The Naked Scientists

The Naked Scientists Forum

Author Topic: Are robots with emotions possible?  (Read 4802 times)

~CB

  • Guest
Are robots with emotions possible?
« on: 17/04/2015 09:29:35 »
If so, would their emotions be considered real? Let's just say they start thinking as humans do. The only difference between them and us would be that of materials... We would be natural and they, man-made. Except this I do not believe there's any other reason we could say that their emotions aren't 'humanly'. Is there?
This also leads to another question... Can, how we think be substituted by algorithms? Basically can our brain be simulated by algorithms? I have also asked this question to some of my friends... The usual answer is no. Their reasoning? Humans make irrational decisions but robots won't be able to do so. I consider that reasoning feeble. What are your thoughts on this whole thing?


 

Offline David Cooper

  • Neilep Level Member
  • ******
  • Posts: 1505
    • View Profile
Re: Are robots with emotions possible?
« Reply #1 on: 17/04/2015 18:37:05 »
Robots with neural network brains may be just like people - they will make lots of silly mistakes in their thinking and if they're closely based on the design of the human brain they may make the same claims about having feelings and consciousness.

Robots whose intelligence is more directly programmed (using conventional computers) will have to fake any feelings and consciousness that they claim to have - there is no way to interface between the information system side of the machine and anything in there that might be experiencing feelings. Given that neural networks can be simulated on conventional computers, it doesn't look as if such an interface can be built on any kind of robot at all, and by extension it looks as if it must be fake in all animals too. Perhaps there is some weird mechanism hidden in the quantum realm though where the necessary magic can be performed.
 

Offline PmbPhy

  • Neilep Level Member
  • ******
  • Posts: 2773
  • Thanked: 38 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #2 on: 18/04/2015 23:22:03 »
Quote from: Jasper Hayden
If so, would their emotions be considered real?
I'm certain that it can and will be done someday. After all, we're all robots in the sense that we're all created from atoms which are arranged in a manner which gives us life. I see no reason why that can't be duplicated in a manner other than an organic construction.
« Last Edit: 01/05/2015 04:30:21 by PmbPhy »
 

Offline jccc

  • Hero Member
  • *****
  • Posts: 990
    • View Profile
Re: Are robots with emotions possible?
« Reply #3 on: 19/04/2015 09:12:53 »
i hope my future robot is not as moody as mr right.
 

Offline Colin2B

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1921
  • Thanked: 125 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #4 on: 19/04/2015 11:27:31 »
I suspect self awareness comes above a certain level of complexity of network.
Many emotions in humans seem to be triggered by an interaction with certain hormones. I say many because I don't know enough to say all.
It would be interesting to speculate what would happen in a self learning network that would form the equivalent of emotions.
Would feelings develop as responses to situations, or would androids find it useful to fake feelings to aid interactions with humans. Certainly it's a tactic many humans follow (politicians?). A famous comedian once said "in this business sincerity is everything, if you can fake that you've got it made".

There are some interesting ideas in the Asimov Robot series (that's the Robert series aasimz!) and in books by Robert Silverberg.
Maybe we have to ask Philip K Dick's question "do androids dream of electric sheep".

 

Offline cheryl j

  • Neilep Level Member
  • ******
  • Posts: 1460
  • Thanked: 1 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #5 on: 30/04/2015 07:02:29 »
Some neuroscientists like Damasio make a distinction between feelings and emotions. Feeling being the result conscious awareness, and emotion being just another form of information. That information is a representation or evaluation of the body's physiological state, or over all evaluation of objects and situations in a particular context. Emotions may be the primary “language” of any animal that has consciousness, but no language, a way of flagging an event or an object in conscious awareness, memory, or imagination as being good, bad, delicious,  risky,  threatening,  effective, ineffective, sexy,  or surprising, without the words "good, bad, delicious, risky, threatening, surprising", etc.  What would be the “short form” for qualitatively flagging an event instead of replaying an entire memory along with its myriad associations in ones head? What would you use if you had no words to do that?

So the question might be, would a computer need to a process anything like that?
 

