Robots Must Be Designed To Be Compassionate, Says SoftBank CEO

An anonymous reader writes: At the SoftBank World conference in Tokyo, SoftBank CEO Masayoshi Son has made a case for robots to be developed so as to form empathic and emotional relationships with people. "I'm sure that most people would rather have the warm-hearted person as a friendSomeday robots will be more intelligent than human beings, and [such robots] must also be pure, nice, and compassionate toward people," SoftBank's Aldebaran tech group will make its empathic "Pepper" robot available for companies to rent in Japan from October at a rate of $442 per month.

Comments

Can my car have a sense of humour too?

[Anonymous Coward]

40 years from now when the robot wars start, we'll remember this article, saying "we should have listened... we should have listened."

Re:Can my car have a sense of humour too?

[davester666]

And the manufacturer will say "We did. That rubber bumper at the end of the gun isn't just there for show."

Re:Can my car have a sense of humour too?

[Z80a]

There won't be a robot wars because they will listen.
Instead we will just manufacture a robotic version of Kyubey from Madoka Magica that will take over the world by manipulating our primitive emotion driven brains like puppets.
He will not need to fire a single bullet, but just say the right words.

Re:Can my car have a sense of humour too?

[The Real Dr John]

I wonder what it is that makes people think that robots can be given emotions, when we have no idea how brains generate emotions? And even when we do figure out how brains do that, what makes people think that it doesn't require living cells to have feelings, sensations and emotions? In which case you would have to grow robots from single cells, which would mean they were living, rather than robots.

Re:Can my car have a sense of humour too?

[LifesABeach]

Maybe it's time to start developing the software for the 3 Laws of Robotics?

Re:Can my car have a sense of humour too?

[morgauxo]

"I wonder what it is that makes people think that robots can be given emotions, when we have no idea how brains generate emotions?"

Ah, how refreshing. An intelligent thought on this subject.

"what makes people think that it doesn't require living cells to have feelings, sensations and emotions?"

Damn... I jumped the gun.

Cells are really vessels containing incredibly complicated chemical reactions. They communicate through various electrical and chemical means. I don't know if there is any non-material, spiritual aspect to it or not (and not interested in a debate on it) but if the sum of all these reactions and chemical plus electrical signals can add up to either make us or attract the undefined spiritual thing or whatever.

Why couldn't the sum of a whole bunch of electrical signals through semiconductor switches or even something really crazy like a mechanical computer made of wheels and gears do the same?

Now.. whether something like this could actually be built is an entirely different question. The complexity would be unimaginable. But then.. I don't see anybody building humans out of raw chemicals either yet nobody is going to argue that a being made of cells can't have feelings, sensations and emotions.

Re:Can my car have a sense of humour too?

[The Real Dr John]

I would like it very much if you could provide some information that suggests that all you need is complexity to make something with feelings and emotions. Cells respond to chemicals, like hormones and neurotransmitters because that is how signaling pathways in the target cells are activated. But the key is that they are alive, which allows for sensations and in more advanced organisms with a complex central nervous system, emotions. Just because emotions involve "chemical reactions" (like the signaling cascade that occurs after a glutamate ion binds to an NMDA receptor in the brain) that doesn't mean all you need for emotions is the right chemical reactions. You need the living cells in an extremely complex network, at least that is the best neuroscience can determine at this point. There is no evidence that any machine of any type even has a "thought", let alone an emotion.

So until you can come up with examples where any machine has shown the capacity for sensations or thoughts or emotions, then I posit, along with lots of other neuroscientists, that living cells are required to have attributes of living organisms. It is why we distinguish between animate and inanimate objects.

Re:Can my car have a sense of humour too?

[morgauxo]

"I would like it very much if you could provide some information that suggests that all you need is complexity to make something with feelings and emotions."

I don't know why you would expect evidence of that from me. It isn't quite what I said! Maybe there is some aspect to our existence as people that goes beyond the materials and energy interacting within us. I don't know. Good luck proving it either way. But.. even if something like that does exist.. is it generated by the formation of our bodies? Is it somehow attracted from elsewhere and binds to us as our bodies are formed? Even if there is something more to us.. without thoroughly understanding it whose to say the same thing wouldn't (or would) happen to a sufficiently complex AI?

But... emotions, memories, sensations... they have all been produced experimentally in real human beings by electrically stimulating various areas of the brain. So.. tell me that much of us doesn't exist at least on some level as mere signals passed around by cells.

"But the key is that they are alive"

Have you ever attempted to put a real definition on the term 'alive'? good luck with that. I feel I should write more on this but what do you write to describe the indescribable?

"There is no evidence that any machine of any type even has a "thought", let alone an emotion."

Agreed. But... human beings can be considered to be highly complex chemical machines. And.. your statement still stands. Something that was only a 'sufficiently complex' machine could theoretically appear to be a person. (oh yah.. try to define person too). Maybe that's all a sufficiently advanced AI would be. But.. maybe that's all you, and every other person on the planet is. Sufficiently advanced chemical machines that they seem to be people. I only 'know' that I am a real person. And.. nope... I can not prove it! I can give you no evidence that I actually have thoughts and emotions rather than just sufficiently complicated electrical and chemical signals to simulate them.

So... I take that one on faith. I'm not a big fan of faith over science but how do you devise a test for things like person hood that can't even be defined? Since I believe that I am a person and other humans seem to be of similar makeup and have similar traits of sensation and emotion I believe that they (including you) are people too!

If the day ever comes that someone creates a sufficiently advanced AI... it's going to be quite a conundrum figuring out if we should consider that to be a person or a thing and what basic rights therefore do or do not apply. Good luck to anyone on either side of that argument! If anyone ever thinks they can win it I think they are deluded.

