 I must admit, first of all, that I don't really know the answer that is to the question that is in the title of our session, what does it take to build an intelligent machine? But I would say that an intelligent machine should be able to sense the environment and to understand the environment it is, it is founded. If we look at the, at the sensorial information that our body is exposed to, that we perceive through our eyes, that we perceive through our nose, ears or tongue or our skin, the majority of this information, over 90%, actually comes from vision. I would say that it would make sense to equip an intelligent machine with the ability to see and to understand the world around it. And we humans, our visual system is extremely developed, but actually it starts evolving from more or less age zero as we are born and actually we acquire the capability to analyze visual information way before we, for example, start walking or learn how to talk. So it is so natural to us to be able to recognize objects that we are actually not fully aware of how complex this task is and what does it take for the brain to do this task on an everyday basis, basically every second that we open our eyes that this is an ongoing process. And I work in the field of computer vision where we try to replicate or imitate some of these wonderful capabilities of our visual system on a computer. And maybe a curious anecdote that the, one of the first works on computer vision actually started as a summer project at the MIT back in the 1960s where a group of scientists connected a television camera to a computer and think that in that period it was a super high tech and they tried to make the computer describe and understand what it sees through this camera. And apparently didn't work well. Even 50 years since the problem is still far from being solved. And you may ask of course what does it make this problem so difficult? Why computer vision is so difficult? So of course there are many answers to this question. One of the answers has to do with the fact that the objects that we see around us are three-dimensional and their appearance depends vastly on the viewpoint. It depends also on the illumination conditions and so on. So the same object can appear actually in almost infinitely many appearances. Just to give you an example, so I'm sure that you will immediately recognize the faces of yourself or maybe of your friends. But if you ask the computer to do the same thing, the computer will see faces that are smiling, faces that are sad, faces that turn left or turn right. And actually it will be a very difficult task for a computer to extract some features or characteristics of these images that allow to distinguish between different people versus other properties, other characteristics that can be attributed to external factors such as viewpoint or illumination. One of the classical illusions actually that visualizes how our visual system can be fooled even though we learn for many years. It takes many years of training the neural networks of our brain to deal with these different ambiguities and different degrees of freedom in the appearance of objects. It can be easily fooled by, for example, in this illusion, you can see that the same object can appear either as convex or concave depending on the illumination. An interesting manifestation of this phenomenon was when the Viking spacecraft took pictures of Mars. One of the features that was discovered there looked very much like a giant face. And actually many people for many years believe that this is an evidence of extraterrestrial life. But of course, subsequent photos reveal that it's just some arbitrary rock formation and the face-like appearance resulted only from some very particular lighting condition. So vision is really hard, not only for a machine, it is hard for humans. And we can easily be fooled and we can easily get it wrong. I'm sure that many of you have seen this movie, Minority Report, that appeared in 2002. And this is a little bit dystopian vision of how our world, hopefully not, but it might look in the 2020s. And one of the most famous scenes in this movie is when Tom Cruise is using his hands to manipulate virtual objects on a giant holographic screen. And this is actually also a vision of the producers of this movie, how our interaction with our future intelligent machines might look like. And of course the ideas of gesture-based interfaces have been around for a while. And if you want to design such an interface that is based on computer vision, you need to solve what we call the hand tracking problem. Basically you need to detect and recognize different parts of our fingers that constitute our hands. And of course the hand is a deformable articulated object. And there are many degrees of freedom, many ambiguities. For example, the fingers can be hidden from the view. So that's why this problem is quite challenging and it's notoriously hard problem in computer vision. Fast forward several years after the movie Minority Report, Microsoft came up with a very successful product called Kinect. And basically it was an add-on to their Xbox, a gaming machine, that allowed the users to control their games using their bare hands. So basically you could move your hands and animate or activate your virtual self in your game or interact in this natural way with your computer. And basically this was the same capability that Tom Cruise had in the very futuristic science fiction movie, but without any light gloves that basically made the task in the movie much easier. So in this case no gloves were, no special equipment were required to interact with