Offline Colin2B

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1921
  • Thanked: 125 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #6 on: 30/04/2015 10:00:21 »
So the question might be, would a computer need to a process anything like that?
If we consider a robot rather than a computer we have senses and motive power.
Being able to move potentially puts it in danger and self preservation might develop along with awareness.
Your short form might be a small subroutine or neural net response that initiated avoidance movement if it detected a rapid movement towards itself, rather than wait for full visual processing and decision making.
The interesting question might be how would it develop. Without the natural deselection of creatures that didn't have this reaction would a robot precalculate that certain situations could be dangerous or would it calculate from reading reports of human accidents?
 

Offline cheryl j

  • Neilep Level Member
  • ******
  • Posts: 1460
  • Thanked: 1 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #7 on: 02/05/2015 05:52:44 »
Stanislaw Dehaene describes a particular brainwave called the N400. N refers to its shape, 400 to its peak latency, 400 milliseconds after the stimulus. It occurs when there is a discrepancy or contradiction of some kind - such as words that don't normally go together like "happy war"  - something surprising that violates expectations. If surprise is an emotion, I can see why a robot might need this emotion.
 

Offline Colin2B

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1921
  • Thanked: 125 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #8 on: 10/05/2015 22:00:25 »
If surprise is an emotion, I can see why a robot might need this emotion.
Did t reply straight away, I've been thinking about emotions.
Surprise did surprise me, I always thought of emotions as driven by hormones eg fear, anger, etc.
I haven't yet come to a real conclusion on what constitutes being included as an emotion. From what I've read there is not a clear cut definition.
If we could agree a list of emotions we might be able to work out whether a robot could find them useful and so develop them
Anyone any ideas for the list?
 

Offline cheryl j

  • Neilep Level Member
  • ******
  • Posts: 1460
  • Thanked: 1 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #9 on: 11/05/2015 04:17:58 »
Well, this is Antonio Damasio's list:
Primary or universal emotions- happiness, sadness, fear, anger, surprise, disgust.
Secondary or social emotions - embarrassment, jealousy, guilt, shame, empathy, pride
Background emotions - malaise, calm, tension, excitement.

The first six are based on research on facial recognition and expression. I don't think the list of secondary or background emotions is meant to be complete, as he mentions additional ones in other parts of his book.This may sound like psychobabble or sort of literary, but it is possible to correlate activity in specific regions on brain imaging with emotions people report experiencing.

 Perhaps there are as many subtle nuances of emotion as there are blends or shades of color. Things like nostalgia or ennui or "shadenfreude". Some may be so nuanced that not every language has a word for them. I thought this article was pretty entertaining.

http://www.popsci.com/science/article/2013-01/emotions-which-there-are-no-english-words-infographic
Among my favorites:

Pena ajena (Mexican Spanish): The embarrassment you feel watching someone else's humiliation

Arigata-meiwaku (Japanese): An act someone does for you that you didn't want to have them do and tried to avoid having them do, but they went ahead anyway, determined to do you a favor, and then things went wrong and caused you a lot of trouble, yet in the end social conventions required you to express gratitude.

Yoko meshi (Japanese): literally ‘a meal eaten sideways,’ referring to the peculiar stress induced by speaking a foreign language

Age-otori (Japanese): To look worse after a haircut

Gigil (pronounced Gheegle; Filipino): The urge to pinch or squeeze something that is unbearably cute

 
 

Offline Colin2B

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1921
  • Thanked: 125 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #10 on: 11/05/2015 04:52:41 »
Perhaps worth looking at the primary and secondary ones. I'll have a think.

For example, surprise. A few posts back I was wondering whether a mobile robot (android) might develop a response to rapid movement in order to avoid danger, before the neural net does any detail processing. Would high speed processing of visual images and evaluation of tactical response mean an automatic response was unnecessary?
However, if we look at an evolved surprise response we tend to freeze when something surprises us. A good forest response as you don't want to be seen, but if the surprise is an oncoming car ...
Another good survival response is disgust, in it's basic form it keeps us away from dangerous bug ridden objects.
Would an android need to develop these emotions? Or would a complete database of dangers + fast processing be enough?
Maybe any automated response cuts down processing requirements and gives more power for other tasks?


