Watch videos with subtitles in your language, upload your videos, create your own subtitles! Click here to learn more on "how to Dotsub"

TEDxLakeComo 2010 - Giorgio Metta sui robot antropomorfi

0 (0 Likes / 0 Dislikes)
Given that we are speaking of engineering I would like to tell you the story of two engineers and then we will see what this story has to do with Robotics. These two men, well if I only show you only one photograph in reality perhaps you will not understand well who they could be, but in either way they discovered a technology that changed the world. It was a technology that more or less makes us live in a slightly different way today. And if instead I show you this picture I believe that it is a lot more clear who these two men could be. They are the Wright brothers and it is more or less obvious in one way or another what is the technology that they invented. The story of engineering that is behind this invention is interesting enough and it can be revisited in a single slide that contains three fundamental messages that we can say brought about success of their invention. For long man dreamed of flying and to be able to fly by copying nature well enough. What the Wright brothers did wasn't very different. They were observing nature and, in particular, the birds’ flight, to understand how they could apply this to a machine. The observation of nature was always a fundamental aspect, let’s say, of the work of engineers. There is an English word, reverse engineering, which refers to how we find what is needed to understand the inner working of a system in order to be able to then reproduce it in an artificial device different from the original one. It’s also the same thing as when inventions are copied. This is called reverse engineering which is a way to observe a system and copy it in order to remake it with the same functionality. This doesn’t probably have a very positive connotation for the person that is copied but here we are copying nature therefore everything is ok. According to the second fundamental message in this story, the thing that brought about success was the possibility to write equations and therefore the Wright brothers did not only attempt to build wings for flying but they also began to write equations to write mathematic formulas that could forecast the behavior of a certain system. The system in this case once again was the wings, but the Wright brothers were very careful seeing that their colleagues at the time had had an unpleasant end trying to prove their inventions. They wanted to do experiments, so they constructed a wind tunnel, which was very modern for their time, where they could actually test the functionality of the wings they made. This is to say that the theory of flight should go together with, in some way, experimentation, in order to measure and to therefore verify the equations of the models. The third message is important, concerning a delicate matter linked to a problem that they had to resolve in order for their airplanes to fly securely. This problem had to do with the fact that for a certain period of time they were able to make a machine that could fly, but it only could fly straight forward. As soon as they tried to turn, new problems of stability arose. This was because in their models - we are entering a little into the technical part here - they did not realize somehow a series of effects that happened when the airplane tried to turn, for which they decoupled the problem of having to go straight in order to be stable from the problem of turning and controlling the turn. In the end they resolved the problem because they understood that in order to turn they needed to bend the airplane. This was exactly the solution that they found. What this answer tells us, all things considered, is that the problems, overall if we are talking about complex systems, cannot be analyzed alone, but must be analyzed holistically: we cannot break up the problem into pieces, isolating the pieces and hope that the final solution will still work. Or better, this approach will work in some cases, but in other cases the system is simply too complicated. These three messages, all things considered are valid even if we put ourselves in a different setting. The setting that we are talking about is about the construction of, one more time, a complex system, then what do we want to do? We want to observe the human being and the first image shows us in some way two important aspects: the observation of the adult humans, therefore of behavior and how humans grow and develop (the infants). We would like in some way to see how it functions, how behavior is controlled and therefore we must look at how the brain functions and attempt to copy it in artificial systems, we see here, that they are robotic systems. This is similar to what the Wright Brothers did, going to observe nature, looking to construct mathematical models, but also performing experimentation. Experimentation in this case is the robot, therefore not only writing equations but also making something that implements these equations and that behaves naturally. These systems we call them interactive artificial systems and we would like just that they would behave like humans. Now these images are taken from a film, in reality, we would like in some way to make systems that can help us, hopefully that can also dance with, but more than anything that can be helpful to humans, whether or not in industry, whether to go in places somewhat unpleasant for humans, or whether instead to help, for example, in the household. The robot that you see in the middle is not taken from a film; actually that is a robot that we made. It is the system that we are working on. One of the other things that we have to consider or that we ask ourselves is if this activity, that I would like to have said that in reality is new, but unfortunately is not, of studying the brain constructing it in different models, but it did not happen like this. The first image at the bottom is an image from a document of Descartes, therefore it is not exactly very new, at that time it was thought that the brain controlled movement by pushing some fluid in the muscles, in some way, and in effect causing movement. This fluidics model was the model of that time, it was the technology that one knew and therefore the technology was imported as the model for studying the brain. Then the story rapidly progressed, the second image is a model of the spinal cord, made from switches and coils, similar to the technology of the telephone, of the telegraph, of the telephone switchboard where there were many switches that were supposedly going on and off. The processing of information changed a little with the rise of the computer, therefore the other image that you see in the center represents the study of “the systems theory,” therefore the possibility to study, using equations, systems in which there is a feedback, therefore it is possible to imagine that there will be sensors, that these sensors cause a modification in the behavior of the controlled system, the robot instead of humans. The other model that you see there is a variant, if you will, of the first one, where if it is imagined that the object that embodies the elaboration of information isn’t a monolithic object at all, a single equation, but rather a compilation of basic objects, similar to neurons in the brain, in this case we will consider, let’s say, interconnections and therefore many simple systems connected among themselves. Obviously what we are doing now, the robotics, we say is the thing that arrived a little later. I want to show you very quickly, some of the ideas from which we have taken our inspiration, of which we are inspired, that in particular changed the way in which we would like to build our robots. The first example has been given the name of Mirror Neurons; it is a group of neurons found in the monkey and is very interesting, not because it has a fairly particular response, but because it was found to be responsive, or anyway that activates, or better it is a set of neurons that