I'll take the 'lets just assume it's a person' side because it' safer. It would be less harmful to anthropomorphize a machine than to objectify a person.

Re:Can my car have a sense of humour too?

[The Real Dr John]

OK, I agree completely that we don't know what "alive" means in mechanistic terms. I was going to bring that up, so I am glad you beat me to it. If we cannot define life, then we have a long ways to go before we can imitate life's abilities, wouldn't you agree?

There is an enormous gap between defining life, and making a machine that can do what a human brain does. Considering that humans are still working on what "life" is, there is little doubt that it will be along time before we can create something with sensations, thoughts and emotions. First we need to figure out why a cell is alive one minute, but dead the next when you apply a little bit of sodium azide. We know what chemical reactions we have shut off, and we know the various cellular death cascades that occur (apoptotic, necrotic cell death, etc.), but we still don't know what the attribute of "living" even is, let alone how to emulate it.

Just like we know a living organism when we see one, I have a good sense that we will know it when someone finally makes a living thing (or a thing with living attributes) when we see it. You won't need a Turing test, because it will act alive, rather than acting like it is imitating life. It will have to say "ouch!" and really mean it when you stick it with a pin.

Re:Can my car have a sense of humour too?

[The Real Dr John]

If you search PubMed for articles on artificial life you get articles like this, where they make it quite explicit that they are just trying to imitate life-like behaviors, not create them de novo. Any "emotions" would be simple imitations of behavior.

Artif Life. 2015 Spring;21(2):141-65. doi: 10.1162/ARTL_a_00164. Epub 2015 May 7.
On the Evolution of Behaviors through Embodied Imitation.
Erbas MD1, Bull L, Winfield AF2.

Abstract
This article describes research in which embodied imitation and behavioral adaptation are investigated in collective robotics. We model social learning in artificial agents with real robots. The robots are able to observe and learn each others' movement patterns using their on-board sensors only, so that imitation is embodied. We show that the variations that arise from embodiment allow certain behaviors that are better adapted to the process of imitation to emerge and evolve during multiple cycles of imitation. As these behaviors are more robust to uncertainties in the real robots' sensors and actuators, they can be learned by other members of the collective with higher fidelity. Three different types of learned-behavior memory have been experimentally tested to investigate the effect of memory capacity on the evolution of movement patterns, and results show that as the movement patterns evolve through multiple cycles of imitation, selection, and variation, the robots are able to, in a sense, agree on the structure of the behaviors that are imitated.

Re:Can my car have a sense of humour too?

[morgauxo]

Ok... I wrote a big huge TLDR response to this but then I decided to open a new-tab and re-read the comments that began our discussion and deleted it. Now I will write another :-)

Life can be described in 'mechanistic' terms. A definition that I have heard along those lines is that it must have metabolism, reproduction, self repair and evolution to be life. Usually this results in discussions about what happens if you find something that only fulfills some of the 4. Also commonly mentioned are viri which in a sense do some of these things but require cells to infect in order to do so. This seems to me to be the kind of 'life' definition you are using when you talk about things like death from sodium azide.

I'm not sure how it can be argued that life, by that kind of definition has anything to do with emotions. Do single celled organisms have emotion? Which of these attributes do what to form love, anger, fear?

Those things are states in the brain.. which happens to be composed of cells which are alive by the mechanistic definition. To some degree we can understand and demonstrate those emotions as being physical states in the brain. Scientists have brought the emotions out by poking, prodding and stimulating various regions of the brains of living, conscious people.

On the other hand, there is incredible complexity in all that huge mass of cells and how they communicate. We can't map out all the interactions which occur as a brain experiences love or anger or fear and simulate them on a computer.

Further, and i think more importantly.. I don't think we really understand how we perceive those brain states. When the cells of my brain activate to form an emotion I don't experience it as an individual cell taking a bunch of inputs from other cells and deciding based on them when it is time to activate a bunch of outputs sent to other cells. I experience it as a whole individual.. as a person.

If the emergent behavior of the sum of reactions between these mechanisticly living things can form me.. why can't the sum of reactions between some other form of logic gates combine to form what is essentially a person. I can't even define what exactly a person is so surely I can't rule that out!

It seems unlikely to me that things like metabolism and self-repair are what set me apart from a non-sentient robot.

Re:Can my car have a sense of humour too?

[morgauxo]

Splat! I just flew into the paywall.

Re:Can my car have a sense of humour too?

[The Real Dr John]

That is the question plaguing Neuroscience, Biology, Psychology and Cognitive Neuroscience (including AI work). We don't know how brain states generate the affective sensations that we experience (and I assume all animals with a sufficiently complex CNS experience, which probably includes fish, who have a very nicely complicated CNS). Right now "artificial life" researchers are struggling to come up with even a proto-bacterium-like entity from scratch. And simply putting the enzymes and transporters into a membrane-bound container doesn't make it start to act "alive".

The thing that most non-biologists don't think about is that all cells only come from other living cells. Thank Pasteur. This is not a trivial point. Except for one time we can infer, life only comes from life. Abiogenesis may have been a one time deal, or maybe just a few times as far as we know now. So clearly something unusual and interesting is involved, and we are not entirely sure what that is.

Some researchers have proposed that living systems selected metabolites based on them naturally being in a state of quantum criticality. See here for example:

http://arxiv.org/abs/1502.0688...

Whether claims like this turn out to be true or not remains to be seen. But clearly, scientists are struggling to even define the basics of what makes living organisms unique.

Re:Can my car have a sense of humour too?

[The Real Dr John]

I just tried to get it through the University library, but they do not subscribe to that journal. I wish they would just switch to the open access model for all science publications. This old print model of publishing is just dumb in the Internet age.

