Department |
People |
Curriculum |
Research |
Resources |
Faculty Current Students Alumnae |
|
|
||||||||||
|
|||||||||||
|
Course Navigation
Week 2 (September 10, 12)
Timothy AmbrogiI was wondering about the concept of emergent systems, as opposed to learning systems with a data set. In particular, the example of www.20q.net springs to mind. While it clearly has its own knowledge base, it seems achieved it mostly through interacting with its environment (people on the internet) and learning from their responses. Is it simply that the people are serving to broaden the knowledge base, and that the technique is not emergent at all? Is learning always referred to as emergent, or as knowledge-based? I guess it's a question of language and definitions more than anything else...On the topic of AI in general, learning like a human is one of the things that impresses me most in an AI system. The concept of a state/memory that can be used in future decision-making feels more to me like "intelligence" than most of the examples I have read about and seen in the video. It seems that as Prof. Kumar added layers to the S-R diagram, it closer approximated the thing that I consider to be true "intelligence". Adaptivity and self-sufficiency seem to be underrated elements of intelligent systems. Also, it struck me as ironic that when we made our robot, we first gave it a goal (moving), and then proceeded to teach it futility (the robot stops trying to move when it runs into a wall). P.S. On the topic of emotions: Emotions are little more than chemicals that steer the will of lifeforms to act in a way that promotes survival and prolongation of lifetime. Just because the medium (blood) and the expression (indirect chemical reactions in the brain) are different, doesn't mean that the net result isn't the same. Also, allow me to pose the question of what happens if you perfectly emulate the chemistry behind how emotions and chemicals influence decision-making. Is the digital representation of the system less "organic"?
Lawrence BombackWhen we kept adding more deliberations to our stimulus-response agent in class on Tuesday, I was pretty clear up until we spoke about the notion of "How happy will I be if I did ______?" This "happy" concept makes me a little skeptical. Their is a major distinction between cognition and emotion, two very different fields of psychology. I believe that computers can and do "think" according to my personal definition of the word "think." I also believe that they can think and act rationally, and that they can think and act like humans. However, I do not believe computers can actually "feel." Without three things, a brain, a body, and a social setting, I don't think it's possible to really be able to have emotions like happiness, sadness, anger, etc. Perhaps of those three, only the notion of a brain is close to that of a computer. For example, they both have memory storage, they both make decisions, etc. But, the other two pose a problem. Psychologist William James asked the question "Would emotion exist without a body to express it?" It's a very good point. For instance, I don't know if anyone would be able to interpret that I was happy if I didn't have a bright-eyed look and a smile on my face. Obviously one could say that a computer could draw this emotion, but it wouldn't be quite the same. A social setting is something that also raises suspiscion. If in a room of computers, is one arbitrary computer even aware that it is surrounded by a similar species? No. If a computer were to tell us that it was happy, would it really be happy, in the same way we express happiness? Of course not. A computer doesn't contain anything organic. It makes more sense to me for the computer to ask itself if it is making the right decision, not if the decision will make it feel a particular way.
Jacqueline ChewMany have been discussing on this forum ways to define differences between humans and computers, and that computers will never be able to match humans in terms of feeling and expressing emotions. To me, it is obvious that a computer, despite any kind of programming or future technology, is not human. I even question the extent to which we, as a society, actually desire for a computer to be more human. Computers are being designed to mimic humans, in methods of processing thoughts, reasoning, learning from experience, etc., but none, as of yet have are able to act as humans in that they can feel emotions, or that they have empathy or memories. I think that such human traits will never be found in computers partly because we, as a society, have no use for a computer that can feel true sorrow or express love or joy. We continue to develop computers that can reason and seem to "grow" from experience because computers with these abilities can be of aid to society. Such computers include the thermostat, which we discussed in class. While i have doubts that the thermostat is actually "thinking;" for all intensive purposes it is "thinking" as much as we need it to. It determines whether to heat or cool the room based on a setting provided to it by humans. We would have no use for a thermostat to, say, feel sorry for us when the temperature drops below a level of comfort, or for the thermostat to have feelings of resentment toward us for giving it such a mundane task.
Catherine ChiuOne key thing that is different between humans and computers is that humans have emotions. Emotions are not something that can be programmed and not everyone's emotions are the same. To say that a computer' choice between two options will be based on the computers happiness after the choice is made seems unlikey. Computer have the ability to evaluate what option would bring it closer to its goal, but is the computer's happiness based on if it is closer to its goal? If this is the case, then the computer's happiness just another name for the computer's process of executing its program.Computers may be able to think in the sense that they can compute what option may be better or use statistics and look at patterns to see what course of action it should take, but humans don't always do that. Humans at times take what they feel into account which makes humans fickle. Computers do not feel anything, they do what humans tell them to. They do not get anything out of completing a task, no reward, no sense of satisfaction, nothing because they have no feelings. When computers can have feelings is when artificial intelligence will be clearly realized and the lines of humans and machines will be blurred.