 

Offline cheryl j

  • Neilep Level Member
  • ******
  • Posts: 1460
  • Thanked: 1 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #11 on: 11/05/2015 05:07:35 »
Yeah, I wondered that, too. For animals, surprise does seem to be an alarm that says "stop what you're doing, cancel that next action, or focus attention on this - because somethings not what I expected, and I need more time to figure out why, and what to do next."
« Last Edit: 11/05/2015 05:09:13 by cheryl j »
 

Offline David Cooper

  • Neilep Level Member
  • ******
  • Posts: 1505
    • View Profile
Re: Are robots with emotions possible?
« Reply #12 on: 11/05/2015 19:03:59 »
Primary or universal emotions- happiness, sadness, fear, anger, surprise, disgust.

Happiness is a reward for doing something that's likely to be beneficial, but it can also put you in danger. Sadness is a punishment for doing something that does you damage, but it also occurs where there is nothing to gain from it. Fear is there to keep you out of dangerous situations, but it can also put you in danger if you're climbing a cliff. Anger is for social interaction and to help you put all your effort into getting rid of a problem (like throwing a perfectly good computer out of a window when the software misbehaves). Surprise is designed to make you think, though there may not be enough time to do so usefully. Disgust is designed to keep you away from things that might make you ill, such as a woman who wears a colour of lipstick that doesn't look natural or who has turned her eyebrows into tadpoles or lines of black goo.

Most of these are advantageous overall, but damaging in many situations. They exist to help steer thinking in what is most likely to be the right direction (on average), but they are not based on detailed calculation. There are many times when thought should go against feeling, but in situtations where there isn't time to do the calculations, the best chance of survival is just to go with feelings - deaths will be less common that way. A robot which is in a situation where it is about to be wiped out be a fast moving object may not have time to calculate the best way to get out of the road, so it could be sensible for it to have a set of standard programmed responses to handle those situations so that it can do something that gives it the best chance of survival - jumping in the air will often work for a small device, and that's what mice do when they hear a possible predator creeping up on them in the dark, landing somewhere the predator won't be able to predict. In robots though, there will be no feeling - just programmed instinctive behaviours.

Quote
Gigil (pronounced Gheegle; Filipino): The urge to pinch or squeeze something that is unbearably cute

That maybe explains why Tess tickles have to be so ugly.
 

Offline Colin2B

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1921
  • Thanked: 125 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #13 on: 12/05/2015 16:05:45 »
Interesting thoughts.

there will be no feeling - just programmed instinctive behaviours.

I was wondering whether a learning robot would develop the equivalent of these reactions. Without the selection of evolution, would it watch the internet and see the dangers to humans and so develop subroutines to handle problem situations; or might it think through dangerous scenarios.
Whichever, it is likely that the 'emotions' and responses could be very different to those that have developed in humans, given the differences in processing speed?

 

Offline David Cooper

  • Neilep Level Member
  • ******
  • Posts: 1505
    • View Profile
Re: Are robots with emotions possible?
« Reply #14 on: 12/05/2015 18:13:06 »
We don't really want robots to have the same instincts as people, because in a survival situation it would be better if the robots don't try to save themselves at the expense of people - they should be programmed to put themselves in danger to protect people every time. If we just leave them to learn from watching video, they aren't going to empathise with anyone and won't think of themselves as people, so they need to be programmed to work on the basis that people can be harmed and that that is something they need to try to minimise, although they must allow harm which people are willing to go through where it's necessary as a route to getting enjoyment out of life - it is vital that they understand why they shouldn't humanely destroy everyone in order to eliminate all suffering. Robots will also need to understand that the rules are different for wild animals - they are allowed to kill and eat each other under the rules of "might is right" as they are not intelligent enough to be able to follow moral rules like we can. However, they could step in to kill prey humanely in situations where animals are being eaten alive (e.g. by wolves).
 

Offline Colin2B

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1921
  • Thanked: 125 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #15 on: 13/05/2015 07:07:28 »
We don't really want robots to have the same instincts as people, because in a survival situation it would be better if the robots don't try to save themselves at the expense of people
I agree, but the thoughts were more about what if they were not so programmed. Or what if there is a loophole in the programming, many programs do have errors.
I assume you have read Asimov's Robot series, very amusing, but does show some of the pitfalls of preprograming a complex system. We don't seem to be able to get it right even for quite simple systems!
 