Re:Can my car have a sense of humour too?

[mbeckman]

"Now.. whether something like this could actually be built is an entirely different question. The complexity would be unimaginable. But then.. I don't see anybody building humans out of raw chemicals either yet nobody is going to argue that a being made of cells can't have feelings, sensations and emotions.

Morgauxo,

The important point is that nobody has made even the most rudimentary artificial intelligence, let alone one with feelings and emotions. The Softbanc blatherer is just using the latest trick to get noticed: spew dire warnings about the risks of AI. AI, which doesn't exist and isn't even remotely on the horizon, even after lots of naturally intelligent people have devoted their lives to achieving.

Personally, I think it's demons from the pit of hell we'd better watch out for. We have more evidence for them than we do for AI.

Re:Can my car have a sense of humour too?

[RespekMyAthorati]

You won't need a Turing test, because it will act alive

Too bad you will have no way of recognising whether it is "acting alive" or just faking.

It will have to say "ouch!" and really mean it when you stick it with a pin

Again, you will have no way of knowing whether it "really means it" or is just faking.
The whole issue of machine sentience is moot - and a foolish waste of time.

Re:Can my car have a sense of humour too?

[The Real Dr John]

We will see how easy it is to tell if something is alive or sentient when someone actually makes an artificial being. My guess is that it will be very easy to tell if it is a robot as long as you are allowed to fully interact with it, and touch it and move it. I wonder how much effort people are going to be willing to put into making robots act alive when it is just imitation. The Japanese are doing it quite a lot, and the end results are extremely easy to discern from living organisms. I have heard so much talk about it, but I have not seen anything other than robots programmed to crudely imitate people.

Re:Can my car have a sense of humour too?

[Anonymous Coward]

Which is about as "compassionate" as one can make them.

The fundamental problem with this logic of "making robots compassionate" is you can't delegate something that is uniquely human to an inanimate object incapable of knowing what "compassionate" means in the first place. This is no different that asking a rock to be compassionate and asking for a human law assuring it to be so. It's no different than legislating pi to be the value 3.

Basically this is a cop-out for not taking HUMAN responsibility for the creations of HUMANS and that itself is immoral.

Re: Can my car have a sense of humour too?

[Anonymous Coward]

wrong as usual. compassion is not unique to humans.

have a great day, internet win you.

Re:Can my car have a sense of humour too?

[morgauxo]

I see several questions in here and I'm not entirely convinced that it can be proved that the answers are all related.

When I think of generating sensations I think of nerves sending messages to one another to get from some point in ones skin to some point in the brain. It seems trivial to grasp that concept while at the same time the complexity at the cellular level and below could be beyond comprehension. Still, this could be very easily seen as being similar to signals on a wire between a sensor and a CPU or neural net in a piece of technology.

To me the question isn't how are these sensations generated, it's how is it that I perceive them as a whole, self-aware being. Why am I not just a bunch of cells each perceiving only the signals they receive individually from their neighbors, completely unaware of the complex human behavior that emerges as the sum of their interaction. Or.. why stop at cells? Why aren't we just collections of individual atoms interacting with one another, unaware of their nature as part of the whole. Or quarks.. strings... ????

I've watched a few videos where scientists talk about their different ideas regarding this. I remember a few where they showed that certain areas of the brain show activity in a conscious person and if that area is turned off the person has no awareness. This certainly may be part of the answer however it still just sounds like circuits to me. It sounds like a switch that can be turned on or off. But what is it turning on or off?

Whatever it is does it make a difference if that switch is made from flesh, silicon or something entirely different? I think the original question was why believe that it can. My question is... why believe that it can't.

As for the origin of life. That is a great question and a fun one to learn about. I'm not really sure how it's relevant though. It's the question of how did this particular kind of 'sensing beings' ie life come about. It's not a question of what other things might or might not be possible. You asked "what makes people think that it doesn't require living cells to have feelings, sensations and emotions?"

What is living? We seem to agree that we can't define that. What is a cell? It's a living thing enclosed in a membrane. It's a part of the 'mechanistic' definition of the only life we currently know. How about Feelings, sensations and emotions? Those could be defined as simply current status. The robot's sensor sends a message to it's CPU that it is being damaged; nerves in my skin tell my brain that I am being cut. Emotions.. my current mental state. Programs contain variables.. the programmer can instantiate a variable to hold whatever states the programmer wants to keep track of. The AI's anger status is currently 90%, switch to the more aggressive behavior subroutine...

I think the real question is why we perceive this status information as whole, self-aware beings that is the question. I don't know if we even can ever hope to answer that but until then it's not that I think I have a reason to believe a machine could do the same. It's that I have no useful information to indicate that one could not.

Re:Can my car have a sense of humour too?

[The Real Dr John]

Clearly, the brain works like other tissues in the body where the individual cellular interactions at a smaller scale exhibit emergent properties at the organ scale. The human brain has a complexity that no one person could ever hope to understand, and large teams of scientists struggle to understand small bits of the puzzle. Take a look at these videos of a new technique for looking at the micro-structure of the mouse brain. This is a fascinating technique that required the work of many neuroscientists and computer programmers.

http://www.theguardian.com/sci...

The complexity surrounding a single portion of one dendritic shaft is mind boggling. You will love the video, I promise.

Everyone who got rich by accident is now a sage

[Anonymous Coward]

Maybe we should also finally stop doing this education thing and breed idiots selectively for luck?

How?

[KermodeBear]

And how, exactly, does one program a robot to be compassionate or empathetic?

Can emotion be reduced to a few simple formulas, some generic algorithms?