Jason ColemanIt seems like "emotions" are a hot topic this week. And while I believe that computers don't have emotions nearly as complicated as humans do, I feel that they can posses some kind of emotion. When we talk about making a robot happy, we're talking about giving a robot purpose, something for it to seek out and/or work for. Giving a robot this simple happy emotion that we preprogram is a good way to have it make complicated decisions... complicated decisions being decisions with more than one possible answer or response. If the robot knows what makes him "happy" and how much, he can decide between to conflicting modes of response. I think that we will be programming systems like this in our class very shortly. So even if our robot isn't crying for no good reason (perhaps an extreme example of emoting), our robots can have emotions that dictate its purpose.There are other analogous systems that could be described as emotions in robots. Consider that when we are frightened, our adreniline flows and we become stronger, quicker to react, etc... basically we become better able to handle the problem at hand. At robot similarly could sense something and get scared... realize a danger, and move into a "fleeing mode" instead of the "exploring mode" it was in. Imagine seeing this robot come close to a source of heat, slowly, slowly, and analyzing everthing, but when it gets to close and it becomes too hot... the robot would turn around and leave quickly. It would probably look like a scared little robot. Alright... I'm done.
Nicholas Kerr
Ananya Misra
James RacanelliI was surprised, as I read this weeks reactions, to see that so many people seem to think that the idea of computers being able to have emotions similar to those of humans is far-fetched. Granted, it is a far-fetched goal for computers given the current state of technology, but I see no reason why we would not some day be able to understand human emotions on a detailed enough, very mechanical and biological level so that we may be able to put the same type of mechanics into a computer. Similarly, a lot of people seem to be hung up on the whole idea of robots being able to make a decision or being able to act in a certain way that makes it the happiest. At first glance, I too thought that these types of computer procedures were very far from being similar to human behavior and that they were all just a matter of programming. But then I began to ask what it is in our behavior that is not somehow related to some sort of programming. When a person is driven by a certain emotion, then they are operating under a certain set of rules. There are different rules for each emotion, because we tend to act in certain ways when driven by any particular emotion. Those rules vary so much for different people and are so complex that it is hard to imagine what kind of internal hard-wiring can produce them, but if we some day understand these rules on a physical level then we should be able to put them in computers.
Juan RamosA lot of focus has been given to the topic of 'emotional' computers. While I think that currently we don't have the capabilities to create such a system, it seems to me that in the long run, it might be possible. However, I find that if it were to be done, it would be extremely difficult to program. Emotion in and of itself is hardly well defined in mathematical terms, and even on in more philosophical terms it is not completely understood. I think that the closest that computers might come to 'feel' emotions, at least with our current understanding, would be to create specific stimulus response agents that react to a stimulus with a preprogrammed emotional response. For example, referring back to the example of the robotic car bumping into obstacles, we could have that such a robot responds by 'feeling' or showing pain. In a more complex case, a robot that can hold simple conversations with humans might respond with a feeling of offense when the human agent inputs a controversial message. In either case, the computer is mimicking emotion through preprogrammed algorithms and stimulae. Whether one day a computer might actually feel the emotions it could display seems like a part of the question of whether we could build a robot that could mimic most human behavior and physically look like a person.
Matthew RushtonLast week when we were talking about simple stimulus response agents, a word we kept using was "decide". For example I believe we talked about the thermostat and how it regulates room temperature. The word "decide" was thrown around in regards to the machine deciding what temperature to pick. Now this seems like a misleading use of the word. In my opinion there is no decision being made, rather it is just simple physics, cause and effect, that force the machine to do what it does. There is nothing magical going on. For us humans it seems that simple cause and effect doesn't exactly explain everything that goes on when we "decide" to do something. I feel like our free will, assuming we have one, is something physics has a difficult time explaining. What allows us to choose freely between multiple options? It can't be simple cause and effect should we truly be free to think. Now when the robots begin to gain some intelligence the line of decision making does begin to get blurry. Give a robot an internal state as we spoke of and it certainly does appear to get closer to making an actual decision. I imagine there are far more sophisticated machines that blur the line even more. The reality is however that these machines all are still ruled by simple physics, cause and effect. The use of the word "decide" in describing their behavior is a bit misleading though very tempting to use.
Tina TanI agree with Larry above. I don't believe that anything unnatural can possess true emotions. It is one thing to be able to program a computer to respond a certain way to a particular stimulus but another for a computer to, on its own, decide what signal to give off. For example, one can say, if X happened, then the computer must be happy and give off a predesignated reaction. Humans are similar, we usually respond like others to particular situations. Yet, there are those humans that react differently to the same situations depended on a personal decision. But with computers, I don't know how it can be expected that a computer can decide freely whether to decide it is happy or sad or whatever with a certain program.A comment that stood out for me from the last lesson was the one about how modern planes are said to fly even though it does not model the real flight of birds. Then there is the question of whether the same idea could be extended toward AI? My opinion is that the process of free thinking and intelligence is personal and unique to humans, and that people will never really be comfortable with the idea that machines have the same capabilities as we do. I personally would not completely dismiss the possibility that computers can someday think on their own, but I can't imagaine ever seeing it as anything but artificial. ViewWiki | EditWiki | Webmaster@wiki.cs |
Course Home Page