necreia
necreia
Quote:
Original post by Calabi
... They might get the odds quite high but never perfect which is probably why we wont have fully autonomous fighting machines for a long time, if ever.


I don't think line that needs to be crossed isn't "perfect" but rather "higher or equal to human error".

If a soldier has a 96% success rate but the robot has a 99%-- the robot should replace the soldier.
Dim_Yimma_H
Dim_Yimma_H
You can also look at it like this: The amount of time spent on functioning well makes the eventual failure more serious, because it doesn't make (common) sense.

Everything that's new and has no experience is probably held less responsible than that which should know better. Frankenstein's monster didn't get much time to learn anything, while it's creator should've been able to learn before it. The short time of functioning is probably also why the monster will be demolished quite quickly, it hasn't existed long enough for anyone to care about its further existence.

Parents will always have the opportunity to know better than their children, but when the children have existed long enough it's not possible for their parents to to know for them, the influences on the children come from so many places.

Influences, I think that makes an important distinction here:
1. The drone was only "influenced" by its manufacturers.
2. The robot was influenced by the scientist, and a little by whatever information it "learned"... Who decided what it should learn? - The scientist, probably.
3. Frankenstein’s monster was only influenced by Dr. Frankenstein, so he's the main cause.
4. The child of Jones' is influenced by the parents, but more so by the society - It's simply not possible to tell the main point of influence, which places the blame on the central point, the child.

So I think I agree with LessBread about this one.

Quote:
Original post by Toni Petrina
OTOH, in all 4 scenarios you have to try to make sure that it doesn't happen again. That is why we have jails :)

That may be the only thing that, kinda objectively seen, makes sense in this dilemma.
Calabi
Calabi
Quote:
I don't think line that needs to be crossed isn't "perfect" but rather "higher or equal to human error".

If a soldier has a 96% success rate but the robot has a 99%-- the robot should replace the soldier.


I can imagine that being the case if just the bean counters being involved, but general public uproar and old fashioned tradition might get in the way. People generally, are less forgiving and demand more of machines.
Dim_Yimma_H
Dim_Yimma_H
Possibly because they don't care for the machines personally. I wonder if that could change if someone grew up with an ever nearby artificially intelligent machine - Though, that's not really realistic yet.
LessBread
LessBread
Now there is an idea. Instead of unleashing an AI with claws on the world straight out of the box, what if it was packaged in a cell phone (or similar small housing) and then given to a child to teach tamagotchi style or if not for interaction, then for observation. In other words, piggyback the development of such an AI on the emotional development of a child. Establish a regimen of development for AI sans claws that culminates in a series of tests that must be passed before that AI is given claws. That's what we do with ourselves. We don't just hand out guns and badges to people and say they're police.

The obvious rebuttal to this idea (outside of technical issues of machine learning), is that the inspiration for these kinds of AI is that they can be manufactured, that we don't have to wait 20 or 30 years for them to "ripen". That brings up questions about deploying manufactured inorganic autonomous devices to kill human beings and thus destroy the decades of development that goes into each person.

Dim_Yimma_H also seems to be hinting at the possibility of using AI as a kind of "machine assisted ethics". So as above, the AI sans claws in a cell phone accompanies a child observing but also advising the child in regard to the legal, ethical and moral implications of the situation the child find him/her-self in, a kind of AI-Nanny (or AI-Nag).
"I thought what I'd do was, I'd pretend I was one of those deaf-mutes." - the Laughing Man
LessBread
LessBread
Here's another recent news story regarding robots, ethics and the sentience of readers: Robot developers learn perils of new media

Quote:

Somewhere in the blogosphere the biomass-fueled robot being developed for the military turned into a battlefield corpse-eater. Their lesson was that information, once released, can't be controlled.
...
EATR, for Energetically Autonomous Tactical Robot, is a robotic ground vehicle that Finkelstein's small company is designing with U.S. Defense Department funding; it can sustain itself on long missions by foraging for twigs, leaves and other kinds of vegetation.