Offline David Cooper

  • Neilep Level Member
  • ******
  • Posts: 1505
    • View Profile
Re: Are robots with emotions possible?
« Reply #16 on: 13/05/2015 18:01:39 »
I have not read Asimov, and I don't watch films about A.I. either because they annoy me by getting things so wrong - I hear enough about them from other people to know not to bother with them. They draw attention to simple problems which won't happen because any programmer capable of creating AGI will not be so stupid as to make them, and they propose solutions that don't work (e.g. Asimov's laws would make an armed robot stand back and watch people being murdered by a gunman rather than shooting the killer).

You say that "we" don't seem to be able to get it right even for quite simple systems, but that isn't true. A lot of systems don't work because the programmers are sloppy - companies bribe governments to give them contracts and then fail to produce the goods because they are incompetent. Operating systems are generally cobbled together by large teams and it's hard to get everything to mesh together correctly once they're too big for one person to keep control of, but they are also involved in a continual race which means that things are rushed out without being fully debugged - they can be mended later when the problems arise if they cause enough trouble to be worth fixing. There are other programmers who have written operating systems that don't crash and apps with no bugs, but they are behind the race because they are unable to keep up with the people who take shortcuts. AGI will be different though, because as soon as you have human-level intelligence in a machine, it can become an expert programmer and be a perfectionist, making sure that all the code it runs on is bug-free. At that point, the only thing you have to worry about is how it calculates what is moral, so that's the thing people need to focus all their attention on. If you hunt around on the Net, you can find the leading experts on this (Facebook has some groups for this), but you might be shocked at the level of stupidity on display there. I've given up trying to influence them, because they don't understand the meaning of simple words like "harm", "suffering" and "sentience". It's almost as if they are robots themselves.
 

Offline jccc

  • Hero Member
  • *****
  • Posts: 990
    • View Profile
Re: Are robots with emotions possible?
« Reply #17 on: 13/05/2015 20:45:47 »
if one can think, one will have emotion.

if robots follow logic, they will have emotion.

too much emotion, will delete stuff.

 

Offline David Cooper

  • Neilep Level Member
  • ******
  • Posts: 1505
    • View Profile
Re: Are robots with emotions possible?
« Reply #18 on: 13/05/2015 21:53:35 »
if one can think, one will have emotion.

if robots follow logic, they will have emotion.

How will they detect this emotion stuff to guage how they feel? It's easy to imagine something experiencing feelings, but how does that get translated into data documenting the experience of those feelings? Where is the "read qualia" machine code instruction?
 

Offline chiralSPO

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1880
  • Thanked: 145 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #19 on: 13/05/2015 23:08:02 »

How will they detect this emotion stuff to guage how they feel? It's easy to imagine something experiencing feelings, but how does that get translated into data documenting the experience of those feelings? Where is the "read qualia" machine code instruction?

I think that is a very good point. My suspicion is that the more self-analytical  software and hardware the robot has, the more likely it would be to have "emotion" (dare I say self awareness?) even if those emotions are nothing like anything we experience. Whereas we may experience having a "pit in our stomach," a robot might sense heat in its predictive processors, and associate that with something akin to "anxiety."

We are able to perceive both internal and external stimuli, and I think that our language has somewhat misdirected our perception of "qualia" A whole system of primary and secondary (and so on) neurological and biological responses to stimuli causes us to "feel" a certain way, and we come up for a word that describes that feeling, like "angry" but I don't think that there is any part of our own code that says "set anger = 1" In fact there are many different flavors of anger, and often we don't even realize that we are angry (or how angry we are), which indicates to me that the emotion is a construct rather than a fundamental thing, and it is really the sum of many physiological and neurological responses that gets bundled somehow...
 

Offline cheryl j

  • Neilep Level Member
  • ******
  • Posts: 1460
  • Thanked: 1 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #20 on: 14/05/2015 00:40:55 »
if one can think, one will have emotion.

if robots follow logic, they will have emotion.

too much emotion, will delete stuff.