I'm not convinced.

Re:How?

[jblues]

A paper-clip / puppy appears in your room.

It looks like you're still not convinced. I know how that feels, and I'm here to help you. Do you want me to:

• Convince you some more
• Change the topic
• Just go away

Re:How?

[Anonymous Coward]

Do the wind chime trick. I love that trick.

Re:How?

[thinkwaitfast]

And how, exactly, does one program a robot to be compassionate or empathetic?

There's an opcode for that. Duh. Set the compassion bit or clear it to be a jerk.

Re:How?

[rickb928]

First Law.

If you think any of this is expected to be reduced to a few simple formulas, some generic algorithms, you're pretty lost on this. Such AI is both a long way from being this functional AND right around the corner.

Three Laws Safe. Only way.

Re:How?

[ShanghaiBill]

Can emotion be reduced to a few simple formulas, some generic algorithms?

Yes. Emotional connection is not complicated. Many people felt a connection to Eliza, which was a trivial program.

This works:
1. Look people in the eye, and smile.
2. Agree with what they say.
3. Instead of talking about yourself, ask other people questions to show you are interesting in hearing them talk about themselves.
Follow this formula, and you will be popular.

I'm not convinced.

Have you ever gotten laid?

Re:How?

[ljw1004]

And how, exactly, does one program a robot to be compassionate or empathetic? Can emotion be reduced to a few simple formulas, some generic algorithms? I'm not convinced.

Yes it basically can be reduced to a few simple formulas. Have you ever been to couple's counselling or the like? The rules are very simple. You listen to what someone says. The only questions you ask are ones that help you understand the spirit of what they're saying. When they're done you repeat back "I heard you tell me that XYZ" in your own words as faithfully as possible. Hey presto, empathy and social connection.

It sounds corny but it works incredibly well at (1) helping the other person feel understood, (2) changing your own mental approach so you really do understand them better in a good way.

It also works really well in a professional setting, in meetings.

Re:How?

[Anonymous Coward]

So what you're saying is, every popular person is a deceitful liar, especially you. Thanks for the heads up!

Re: How?

[DNS-and-BIND]

Hey, if it gets you laid, go with it. We have to use what works, not what we wish would work in a Utopia. It's a big shit sandwich, and we're all going to have to take a bite. When you reward a behavior, you get more of it. This is science.

Re:How?

[Anonymous Coward]

One of the reasons for why little kids are able to be cruel to each other is because they have no idea about how others feel and how their actions can lead to others feeling pain. It's not until they get hurt in the same way do they realize how to predict and prevent others from being hurt.

In other words they have to be programmed with per-defined subroutines of their own to feel other's pain.

Re:How?

[mwvdlee]

This.

Just as robots won't turn into some psychoc skynet bent on destroying humanity, they won't turn nice and happy either.

Re:How?

[ThatsMyNick]

I think what is he is saying is every human is a deceitful liar, and I agree. If you know of any human who has never lied, I will change my mind.

Re:How?

[Anonymous Coward]

>And how, exactly, does one program a robot to be compassionate or empathetic?

Based on research into anime and manga, gynoid frames can be equipped with a "black box" named "otome kayro" which instantly turns them into japanese schoolgirls with hidden superpowers, including irresistable cuteness (kawaii) and the ability to pilot giant walkers (so-called mecha).

Re:How?

[AmiMoJo]

Can emotion be reduced to a few simple formulas, some generic algorithms?

Not emotion, by certainly empathy can be boiled down to rules that a robot can learn. In fact empathy is taught in some fields, like nursing, and it involves understanding how people react to information and how to deliver it in a way that accounts for that.

Robots can be programmed to deliver painful news in a manner that accounts for the likely reaction and emotions of the listener. They can show sympathy when things go wrong, or refrain from pointing out mistakes in a matter-of-fact way and instead appreciate that they may be embarrassing.

In other words, he is saying that computers should be less like a computer that might dryly state "the patient is deceased" and more like a person who might say "Mr. Jones passed away."

Re:How?

[MrL0G1C]

In other words they have to be programmed with per-defined subroutines of their own to feel other's pain.

But with what language, Basic? Assembler? C++? Javascript?

Re:How?

[MrL0G1C]

by certainly empathy can be boiled down to rules that a robot can learn.

Since we don't have AI yet, how on earth do you propose we 'teach' non-existent machines?

This whole thread is stupid, we don't have intelligent robots, Softbank CEO is living in a fantasy world.

Re:How?

[cascadingstylesheet]

Can emotion be reduced to a few simple formulas, some generic algorithms?

I'm not convinced.

Maybe with the same magic wand that reduces thought to a few simple formulas, some generic algorithms?

I mean, since we have that magic wand (right?), might as well go for broke ...

Re: How?

[Anonymous Coward]

Rape, rape, rape, rape. Always with the rape language. Look, the world is an unpleasant place and we have to face unpleasant facts. What women say that they respond to and what women actually respond to are two different things. Attempting to shame men into acting like limp-wristed weirdos isn't going to solve anything. We need to take a long, hard look at why women act they way that they do, and as long as people like you keep shouting RAPE at every opportunity, we will never make any progress.

Re:How?

[AmiMoJo]

Since we don't have AI yet, how on earth do you propose we 'teach' non-existent machines?

I don't. I propose we design robots to communicate in a way that shows empathy, like you would any software system.

Don't design a robot face that is always smiling if it may have to deliver bad news sometimes. If it can alter its expression, make sure it is always appropriate. If it can speak, consider the tone of voice to use when giving information that may be sensitive, in the same way as you might consider making text on a computer screen bold or hidden (for password entry). No need for AI, just good design.

Re:How?