the machine. And this capability came from a novel three-dimensional sensor that projected the invisible laser light on the objects in front of it. And using triangulation techniques extracted the geometric structure of these objects to the accuracy of several millimeters. And it appeared that this three-dimensional information actually solved many degrees of freedom, many ambiguities that exist in standard two-dimensional images. So suddenly the hand tracking problem becomes much easier in three dimensions because many of these ambiguities are gone. And in fact this allowed Microsoft to use, I would say rather simple computer vision and machine learning techniques to do accurate hand tracking. Now of course the principles on which the Kinect 3D sensor was based have been known in the field of computer vision for many years. But the commercial breakthrough came from the ability of Microsoft to design a device to design a product that was reliable, that was affordable. And was also suitable for mass production. So Kinect was really a revolutionary product in this sense. And basically technology that existed only in the lab and costy fortune has suddenly become a commodity. But of course it was designed for specific applications in mind. And that was gaming. And you know that these manufacturers of laptops and tablets or smartphones are fighting for every gram and every millimeter in the design of their gadgets. So nobody would really want to use a smartphone that weights a kilogram and requires an external power source. So I was involved with my colleagues in Israel in a startup company that tried to take this dream of a three dimensional vision or 3D seeing machine one step further. And we designed a technology that would allow to shrink the size of the 3D sensor to a form factor to dimensions that would fit into the display of a laptop or a tablet. And our company was acquired in 2012 by Intel. And you might have seen Intel presenting this technology that is branded as real sense at the consumer electronics show in Las Vegas. And this is actually a commercial product. You can nowadays walk into an electronics store and buy a laptop that instead of the standard webcam has our 3D sensor. 3D sensing is an enabling technology that will bring qualitatively new capabilities to our future computers to our future intelligent machines. And actually what is probably important to emphasize that these technological capabilities, they all exist today. So we are really not talking about the future. We are talking about the present. And maybe several or many products that are based on this technology will become widely accepted in a matter of a few years. So our laptops, our tablets, our smartphones will become precision instruments that measure very accurately to sub-millimeter accuracy objects in our environment, three-dimensional objects. So we'll be able to scan our objects, we can scan ourselves. We can create a three-dimensional copy of objects in our environment. So next time you want to do a selfie, forget about selfie, think about Shapy. You can make a three-dimensional selfie of yourself like this person who replicated himself in thousands of near identical copies. Next time you want to buy a dress, you can literally do it without leaving your home. You can sit in front of a computer that, of course, in the future, or maybe even in the present, will be equipped with a three-dimensional sensor. And this sensor will allow to measure precisely your body in order to suggest the size of the garment. And not only that, you will be able to see yourself virtually as if you were stood in front of a mirror in a fitting room in a standard department store. So you can virtually measure the stress on yourself and see if it fits or not. And this brings actually an interesting class of applications from the domain of virtual reality. I believe that many of you have seen virtual reality, for example, the Oculus VR glasses. And if you have not, I strongly suggest you to go to the first floor and try the virtual reality movie. It's really incredible immersive experience. However, one of the flow or one of the disturbing parts of these experiences is that you don't see your body. So you appear a body less in this virtual world. And if you look down, you don't see your legs. You don't see your hands. And of course, this is not how it should be. So with a 3D sensor in the future, you will be able to scan yourself in real time, insert a virtual copy of yourself into this virtual world. And you will be able to see yourself and interact with objects in the virtual world. So this actually allows to blend the borders between the physical and the virtual reality and think of user interfaces. What kind of an impact it will make on them. And here you can see an example from the recent concept of Microsoft HoloLens. So it's a user interface that is based on augmented reality. You can see both the virtual and the physical world at the same time. I believe that 3D sensing technology is a key ingredient that might be needed for a paradigm shift in the way we interact with our intelligent machines. Basically it will bring us closer to a more natural way of interacting with computers, replacing more traditional ways of input devices such as keyboard or touchscreen or a mouse with ways of interaction that are more natural and more similar to the way that we interact between ourselves. So to summarize, I would say that an intelligent machine should be able to see the world and in order to do it at least as well, or it may be even better than we humans do it. It should be able to see the world in three dimensions. Thank you.