activate when the monkey sees a certain action done by a human being, but at the same time it is the neuron of a motor area, therefore you can ask, but why if the monkey is not moving, it is just observing a person performing an action, is this neuron activated? They are called Mirror Neurons because they are the same neurons that control the movement of the monkey performing the same action. Now the actions in particular were observations of grasping, therefore the monkey picked up a certain object, or rather the human showed the same grasping action to the monkey. The interesting thing is that this way of looking at things influenced the way in which we see another series of behaviors, for example the way in which we try to study how we perceive words. Therefore some gentlemen, our neurophysiologist colleagues, showed us that in reality, while you are here listening to me speaking, your motor areas are activating and in a way that is identical to how they would activate as if you were talking, saying the same things that I am saying. Therefore the perception in reality isn’t univocally determined by the senses, but has also to do with the activation of our motor area and this we say changes the way that we want to build our robots. With this other series of images, I wanted to express another concept that stems not from neurophysiology, this time, but from experimental psychology, that is how we perceive objects. Also in this case, in reality, two objects that for the computer could be completely the same, because they are two long objects with a form, all things considered, not very different one from the other, that in reality for us are very different, but they are different because we associate the action that we perform with these objects to the object itself. And it is evident that if we have a useless or impossible object, like this one here,in short it is clear that for us the object whether or not a computer could be entirely plausible, but for us who have to use it, it is exactly like a teapot that doesn’t do its job, or anyway, that burns us every time we try to use it. The other important element is concerning psychology, and in particular developmental psychology. I will describe it to you very quickly with a metaphor. The metaphor, let’s say, is as follows: to us it is interesting to understand, not how to create an artificial system that behaves like a human being, but to create a system that, when put into the world, develops just like a human being. This could be a much simpler problem, also because studying development shows us, still using a metaphor, how we or how the brain structures develop to determine our behavior. In particular, studying development shows us a whole series of structures in the brain like in the construction of a building, that are in effect scaffolding, which are structures that when you see the final product, for example the adult human being, you don’t see them. Relating this back to robotics it could be interesting to go and see these things, to study these things and that is what we are doing. The projects that we are doing are not only projects concerning robotics; we are collaborating also with neurophysiologists and developmental psychologists, as in this case. From this point on you will see what we have really done, until now nice words have been said but what have we done? We have told you these ideas, the artificial perception, the robot, we want to implement them.The idea was to make a little robot in order to test these ideas in concrete terms, here you see in the middle, some way or another, the design of this robot and the final product on the right. These are now two videos that show us even more in depth, following the same idea, the observation of nature, therefore of a biological system and its production in an artificial system. In this case the robot utilizes a series of particular neural networks, that are similar, or that one thinks would be similar, to the response of neurons of the spinal cord, the ones that cause rhythmic oscillation that can serve to control periodic movements like crawling or walking. The robot that we made has a series of sensors, this image shows you a glimpse of what there is; the eyes are video cameras in reality, the ears are microphones, it has an inertial system that allows it to balance. Therefore it has a whole series of sensors that in some way make it resemble a human. I will show you some more videos that are examples of the things that it can do, things that we have worked on a lot, for example the hands.The hands for us are the way in which the robot should interact with its environment, how it should acquire its knowledge of objects, and for this I put particular attention on the design of the hand. The hand has independently operating fingers, something you don’t see often with traditional robots, and it was a lot of work just to give the robot hands, in the end. In the other video instead you see the robot that is autonomously looking at the world around it, substantially processing the images it receives through the video cameras to search for interesting events in its environment and therefore shifting its gaze in order to observe these interesting things. This is another video, somewhat long if you will, of a series of experiments where the robot comes in contact with the world using touch, therefore it touches the table to know where it is, what is its height, and well, this without using touch, it looks at an object to acquire a visual model of how the object is made, the robot has never seen that object, he looks at how it appears, learning to recognize it in the future. The person that is in front of the robot uses a speech-based interface, through which he is speaking to the robot asking it to do certain things and here the robot again comes into contact with objects, accepting in a kind way the help of a person who explains how to hold that particular object, touching and pushing it in a particular way, there you have it. Well, to conclude, the question of “what can we do”, let’s say, whether these results are helpful, the question “what can we do” is if this technology really allows the construction of a model of the brain, if in the end they are useful to do something else, that in the end isn’t just building a nice robot. Some aspects that we might have forgotten is the form, therefore we can ask ourselves about the form, in nature we know that the form bears function, therefore as the biological systems develop, form and function are connected. We should maybe do the same in robotics, a thing that at the moment we are not doing. The development of the body, we should not only consider the development of the brain, but also the development of the body, the way in which the bones develop, for example, has to do with their use and what you do with bones, which is why they are more resistant in certain directions than others. The same thing, the processing of the information in the brain in reality comes about because there is a change of morphology, of connectivity between the neurons. This isn’t a thing that we do obviously with a computer, but can be important as a technology. Here I conclude, the Wright Brothers, what is there to this day? A hundred years of work. The dawn, if you will, of artificial intelligence of Google and that which we would like to obtain, we hope, in less than one hundred years.

Video Details

Duration: 18 minutes and 38 seconds
Country: Italy
Language: Italian
Producer: TEDx
Director: Gerolamo Saibene
Views: 246
Posted by: tradottiinitaliano on Dec 22, 2010

Giorgio Metta svolge le sue ricerche nel campo dei robot umanoidi (e in particolare nei sistemi artificiali che nel corso della loro durata mostrano alcune abilità proprie dei sistemi naturali), collaborando con i più riconosciuti scienziati a livello europeo e internazionale nell'ambito di diverse discipline come le neuroscienze, la psicologia e la robotica. È Senior Scientist presso l'IIT, Istituto Italiano di Tecnologia di Genova.

Caption and Translate

    Sign In/Register for Dotsub to translate this video.