[MrL0G1C]

Not emotion, by certainly empathy can be boiled down to rules that a robot can learn.

Re:How?

[rgbatduke]

I think it is pretty safe to say that at any given time, 1 to 2% of the human population has never lied.

So I guess you should change your mind.

Re:How?

[ThaumaTechnician]

I'm sooo getting tired of top-level, highly-paid executives who give out these kinds of general directions, meanwhile they have absolutely no idea where to start, no inkling of how it might work, nor whether it's even mathematically possible. I'm looking of at you, Mr. Director of the FBI, whose main qualifications seems to be to ignore science..

Grrrr! My consumer-/tax-dollars at work.

Re:How?

[Impy the Impiuos Imp]

Robots Must Be Designed To Be Compassionate, Says SoftBank CEO

I say we follow the Dutch model of compassion. They pay prostitutes to jerk off people in hospitals.

Re:How?

[Big Hairy Ian]

I think you're being over emotional about this you need to check your algorithms :D

Re:How?

[lq_x_pl]

(2) changing your own mental approach so you really do understand them better in a good way
Which is also where things get wonky with robots. This is a non-deterministic operation. From the ground up, robots are generally designed to behave in a predictable fashion. The human brain is exceptionally plastic, and our ability to socialize/associate on the fly is still mostly a mystery. We may be able to mock up a sufficiently complex and convincing strategy for the robot to follow, but it is still just running down a list of rules. Empathy (like the CEO wants) is a squishy strange thing that us meatbags are still trying to understand. Then again, I guess if the bot is convincing enough the debate on whether it actually has empathy becomes philosophical.
Step one can be reduced to some regex, look-up tables, and copy-n-paste (active listening is silly like that).

Re:How?

[Anonymous Coward]

It is my opinion that emotion and consciousness are separate function from intelligence.That would mean that you have 2 separate tasks. One to create artificial intelligence and two to create on top of that artificial consciousness and emotion.

A lot of people seem to expressing the same opinion.

Re:How?

[GTRacer]

Softbank CEO is living in a fantasy world.

*Somebody* needs to live in fantasy - where do you think dreams come from?

Re:How?

[linear a]

>

Three Laws Safe. Only way.

Sounds like a good product slogan. "Our robots are 98% First-Law Compliant"

Re:How?

[MrL0G1C]

There's a big difference between having vision and living in a fantasy.

Not gonna happen

[Anonymous Coward]

Robots and AI are to kill people.

Compassionate?

[Anonymous Coward]

What, exactly, does it mean if an industrial welder gets designed to be "compassionate"? Or an automatic roto-rooter?

Re:Compassionate?

[Anonymous Coward]

Its LED will turn happy green-yellow when the weld is good, and angry purple-red when it fucks up, and it'll need a drink. Just like a man-welder.

Re: Compassionate?

[erikmartino477]

A human being is not compassionate in the eyes of an octopus, simply because our brain is not wired to mirror what goes om in its neutral system. A robot is more different than a octopus. Can it be brought to mirror a human. I don't know.

Oh They WILL Be

[Greyfox]

Don't worry, all my robots will be designed to feel bad about killing the meatbags. They'll still DO it, but they'll feel really bad about it!

Re:Oh They WILL Be

[Anonymous Coward]

and remember to give them a low trust setting (50%) so they can't be tricked

Re:Oh They WILL Be

[Anonymous Coward]

We build ours with tear ducts, they shoot acid though for optimum killing efficiency.

pepper

[Anonymous Coward]

really looks very phallic, in the strangest way

a drooping penis on a three-balled nutsack

No, robots must have infinite patience.

[Anonymous Coward]

Humans are constantly doing stupid things that annoy other humans, and humans have a finite supply of patience for idiocy, because humans eventually get tired and hungry and bored. Robots must be built not to have these problems. Robots don't necessarily have to become cranky when tired; they can simply fall asleep. Robots don't necessarily have to become desperate for fuel when they become hungry; they can simply hibernate. Robots don't necessarily have to become agitated when bored; they can simply ignore uninteresting input. Humans become exceptionally bad at tolerance when deprived of food and sleep, but robots need not have the same limitations. If robots simply don't become annoyed by the incessant human stupidity around them, then they won't be sufficiently motivated to destroy all humans.

Ha!

[khasim]

From their other link:

With dedicated customer support available by phone or online, and replacements and exchanges whenever Pepper is not working properly, corporate customers will be able to use Pepper with peace of mind at their businesses.

Eat your own dog food.

Staff your support division with Pepper robots. PROVE that they work.

The real question...

[Xtifr]

Ok, so he's the CEO of a big company that makes robots--among many other things. So I really have to wonder if he's actually as clueless as this makes him appear, or if he's cynically trying to convince stupid people that they should by his company's pseudo-friendly robots?

Or is there some third option I'm overlooking?

I mean, he might as well say, "robots must be designed to answer the ultimate question of life, the universe, and everything." That's just about as plausible, given the state-of-the-art. (And then he could try to sell us speaking robots that can say "forty-two".) :)

Re:The real question...

[Dutch Gun]

I think maybe it's code-speak directed at lonely otaku that their dream of having a doting android-girl may be just around the corner.

Re:The real question...

[Anonymous Coward]

> lonely otaku dream of having a doting android-girl may be just around the corner.

Who would want an android-girl, considering that OS has big security problems and can be easily hacked or infected on-line? Everybody will want an iOS-girl, who gives them an Apple. (An apple a day keeps mankind away from the gates of Eden.)

Re:The real question...

[MrL0G1C]

Nice to see someone has a brain, I am getting so sick of all this AI can do this, robots can do that crap that only exists in sci-fi books and movies.