But wild speculation on the Internet this month was that Finkelstein's Robotic Technology Inc. and a partner were building flesh-eating robots for the Pentagon.

Scores of blogs and news sites, including FoxNews.com, ran with the unfounded report. The online furor caught the companies off guard and turned into a major distraction.
...
The speculation that launched EATR into the popular consciousness as a carnivore can be traced to a straightforward news release from Cyclone on July 7. The word "biomass" in the release was misinterpreted to mean EATR could feed on corpses on a battlefield.

It reached a fever pitch in mid-July, when FoxNews, Fast Company and CNET published online reports repeating the speculation, without first checking with Robotic Technologies or Cyclone.

The companies' websites were swamped. The project's main sponsor, the Defense Department's Defense Advanced Research Projects Agency, wanted the public record corrected.

Cyclone issued a second news release July 16, calling EATR a "vegetarian" -- leading to even more news coverage. A story about EATR being a "vegetarian" ran in Britain's Guardian newspaper.
...
"I thought what I'd do was, I'd pretend I was one of those deaf-mutes." - the Laughing Man
Machaira
Machaira
Quote:
Original post by Toni Petrina
It still doesn't explain the fact that someone who did know the difference still had to act, not choose, but act.

What's to explain? If they commit the act of murder, knowing it's wrong, they're to blame.

Quote:
Original post by Toni PetrinaAnd whatever criteria was used, you can't say that the being in question is responsible for that criteria since that would lead to infinite regress.

I never said they were responsible for the criteria. You follow the laws of the land though or you pay the penalty.
Former Microsoft XNA and Xbox MVP | Check out my blog for random ramblings on game development
slayemin
slayemin
Quote:
Original post by necreia
Quote:
Original post by Calabi
... They might get the odds quite high but never perfect which is probably why we wont have fully autonomous fighting machines for a long time, if ever.


I don't think line that needs to be crossed isn't "perfect" but rather "higher or equal to human error".

If a soldier has a 96% success rate but the robot has a 99%-- the robot should replace the soldier.


This is extremely frightening to me. Imagine if we managed to build robots with nano-technology and imprinted an AI which had a "seek and destroy bad guys" algorithm with 99% efficiency. The robots would fly like humming birds and be the high speed projectile (instead of bullets). They could be dropped onto a city (via missile payload) and they'd scour the city like a massive swarm of locusts, killing every hostile person. Air forces, navies, armies, marines, etc. would all be rendered obsolete.

The scariest part is that these little buggers could be mass produced like twinkies and they wouldn't cost more then $5. Suddenly, politicians of technologically advanced nations can wage a "war" (more like slaughter) without loss or risk of life to their citizens, and maybe at a cost of a few million dollars. The only restraint would be a moral restraint, and based on what we see with corrupt politicians, we'd all be doomed.

Suppose that you do replace all your infantry with robots and fight a war against a technologically inferior nation. Is it even fair that death on one side results in scrap metal and death on the other side results in severe loss of life and families torn apart? I mean, you'd have a bunch of bots on one side that go "boom, head shot!" with mathematical precision and "spray and pray" on the other side, hoping they get lucky.

I think the future is too scary if war is still being waged. People need to STFU about their petty beefs with each other and find better ways to get along...before it's too late and we extinct ourselves.

---------------

Quote:
Influences, I think that makes an important distinction here:
1. The drone was only "influenced" by its manufacturers.
2. The robot was influenced by the scientist, and a little by whatever information it "learned"... Who decided what it should learn? - The scientist, probably.
3. Frankenstein’s monster was only influenced by Dr. Frankenstein, so he's the main cause.
4. The child of Jones' is influenced by the parents, but more so by the society - It's simply not possible to tell the main point of influence, which places the blame on the central point, the child.


In the first scenario, the predator drone is just a set of instructions. Lot's of If-then statements. There's not really an "influence", other then the programmer who wrote the instructions.

