Michael eagle sten. What is it that you do here . Background . Our physicsackground is and devices. U. C. Berkeley and a p. H. D. Iowa stateat university. I am from minnesota originally. Are here at bell labs nu and working again on . Sensors. G on optical which is what . Anything sore maybe familiar with cameras. They are twodimensional imagings. 3d opticalng on imaging technique similar to radar. Light. E arent those in practical use today already . Yes. These. Re uses of one of them is in the medical technique called top graphy and you can use it eye,00 d imaging in the for example. You can get a scaled entireion of the eyeball all come notly nonevasive just the lights. Yes they are more and more common. So what is your research doing . Well, what i am working on system andn entire shrinking it down to the size of a chip. Integrate the phones in the ture. Future. Then allow us not to monitor ourselves but the land around and three dimensions. How far away are we from that type of technology. Of challenges we still have to overcome. Buildingll the blocks that we kind of develops with the cop technology and pioneered in the past. Actually now repurposing this technology for this new technique. So in your work, completely related is that where you are headed . That is where i am headed. Really the biochemistry of the and m Monitoring Health on a daytoday basis. To me, it is one of the Biggest Challenges we face as health, is how our how we feel, um, and it is really something we monitor. Is a lot ofere opportunities and challenges to evercom to really make Continuous Health monitoring a reality. So in a sense, is it the math thaten goes into the optics . Yeah. Yeah. Really, a whole spectrum of work that needs to be done that can benones smell and integrated in the system and actually developing algorithms to use to actually diagnose things within. Doctor. Didnt have a medical background. What i do is really i make tools that can help those professions to actually job. Eir mr. Eggleston, you image. About the 3d eye where he is can this and is this being used Fire Department yeah, really a lot of use. The second most common is in skin care. Abnormalle, an growth or something in my hand. I can do a 3d scan of it until cancer, is it benign, is it something i need to look at . Nonoffensive. You dont have to cut anything off. You dont have to. Well, there is many other fields that people are using it. Den try, you can actually look teeth and of the see it before they develop insidevities and even of the body, they are using this to do surgery where they what theyg to see are doing on the inside of the body. So really, tehran g of the kind ofities is endless. Anything you want to see or this, you can use technique. How, a lot of us are familiar mris. How old is that technology. Fis it related to what you are an old well, mri is technique. Actually and it is a great technique. Is lets you scan the entire body. Then, the resolution is on the millimeters to centimeters and the entire system is massive. Something that is very, very challenging to scale down. Then, they are expensive and people dont have access to them. Obstaclese techniques can be very small allow use cheap and to use the things to do them much lower costs. So what is more important work . He the skill or the technology . Well, to me, i think, they of go handinhand. At the end of the day, when you can scale something down small prize, you can make it cheap. Then, you enable it to use audience. It is great. If we have access to our daily significantly enhanced the health. Toll me, that is helping one one of the main goals. Why i think with the technologies we have is so important. Here is ar door sign that i want to ask but. Danger invisible laserreddation happening in here . What . Well, a lot of the sensing we do is in the infrared. So it is longer wavelength of than what you can see and we tend to use highpowered system. This is just a warning that depending on what ex perms we here. Nning in you may have to wear safety goggles. Now as you can see, my door is open. Usually work. Usually we done use these lasers because we are looking at something that is used to it ishe eye so complete. Where is the layser . One now . Ve yeah. We are surrounded by lasers. Over here andu show you examples. Date back foo dont see . Yeah. Lates date back to the 50s early hes 6s. Work . Abouted on his yeah. His idea. They come in a factor like this. Laser in this. This is a laser. This is a high laser that is used for medical imaging technique. Laser is much, much smaller. The lasers can actually be, actually, even, even for example. Iphone, those in them. E lasers you can make them very, very small which is why they attractive. We can embed them in every day devices in add functionality. So, yeah, this is a laser. We have several lasers on this table here. And all different factors and packages. We put them in large bobs then are easier to handle. Before we interrupted the day today. Working son in so, i am working on a new actually implement the systems on chips so a lot of it actually when you come system integration, writing codes can and algorithms to do what you want to do because everything is tiny. Cannot actually physically change it. Be automateds to and writing code to do this. Where did your interest in topic, in these topics come from . Well, i have been lights. Ed with with lights. With lights . With lights to. Me, my vision is my number one thing. And so when i went to school shall i learned about lights and light is amazing because everything it touches it actually picksp the signature material, so every beam of light that bonses off of you and hits me carries you. Mation about so if i look at it in the actually, i can decertain incredible amount of information about from that so task. Ed to the eggleston, thank you. Thanks. Recently the communicators visited bell labs. The labs were created by at t and they are owned by nokia, mr. Kennedy, what is your position here. I am the department head. I lead the team. In a lot interested of things artificial host ofence, a whole other network algorithm wes try develop here. Before we get into those. What do you mean networks . Just a structure you can look at. Structure. We study the properties. Then we build algorithms to it