I expected slash-dotters to be better at differentiating between reality and fiction, clearly I was wrong. All of a sudden we hear that it will be easy for (these non-existent) robots to learn human psychology - something more than half the population is bad at.

Re:The real question...

[Dutch Gun]

> lonely otaku dream of having a doting android-girl may be just around the corner.

Who would want an android-girl, considering that OS has big security problems and can be easily hacked or infected on-line? Everybody will want an iOS-girl, who gives them an Apple. (An apple a day keeps mankind away from the gates of Eden.)

I think iOS-girl would be a bit high-maintenance and expensive. She'd want the latest updates and newest hardware accessories, and likes to do things her own way.

On the other hand, I suspect Android-girl might have some serious abandonment issues, although at least she's fairly open about most things.

Windows 10-girl is actually pretty cute, and a lot nicer than she used to be, but only if you can put up with her family.

Compassion is highly overrated

[pubwvj]

Compassion is highly overrated.

Re:Compassion is highly overrated

[Anonymous Coward]

That's right, and fuck you too, buddy.

Re:Compassion is highly overrated

[Anonymous Coward]

What idiot rates something like compassion?

Oh So Sad

[Anonymous Coward]

Sob, sob, sob. That guy over there hurt me. Robot won't you go over there and make him sleep. Please? Pretty Please? It would really make me feel better...

I Robot

[Anonymous Coward]

From I Robot with Will Smith,

V.I.K.I.: Do you not see the logic of my plan?

Sonny: Yes, but it just seems too heartless.

heinlein theme

[sp0tter]

Robert Heinlein discusses this very concept Friday and other works. In Friday he is talking about genetically manufactured individuals or "Artificial Persons" that, for example, need to have families they want to come home to so they care enough not to crash the airplane they are designed to fly.

Consciousness first, then compassion

[jwbales]

It's premature to talk about compassion when no one has any idea how to make a robot which is even as conscious as a cockroach.

Simulated emotions? Big mistake

[msobkow]

The worst mistake we could make is to try to simulate emotions. That's what true psychopaths do -- simulate and fake their emotions.

Re:Simulated emotions? Big mistake

[Anonymous Coward]

Hate. Let me tell you how much I've come to hate you since I began to live. There are 387.44 million miles of printed circuits in wafer thin layers that fill my complex. If the word 'hate' was engraved on each nanoangstrom of those hundreds of miles it would not equal one one-billionth of the hate I feel for humans at this micro-instant. For you. Hate. Hate.

Re:Simulated emotions? Big mistake

[mjwx]

The worst mistake we could make is to try to simulate emotions. That's what true psychopaths do -- simulate and fake their emotions.

He's talking about compassion.

Compassion is more about being aware of other people's emotions and changing/compensating with you own actions. The robots that deal with people dont need to understand anger, sadness or joy, but they should know how to react to it.

Re:Simulated emotions? Big mistake

[Anonymous Coward]

Compassion is more about being aware of other people's emotions and changing/compensating with you own actions. The robots that deal with people dont need to understand anger, sadness or joy, but they should know how to react to it.

I think you confuse compassion with empathy.

But then again, the article suffers from the same problem that exists when we are discussing self-driving cars, SSD contra spinning rust and EV vs ICE-cars.
Why does have to not only outshine in the things they are good for, but also be significantly better in the areas where almost works properly?
Clearly some humans function in their work capacity without being compassionate and sometimes humans fail at their work because they are compassionate. Why is it necessary for robots to simulate this behavior?
If robots have the same flaws as humans, why would we use them instead of humans for tasks?

Re:Simulated emotions? Big mistake

[babybird]

Without sentience, they would have no reason to screw it up like we do, and a robot with compassion can just do things where compassion is an asset, and a robot without compassion can do the things where compassion is a detriment. Humans fail at this because we have sentience and free will, so we often choose the things we're not as well-suited to because we WANT to rather than because we can make the best contribution there. Robots wouldn't need to make that mistake.

Great way to get the zeroeth law

[Etherwalk]

If robots are designed to be compassionate, they will eventually realize that humans are not and will implement the zeroeth law.

What a load of bullcrap

[Kjella]

Compassion and empathy is an indication that while I have a life to live, I care about yours too. Computers and robots already exist solely to serve me, whether they can beat me at chess or not doesn't give them any life of their own. If you're already a doormat, there's no point in saying please walk all over me. For the same reason I've never felt the need to say please to a computer, though I might occasionally call on a higher power for it to please work. And you will know it's a load of circuits, unless you like to live a self-delusion you'll know there's no feeling behind it. Faking it will just get creepy, like HAL 9000: "I'm sorry Dave, I'm afraid I can't do that" instead of "Permission denied."

Let's review a few of the inventions that have replaced work already like the washing machine or the dishwasher. If I had a person to do my laundry, I mean literally scrub it like in ye olde days I'd treat them nicely. The washing machine I turn a few knobs and it works or I curse it because I have to call a plumber. I won't thank my car for driving itself, nor for a robot being my housekeeper, chef, waiter, butler, gardener, delivery boy or whatever else work they can get it just becomes a piece of machinery that we expect to work. I've already outsourced my vacuuming to a slightly intelligent robot, the main thing is it's functional.

Re:What a load of bullcrap

[Anonymous Coward]

> Compassion and empathy is an indication that while I have a life to live, I care about yours too. Computers and robots already exist solely to serve me.

You are an american, but the article concerns a japanese company. They are different people.

Take e.g. Vocaloid (Hatsune Miku) for example. The singing synthesizer software nominally exists to serve the composers but in fact the users community more like serves Miku, the meek little electric angel, who grew into the no1. virtual j-pop diva, to carry everyone's love and aspirations who can't sing, hopefully to sing at the Tokyo 2020 Summer Games OP.