The other three situations are a thought exercise in exploring our humanity. You seem to believe that our sense of humanity is bestowed upon us by our nurturing. If you give each being the exact same nurturing in every scenario so that nurture isn't a factor, is there anything about our nature as human beings which distinguishes us from the Frankenstein monster and the sentient robot?

Quote:
The obvious rebuttal to this idea (outside of technical issues of machine learning), is that the inspiration for these kinds of AI is that they can be manufactured, that we don't have to wait 20 or 30 years for them to "ripen". That brings up questions about deploying manufactured inorganic autonomous devices to kill human beings and thus destroy the decades of development that goes into each person.


Let's say that the sentient robot spends 20 years learning the ways and customs of people by being a part of a family. Or maybe you even have fifty robots learning our ways and customs. Then you pick the one robot you think exhibits the best character traits, copy their brain, and upload it to every subsequent robot produced. Each manufactured robot would then have 20 years of life experience, yet be one day old.
A) Would it be immoral to "delete" the other 49 electronic brains? (probably an unoriginal question that's been asked a thousand times) What's the difference between deleting an electronic mind with 20 years of life experience and murdering a human being?
LessBread
LessBread
Quote:
Original post by slayemin
Quote:
The obvious rebuttal to this idea (outside of technical issues of machine learning), is that the inspiration for these kinds of AI is that they can be manufactured, that we don't have to wait 20 or 30 years for them to "ripen". That brings up questions about deploying manufactured inorganic autonomous devices to kill human beings and thus destroy the decades of development that goes into each person.


Let's say that the sentient robot spends 20 years learning the ways and customs of people by being a part of a family. Or maybe you even have fifty robots learning our ways and customs. Then you pick the one robot you think exhibits the best character traits, copy their brain, and upload it to every subsequent robot produced. Each manufactured robot would then have 20 years of life experience, yet be one day old.


That's probably how it would be done, just like with a hard drive and a new pc.

Quote:
Original post by slayemin
A) Would it be immoral to "delete" the other 49 electronic brains? (probably an unoriginal question that's been asked a thousand times) What's the difference between deleting an electronic mind with 20 years of life experience and murdering a human being?


The 20 years of electronic mind experience could be stored on disk, so it wouldn't be deleting it. And the machine could be shut off and reprogrammed, per the above scenario. The Sarah Connor Chronicles began to explore this idea right before Fox canceled the show. Murdering a human being on the other hand... to cite Bill Munny, "It's a hell of a thing, killing a man. Take away all he's got and all he's ever gonna have."


"I thought what I'd do was, I'd pretend I was one of those deaf-mutes." - the Laughing Man
Krokhin
Krokhin
Quote:
Original post by slayemin
Scenario 1: A predator drone is flying over Afghanistan. It's equipped with air-to-ground missiles. It's flying on autopilot and searching for bad guys to blow up. After a few hours, it finds a target and shoots. Unfortunately, the predator drone had a false-positive and ended up killing innocent people.
...
Now, the question is this: In all the cases, who is the murderer? Who is to blame for the murder? who is to be punished and how is the punishment to be rendered?

People who ordered to attack.Not a drone,because it hasn't mind and free will.And not a drones creator,because next time they can protect innocent people from terrorists.Don't sure-don't shoot,otherwise one day civil people will buy
">something
to protect themselves,may be get to people who order, and will become "bad guys" [smile]
Dim_Yimma_H
Dim_Yimma_H
Quote:
Original post by slayemin
You seem to believe that our sense of humanity is bestowed upon us by our nurturing.

Ah, but that's not the truth. I believe the source of influence is relevant when it comes to what responsibility can be expected from a being - or the source of "nurture". Genes play a part in the weight a being puts on different moral choices, even after the logic of the moral has been nurtured.

Quote:
Original post by slayemin
If you give each being the exact same nurturing in every scenario so that nurture isn't a factor, is there anything about our nature as human beings which distinguishes us from the Frankenstein monster and the sentient robot?

A. Biological genes
B. Personal effort of learning
- Is there something other you're thinking of?

Topic Locked

This topic has been locked by a moderator. New replies are not allowed.

Sign in to reply to this topic.