I don't think information processing necessarily results in an emotion, but I do think emotion is a form of information. ("Feeling" might be another kettle of fish. ) But if emotion is a representation or evaluation of internal and external states, a robot might well have other ways to form those representations, and therefore not need emotions.
« Last Edit: 14/05/2015 00:44:50 by cheryl j »
 

Offline Colin2B

  • Global Moderator
  • Neilep Level Member
  • *****
  • Posts: 1921
  • Thanked: 125 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #21 on: 14/05/2015 00:47:02 »
My suspicion is that the more self-analytical  software and hardware the robot has, the more likely it would be to have "emotion" (dare I say self awareness?)
I suspect that self awareness comes above a certain level of complexity. I think you are right that whatever 'emotions' evolve they are unlikely to be similar to those we experience, but some might appear similar.
I must confess that experience has taught me to be suspicious of the concept of perfect programing.
 

Offline David Cooper

  • Neilep Level Member
  • ******
  • Posts: 1505
    • View Profile
Re: Are robots with emotions possible?
« Reply #22 on: 14/05/2015 00:56:05 »
x86 code: EB FE (235 254) - this will never crash unless there's a hardware fault or disruption from strong radiation. You can run this on a machine for decades without it failing. It is bug-free. It is possible to write longer programs that are just as safe, except that the risk of a hardware fault is higher the more bites it has to use to hold the code and data, but again a machine running good software can run continually without fault for decades. How complex does the code need to get before bugs must creep in? If you design the code carefully by taking all possibilities into account at every step, there is no point at which you cannot guard against bugs creeping in - it's just a matter of rigorous checking, and while human coders eventually get tired and make errors in this, AGI won't. It will be able to ensure that it is running on perfect code.
« Last Edit: 14/05/2015 18:08:41 by David Cooper »
 

Offline cheryl j

  • Neilep Level Member
  • ******
  • Posts: 1460
  • Thanked: 1 times
    • View Profile
Re: Are robots with emotions possible?
« Reply #23 on: 18/05/2015 13:22:01 »
http://www.sciencedaily.com/releases/2015/05/150514132907.htm

I thought this was an interesting article, particularly their attempt to separate emotion from feeling in the experiment, and view it as "a type of internal brain state with certain general properties that can exist independently of subjective, conscious feelings" and "behaviors that express those states as a readout." They characterize emotion as having five basic components or "primitives" - valence, persistence, scalability, generalization to different contexts, and tran-situationity.

" 'These experiments provide objective evidence that visual stimuli designed to mimic an overhead predator can induce a persistent and scalable internal state of defensive arousal in flies, which can influence their subsequent behavior for minutes after the threat has passed,' Anderson says. 'For us, that's a big step beyond just casually intuiting that a fly fleeing a visual threat must be 'afraid,' based on our anthropomorphic assumptions. It suggests that the flies' response to the threat is richer and more complicated than a robotic-like avoidance reflex. '"

" 'Our work can get at questions about mechanism and questions about the functional properties of emotion states, but we cannot get at the question of whether or not flies have feelings,' Gibson says."

So would a robot need a program that has these kinds of traits of emotional states?
 

Offline David Cooper

  • Neilep Level Member
  • ******
  • Posts: 1505
    • View Profile
Re: Are robots with emotions possible?
« Reply #24 on: 18/05/2015 21:43:45 »
What is fear for? It puts an animal into a state in which it is devoting more of its thinking apparatus to potential threats. A robot should do the same when it is in a dangerous situation (in which it may need to protect not only itself, but more importantly to protect the people it may be with). Instead of using processor time to think about a wide range of things it sees and hears around it, it should be cutting off most of those thoughts short and putting more time into thinking about anything that could be related to the current threat. This mode could be entered on hearing a gunshot (amongst many other ways), or anything that sounds like a gunshot - the robot should be putting more resources than normal into looking for signs relating to that danger, but over time this mode can be modified if it looks as if no one nearer the source of the sound looks worried, and eventually it can return to its normal thinking mode. Emotional states can be seen as modes of operation which prioritise certain tasks over others - there's no point in the robot analysing poetry when there might be a gunman on the loose. In this "fear" mode, there might also be a wider range of defensive instincts in place and ready to be acted on with less delay than would occur from some other mode. What would be missing is any actual feeling of fear, but everything else could be present.
 

The Naked Scientists Forum

Re: Are robots with emotions possible?
« Reply #24 on: 18/05/2015 21:43:45 »

 

SMF 2.0.10 | SMF © 2015, Simple Machines
SMFAds for Free Forums
 
Login
Login with username, password and session length