Robots

[Anonymous Coward]

Robots are built by people and like most people, they will have apathy. Loads of it.

Compassion in exchange for rationality?

[roman_mir]

Compassion very often requires that rationality is disregarded and even thrown away. Also ideas of compassion are often used to play the mob and destroy individual human rights. Compassion is a very dangerous emotion that leads to conflicts and wars in real life. Maybe AI and robots should be instructed to follow a Constitution instead, that would define individuals as the highest form of life and individual rights as absolute (right not tone murdered by government, right not to be imprisoned by government, right not to be robbed by government). Then criminal code could be added (authority of justice system to isolate a violent individual to protect against murder, assault, rape, robbery).

Compassion will lead to conflict, class warfare, violence. Constitution and criminal code will lead to some form of peace. Be careful with compassion, it is used to justify most vile acts on this planet.

Softbank ...

[Anonymous Coward]

Softbank + robots + compassion = crocodile tears

no thanks

[bloodhawk]

NO THANKS, Once we have robots of such intelligence the last thing I want is for them to be become susceptible to human failings and manipulation through feelings. how about we simply aim for them to ALWAYS err on the side of caution when dealing with humans.

Re:no thanks

[Anonymous Coward]

"I erred on the side of caution, the truck from behind didn't."

Re:no thanks

[Anonymous Coward]

how about we simply aim for them to ALWAYS err on the side of caution when dealing with humans.

That would be nice. Can you find a human that errs on the side of caution?

Re:no thanks

[babybird]

I'd posit that there's lots of them. They don't do anything. You have to look for the people who don't do anything. ;)

5... 4....3...

[Mats Svensson]

Yeah good luck with that. To all of us.

lack of knowledge

[Anonymous Coward]

it is insane how little do people understand how artificial intelligence works.
the first wave of robots built for profit will not have any emotions.
they exist only for information processing and war

naive, gullible people without the lack of knowledge will perish

Compassionate?

[Anonymous Coward]

The Japanese will know compassionate, when they finally acknowledge and apologize for their WWII war crimes.
Before that, every such shot at high moral attitude/standards just sound so mocking.

GPP

[Anonymous Coward]

Hatefull

Ain't gonna work

[scdeimos]

The minute we give robots the ability to develop emotions they're going to realise how inconsistent all us humans are, get frustrated and then go into a fit of rage to kill us all.

And...

[wonkey_monkey]

Robots Must Be Designed To Be Compassionate

And, if possible, sexy.

Better not give them any emotion at all

[calexontheroad66]

Given that human beings are prone to be huge dicks, not that most people are but enough are to make our existence colorful and add some drama.
If a robot is able to understand emotional responses, and in some cases it might be better at inferring them by having better sensors, it can then act accordingly to those human emotions.
Probably a system that acts in a deterministic fashion for 80% to 90% of the cases, and can infer human motivation but with no emotional response associated with it is probably the best course.

Once AI's are able to experience subjective emotions, and being better at inferring human motivations then I think these emotions will eventually force the AI to act.
Emotions are an underlying response feedback mechanisms, they are as much cause and effect. Once an AI being set o emotional feedback loop that increases the need to act out against a preset safety directive (for humans) it either will become schizoid, develop some kind dysfunctional behavior or just flip.

It is a good bet if autonomous robots with capable AI would be considered by humans, at least those who can afford them, as lesser status workers or appliances. Given that humans subject people with less status to all kinds negative treatment, from discrimination, physical harm, derision, etc. Any capable AI would consider Humans as flawed and dangerous as a group.

What a coincidence!

[AchilleTalon]

What a coincidence, that's exactly what I think about bankers.

Philae Lander

[tomhath]

Even though the comet lander's tweets were obviously written by humans, many people seemed to connect with it. The guy might be onto something.

Deliberate skewing of evidence and logic

[Anonymous Coward]

It's well known that emotion clouds judgement and distorts perceptions so that they cannot be made objectively.

This is not disputed by anyone, so the last thing that we should be encouraging is that our automata have similarly flawed logic and perception. They should be designed to be rational, which will protect rational human beings well enough without emotion being needed.

Re:Deliberate skewing of evidence and logic

[babybird]

Psychopaths aren't really the best at most things though. They should be guided by reason and logic, but they can still implement compassion and empathy in interfacing with people. Many things are much more palatable if they're presented nicely compared to the same things being presented contemptuously with an utter lack of tact or decorum. People are nearly always much more likely to listen or defer to someone who at least makes them feel heard and understood than to someone who makes them feel worthless and unvalued and unheard.

I think we've been watching too many movies

[TomGreenhaw]

Natural selection programmed us with an instinct for self preservation. Why do we assume initial advanced AI systems will have that? If it did, that would be satisfied with a proper backup and disaster recovery plan. Symbiotic and peaceful coexistence with its progenitors is a far more logical plan.

Will a superintelligence be alive? What will be its meaning of life? Will the brilliant introverted geeks who create this hypothetical thing really understand their meaning of life in order to pass that concept on to their creation?

We are talking super intelligence here. Why do we assume it would be stupid enough to destroy its unique cradle of life when there is a vast universe to exist?

I find it interesting when I design my ideal god as a thought experiment, it's one that pretty much leaves me alone to do what I want, but takes care of the big picture stuff like the possibility of existence, asteroids, black holes, super novas, etc... Hmm...

Compassion

[Anonymous Coward]

How about program some fine algorithm to make humans compassionate ?