isproblems whether eroding problems which is the question how you get information from within part of the net to work the other part of the network or you have a Large Network, so lots of network type problems come structuresng these of these networks and you can you should be friends with next and the other if he questions we experience on a similar basis. U mentioned Artificial Intelligence, how do you define that . That is a good one, actually. I am not sure how i would it. Ne define it in general. People tend this think about machines. And selfdriving cars and things like that. Do you tend to think of it a mathematician, so Artificial Intelligence right problem andnt where we put in a whole bunch of input that is labeled and machines or mathematical models that do a of answering questions, for example, we have a lot of data, we build a model that is able to decide which images are cat images and which ones are not. The way i few view it, it is problem. Serious Artificial Intelligence, is google Artificial Intelligence. Way, think so. Whether they are thinking machines, i would sa no no. In the problem, they fall into that realm. Differencehe between a steking machine and a learning machine . There yet . No, no. Def lnt no i. We dont have intuition built am we do achines good job of confirming the obvious the machines. I show you a picture. Probably do say 596 getting it right, but you machines do an excellent job as well of identifying the pictures. Not really. Sometimest there surprising us. You drive a car very well. You are able to drive carves now. Yeah. I think they are not thinking machines. There is no, they dont really of intuition behind what they are doing. Dr. Kennedy, what kind of research or what, what is the back trend of that machine the face of the at . Technologies . So generally a big network. Met work, no networks dont have a picture here, but the Large Network you input the image by the pixel values. Nothing you have the dot on the screen and they show up with different tins ty into acan turn it digital signal and you feed that digital signal into the and whole bunch of nodes or in the middle of this that use this to compute different, compute and pushes ites to the network. I am not sure if that is clear. The oneslets go to before we look at what is on in board, why is serving ones ander . Res in your world this. We digitize everything. What do they stand snore theydo they mean well, can mean anything, right . They can represent your heart rate, your heart rate as you are going for run or be the we send tos that the screen for an image and produce the pixel image on the screen. Computer Stores Information is in one computer that is normally the way we deal with information. All right. Well, we are in what is called the anomaly room here at bell labs. You going to show us . Well, so what we are talking about. Well, what we have here is what we call augmented intelligence so rather than replace the human thing what we are trying to do is build tools that help understand. So we are not doing this in a want to build a tool that is simple enough for children to look at and indeed of in guesting dat from pokemon that kids to play with and we took it to a robot picks conference me kids were explaining to how it works but we have internal tickets, et cetera. Ingest want to do is it of all types and all paces so sensors or image sensors or the weba, scraping for document, et cetera, even you know, the websites that videos where shall et cetera, we create that information. Build internase allows people to plosion, the information so they can see the information in a way they seen it before and allow them to compare things inside of this data. Why things are showing up on the screen. What we mean by creating and decisions about this information. Areyou can imagine you given some huge spread shet or document you want to ingest and what is useful for you is to take that data and the ones the document and turn it into information but imagine that you can use to answer evolve andnd to some information and so what what we callis august amed intelligence and what we think of as assisted so tools that allow us to think better and better. Uestions well, lets get to pokemon and walk us through what intelligence can do. We got to have a special touch here. Oh, yeah. Yeah. Have to be a mathematician to be able to do that. Ok. Will stand back. You go ahead. Through. Us yeah. This is just a massive information from pokemon like you have thend different creatures or whatever call them of pock can guess. You have a whole bunch of extra features. Screen shows a global ranking of what the smart machines that under pin this thought were the most important thing so something on here that you find interesting and another feature we feel if you are going to play with it. You use dat. You want to interact with dat so perhaps you find guy interesting. Him. Imply touch now by touching him. What you have done. You have under kateed to the the piece ofis machine you are interested in. It is not showing here. We have the ability to tell you are not interested in certain pieces of information. That what you see is the screen was reorganized around the pieces ininformation that were interesting so you can see Different Properties of these pokemons whether the way they breath, i guess, the name, the general category overall. We did this for kids to think about information or to play with information or contact when you are thinking about a problem. You want to take that information. You want to quickly stream line it to the piece of are interested in. Guess you could think about doing interview and you like build the great documents and go through and ex ac the piece was information. Cut your that would eye and you collect that and hope you are able to reorganize the screen to help you filter on these pieces of information. So of course, you thinking about things that are, when au are thinking about problem. We dont want to think about the things that are immediately related to the subject. Want to look further out. You know, that is wha what i am able to do with the machine. Am able to zoom in and zoom out on the data to look for things that are related. Well go. I can see if i can help this here. Do you have questions . Keep going. Keep going . Well see if this one works. I never used to play with this screen. Dont have this in my office. I dont think its the within i want. You different one. So one of the curious artificialt intelligence which get comes from the crutch why the really, you not know, they are not intelligent in a sense, if you are were to network toe recognize, did you a good job in, if you images put in a cat image, you know, the time. Right . You are set. You have this the question is what is going to do . Did you learn anything about that . What about similar things, like what is going to happen by sticking image of a fire truck . Is it going to solve my problems . Am i . You know, have i learned something about the structure that will help me recognize the fire truck . You are curious. It is could kind of return right . Thing, it doesnt really contain any intuition that is helping you recognizingo different types of images. You got a really good job of training on for example. Then here, in this space is we what we should be able to do is trin a big network and use network to understand what thises are similar and what is happening in the end side of the network and something here to catches your attention. Well give you a second chance. Well see what happens here. Hope so. I would i answer, possibly you could ex track. These are images now. See. This is not based on key words. The key words that tag the important on the features that underopinion the images. See why the images are similar on the screen it is not because of the service. Whenan tell as a human you look the images the middle, right they seem very right . To you, perhaps, you could describe it, right . Getting the machine to it is more difficult. Intelligence is being used now, correct . Well, now. Right now. Well, he mean, in a practical sense. I think so. I think so. We build tools as humans naturally, right this the idea of augmenting the limits is being used. You know . Example. Google could be thought of as memory. D you dont have to remember anything. You can go to google and wilt it up for you. The tools try to augment. We believe this is a different on the actual way were using this tool. Intelligence. Are we using it today . Well, like not in this room but on pratt call level . The worldple in using ready Artificial Intelligence . Yeah. Sure, google, facebook, selfdriving cars, these things exist. That is a good example. These are real examples of to does that are able things that humans cannot any more, right . How far along are we in our and our knowledge . Done ai think we have very good job. So we are where we are right now in the Artificial Intelligence space because of massive amounts of dat that are available to us and the massive amounts of power that exists now for eventually free and transportility to and our ubiquitous activity to anywhere fromta any time. These things together have made, i mean, like the idea is help, you oh, reinforcements learning areniques and techniques 3040yearold techniques, righting . Lot more, we have a power and dat to make these techniques actually work well, as far needing to achieve untuition, it is hard to say. I think we will continue along this path. Whatll be surprise by machines can do year over year whether or not this builds intuition. A good question. Sean sean kennedy working Virtual Reality . I am not. You are not . No. I have watched a lot of demos but i dont think should be giving you one. Artificial intelligence, augmented intelligence are the two things are working on here bell labs . My main focus. Yeah. Thank you for your time today. Are very welcome. Introduce you to this man. What is yore title here. When you say that, what is that . Integrated circuit f. These are highly statistics for very specific applications and traditional and over to the tasks and optical designedtion and only. Anything that you worked on now . In the public right are people using your products into well, actually. Certainly stuff that bell labs industry andhe we have released products that communication ideas. Yes. Traffic of the cell phones will go through that at some point and time. So what are you working on you . Well, a lot of things. Forefront, the most exciting is the communications. Is . The 5g is an interesting thing because it is 100 years since we had it. Changed. This is what we do with you are wireless communication. But what we want to do is go ofa new rear communication. That era is directed communication. Toed casting, we want to target the beams and those individuals. We wanted to do this because i searched the data. It is never ending. We have saturated our spectrum. Andave higher frequency the challenges is the law through the years is too much. Cannot do it. If i with president to talk to at youave to direct and the data and get the dat from you and move to the next person. A complete changed communication. Has a huge set of challenges. And the next couple of years . Yeah. Yeah. It is am berb cious goal, obviously. Different vendors want to be out there. And that is both exciting and going to be difficult for think it is a great bell labs program because there are so many things you need to solve mostly in terms of cost, performance, integration, and manage. Means you need an theronment to target different problems. That is why bell labs is so so manyause you have experts in so many areas that that. Gether to resolve now my guess is that bell labs is not the only Research Work . Ty no, not at all. Is it a competition . Well, i mean, everything is competition otherwise it is not worth doing. Everybody is working on it. It is exciting. Certainly, a lot of people, a lot of smart people in the working on. I eventually, everyones in knowvations are going to come create the next, the next revolution in communications. Were all excited about it. It is good. Shahriar, how it did intoe our dally lives what we want to do is give you more data. You can change the way you use freelyrt devices. And would not have to think about it, that is what you need. This does not even include all the other highresolution four k tv series you want to watch, whether you have your next cell phone transmit tens of gigabytes of data all the time. That is the only way to do it. If you want to get to a point where we can do this freely, we have t