Will Robots Finally Feel?
A common issue that arises in discussions about A.I. is that of emotion and sympathy. Everyone is aware of the fear that A.I., whether robots or software could possibly turn on humans if they get intelligent enough to attain sentience - or even before then. But just as likely as that is, is it not also possible for them to feel empathy and maybe even care for us and maybe even each other? This is a subject brought up in various works of fiction containing robots with feelings such as I, Robot, WALL-E, and Detroit: Become Human.
Well, it seems that robots are being built which “feel pain” depending on the level of impact they receive. This is done by connecting sensors within an artificial soft skin to the robot, which in turn displays different emotions based on how hard it is hit. Affetto is a robot housed at Osaka University in Japan, where a team of engineers has enabled this almost too real-looking child’s head to make a number of facial expressions in response to the signals. This is the first step towards empathetic robots that care for other beings, living or engineered. They would be invaluable in places such as hospitals, daycares, and homes for the elderly.
Although one can argue that this is all pre-programmed, another article by Tom Siegfried provides a relatively simple solution to this problem: program the robot to sense danger to its existence. “It would then have to develop feelings to guide the behaviors needed to ensure its own survival,” say neuroscientists Kingson Man and Antonio Damasio. The key to this solution lies in homeostasis, the tendency for life to be maintained within a narrow range of suitable conditions, such as regulating temperatures and balances of internal chemistry. If the machine is aware of its internal state and the changes that occur regarding that state, then it effectively has the robotic version of “feelings”.
Artificial intelligence has so far been mainly constrained to performing designated tasks, but it still arguably lacks the human intelligence that allows us to adapt to different situations, even if we’ve never been exposed to them before. Robots are consistently supplied with the things needed for them to operate (instructions, power, sufficient environmental conditions) so there is no need to concern itself with self-preservation. Feelings are the missing ingredient that will ensure that the robot is aware of its vulnerability, and acts accordingly. Two main areas are especially important to this breakthrough according to Man and Damasio: soft robotics and deep learning. Deep learning is particularly essential since it emulates the human nervous system and allows for recognition of patterns within patterns and learn to relate external conditions to internal feelings. A robot that protects itself will eventually be able to learn to protect - and perhaps even care for - humans.
This comment has been removed by the author.
ReplyDeleteI wonder if these types of robots will consider adopting any of the three rules of robotics from I, Robot. If we want robots to act as human as possible in terms of emotions and self preservation, should they even follow those rules? Humans don't follow the three rules of robotics, so why should AI if they are supposed to act as humanly as possible? If they truly have deep learning capabilities, will they somehow learn to break those three rules. And if that is the case, will we as humans be in danger?
ReplyDeleteThat raises the interesting duality of the choice: If we want robots to be as close to human as possible, then they should have both good and bad qualities of humans. However, there is no doubt that A.I. does pose a threat to humanity. However, is it not true that every human has the potential to become a threat as well? For maximum security and peace of mind, it is necessary that all options be considered before deciding whether A.I. could be trusted, and even then be kept under close watch.
DeleteI think that it would be interesting having AI be able to feel. Would they give different responses to the environment around them? Would these reactions be logical? As humans, when we respond to feelings sometimes our responses to them are not logical. Does making AI feel create personality?
ReplyDeleteThat child head is creepy. If we're not getting Westworld quality of hosts (in terms of looks), I'm not sure about robots in hospitals, daycare, or nursing homes. But between replicating emotions and replicating the human look, I imagine the first will be significantly harder. So, I probably have nothing to worry about.
ReplyDeleteIt's interesting to think about robots feeling as we don't necessarily know what or who they would have feelings for. What I mean by this is that I feel like there are assumptions that empathetic and feeling robots would help humans, nurturing, and fulfilling our needs. That being said, who's to say that they would instead be loyal to other AI beings?
ReplyDeleteThe concept of loyalty and the uncertainty that comes with it is a human quality. Even among humans, there are those whose loyalty is certain, those who make it appear that way, and those who leave people wondering where their loyalties lie. This demonstrates that both sides are pretty similar: loyalty to humans ensures safety, and loyalty to other A.I. also shows that they are capable of human qualities.
DeleteThis is an interesting topic, it makes me think of the whole what makes us humans different from other organisms and even AI. And it raises a question as to how far are we going to push AI to be more like humans. But all in all its an interesting to explore, hopefully we see some AI reaching that level.
ReplyDeleteI'm not too sure about we're an unique organism in terms of the ability to feel. The only way to 'know' what another human being is feeling is through pure guess. We're not animals so we don't know if they have the ability to feel or not. I strongly believe that animals have feelings too. We can see that in nature when a mother of one species has her child taken.
DeleteTying back to A.I., I think it's really hard to know if an A.I. is feeling something, especially with self-evolving algorithms like machine learning and genetics algorithm. We don't really know how the results are created and I don't think we would know what A.I. thinks at that point...
That sounds pretty human, if you ask me. If A.I. is left to its own devices and allowed to make choices for itself based on what it thinks is important, whether that be past events, experience, or what would benefit the most people, how is that different from being human?
DeleteThat logic looks programming possible, but feels more like an animal instinct to me, it's very rational and could not solve complicated emotions like romantic. Rational also makes it feel safe though.
ReplyDeleteI don't think that robots will be able to feel. It is something exclusive to animals. Feeling is really irrational and ambiguous, and not a thing that is programmable or can be learnt.
ReplyDelete