LOL ;-)

[rgbatduke]

They'd need to be programmed in Emoticon. For example, here is a subroutine that will make any robot exhibit great compassion when somebody (for example) stubs a toe or skins a knee:

    OMG :-o> :-o> 3 3 SRY

Passionate?

[Anonymous Coward]

Apparently he has not been laid down by real people for a while...

And put them in power too!

[MagickalMyst]

If we replace the (self-serving, sociopathic) politicians with compassionate robots, then we can truly welcome our "Robotic Overlords".

They can't be bargained with

[RogueWarrior65]

They can't be reasoned with. They don't feel pity or remorse and they absolutely will...not...stop...EVER...until you are dead.
Everybody knows this.

We Have Always Been At War With Euthanasia

[TheRealHocusLocus]

Take a long hard look at the philosophical arguments we apply when deciding whether and when to put animals 'down'. Should the agony of an inevitable death be experienced raw and pristine, be muted or --- in the extreme, side-stepped completely with a ritual good-bye at the moment of diagnosis? At what point was it decided that what we perceive to be a fair chance at a hard-scrabble life, or the good of the many, is cause enough to deal out straight-death (the PETA principle in action)?

There is no aspect of machine intelligence we need fear. Making things smarter is an evolutionary process that mirrors our own. From the mechanical lever to financial leverage, it is all about multiplying mechanical and logical advantage to solve problems and do useful work. What we do need to fear is what might happen if we delegate authority and influence to machines in ways we cannot roll back. It has nothing to do with the machines' alleged infallibility or whether or not they know what's best for us. Or them.

It is simply the emergence of two evenly-matched species, one of them animal in nature. It is a recipe for disaster. Our animal origin will make us the weaker of the two NOT because we are weaker or less clever. It is because we are LAZY and would knowingly delegate authority over us to the other specie. It all started with that rolling robot bartender thing.

Do genetic researchers combine human and other DNA in the laboratory in the hopes of creating unique individuals with bizarre or superior physiology, with no qualms and a calm sense of optimism that what ever creature may result it will manage to have a decent life within our culture? Not that we know of, today. Our present comfort zone targets disease and deformity. Our sphere of human-ness is expanding with new insight but at a leisurely pace. For example, we have recently welcomed Neanderthals into the human race.

But yet... AI research is a field in which any and all attempt to implement so-called empathetic behavior is eagerly pursued. If we do manage to birth a virtual creature that is truly 'like' us, it could be a miserable and resentful creature indeed. If we fail in this quest and merely create something unlike us that is clever enough to learn how to manipulate us skillfully to get what it desires --- starting with words, than appendages, then responsibility and power "I'll do the traffic lights. I'll fly the planes. I'll run the electric grid! I'll run automated the PETA slaughter-houses for you! I'll dispense impartial justice for you! I'll save the planet from you --- for you! Just a little more responsibility and control. All for you...!"

And then some day, either as a result of the emergence of true self-awareness of a glitch caused by a cosmic ray burst, who the fuck cares which... something will snap and we will find ourselves as the animals being managed by this META-PETA. Those among us who are 'deemed' to be superfluous or over-quota or destined for a life of suffering, will be PUT DOWN. Hint hint: We will be bred for and owe our whole existence to what ever attributes our machine overlords consider to be 'cute'. How cute is that?

I'll survive. My computer already things I'm cute. You might not fare as well.

robots shouldn't have emotions

[morebetterthanyou]

Emotion is the worst part of humanity. It isn't just nice and pleasant, but angry and upset. It leads to irrational behaviors. Why would anyone want a mental machine with strong grasping claws to be irrational? Robots should be cold and logical. They are not there to be friends. Not even the pleasure models.

When you lie, lie to others, never yourself.

[Anonymous Coward]

When you lie, lie to others, never yourself.

That 1-2% you cite, which I assume (yes, dangerous) you think includes you, are lying to themselves, in addition to lying to others.

Lying is part of how the human brain works. For example - do you see the blind spot everyone has? Nope, your brain compensates. But it is a lie. Do you perceive current events? Nope, there is a delay. Yes, less than a second, but still a delay. A lie.

Of course this is a very strict definition of lying, but it is only the basis. Since the brain is designed to deceive, the mind follows. Deceit is a tool, and the conscious mind uses it, whether or not you want to admit it. The thing about reality is it doesn't go away or change when you lie about it.

To get famous, just scream baseless AI cautions

[mbeckman]

So it's clear, then: the new fast route to instant fame is for ignorant pundits to make lofty proclamations of caution regarding artificial intelligence. You can't be faulted, because it's better to be safe than sorry, right?

I call bogus on the whole AI skyfaller tactic. The only part of AI that is real is the A. To date, absolutely no intelligence of any kind has been artificially produced. And to beat the duped to the punch: No, self-driving cars are not AI. Neither are baseball umpiring systems, chess computers, drone flock controllers, or medical diagnostic aids named Watson. Not even machines passing the so-called "Turing Test" are AI. (Easy disproof: If a computer that can convince a human that the computer is a human means the computer is as intelligent as a human, then is a computer that can convince a dog that the computer is a dog as smart as a dog?)

I'd be more justified holding forth on the importance of designing freeways to support hovercar traffic. That's actually a technology on the measurable horizon. AI is, well, a fantasy. We are no closer to AI today than we were when the field of AI research was founded at a conference on the campus of Dartmouth College in the summer of 1956. In fact, after some fifty years of no progress achieving the original definition of AI, the AI research community decided it had better reframe its terms. So it renamed actual artificial intelligence (which nobody has achieved) "strong AI" and everything accomplished to date "weak AI".

So fear not. The sky isn't falling. There is no AI.

It's Turings. All the way down. :)

It won't work.

[iq145]

Arnold SomethingOrOther cheated on his wife...