We talked about their Communications Project and looked at some information on Artificial Intelligence. They are headquartered in new jersey. The communicators is on location at Nokia Bell Labs in new jersey. We are talking to the Staff Members about what theyre working on. Joining us now, michael eggleston. What you do here . I am a researcher. I work on. [inaudible] and technology. What is your back ground. Semiconductor devices. My degree is from berkeley. My phd was at berkeley and my undergrad was that i was state university. Are you from iowa . Im from minnesota, originally. Your here at bell labs working on. Optical senses. There anything that detect light. What you are mostly with is a camera. The only see twodimensional images so im working on a 3d imaging technique. Its very similar to radar by using light. Are those in practical use already. Yes, there are, one of them is in the medical field it is noninvasive. That is pretty common, today. Yes, these are becoming more common. They take up a large table and can cost tens or hundreds of thousands of dollars. Im working on taking an entire system and shrinking it down to the size of a chip. We monitor ourselves and the environment around us. How far away are we from that type of technology. There are a lot of challenges that we have to overcome but right now we have all of the building blocks. We kind of have developed with the optimal indication technology. In your work is a completely healthrelated . Is that where your relat headed . Thats where i am headed, healthrelated, sensing the biochemistry of the human being and monitoring our health on a daytoday basis. This was one of the Biggest Challenges we face for humanity which is how we feel and its really something that we dont monitor. Theres a lot of opportunities and challenges that come. We want to make continuous help monitoring a reality. Other sensors and then them math that goes into the optic. Yes, theres a whole section of work that needs to be done both making the components that can be small and integrated, putting it in the system and developing the algorithms to use these new tools, to actually diagnose things so, im not a doctor, i dont have a medical background so what i do they really make tools, new tools that can help those professions do their job. So, you talk about the 3d i image, where else can or is this being used today . Its really seeing a lot of use, the second most common is in skincare so for example if i have an abnormal growth on my hand i can actually do a full 3d scan and tell of its cancerous or benign or something i need to look at. Its completely noninvasive. You have to cut anything off or take a blood sample. There are many other fields people are using, dentistry, you can actually look at the enamel of the teeth and see micro factors before they develop into cavitie cavities. Even inside the body there using this to do internal surgery so they can see what theyre doing on the inside of the body. So really the range of possibilities is in less. Anything youd want to see you could use this technique. A lot of us are familiar with mri. How old is that technology and how is this related to what youre doing. Mri is a pretty old technique, it was pioneered. [inaudible] its a great technique. It lets you scan the entire body, but the resolution is millimeters to centimeters and the entire system is massive and something thats very challenging to scale down. They are very expensive and people dont have access to them. Now with these optical techniques, they can be very small and very cheap and can actually allow us to do a lot of the same thing that were doing with mris and do them at a much lower cost. Whats more important in your work, the scale or the technology . To me i think they kind of go handinhand, but at the end of the day when you can scale something down to a small size, you can make it cheap and even enable it for a mass audience. If we all had access to mri in our daily life it would significantly enhance our health. So to me helping everyone is one of my main goals. Thats what is driving the size down with the technology we have. On your door is assigned that i want to ask you about. Danger, invisible laser radiation happening. A lot of the sensing that we do is in the infrared so its longer wavelengths of light than what you can see and we tend to use rather highpowered systems when we are demoing them in the lab. This is just a warning that depending on what experiments we are running, you might have to wear safety goggles. Now as you can see my doors open, thats how it usually works, usually we dont use these lasers because were really looking at something that is used to scan the eye. Where is a laser . You have an inherent . Were actually surrounded by lasers. I can bring over here and show you some examples. In these date back to einstein, dont they. So yes, lasers date back to the late 50s, early 60s. Based on einsteins work. Yes, stimulated emission. Usually, the factor this. So that the laser. This is a laser. Its a pretty nice laser that is used for optical and the actual laser is much much smaller. The lasers can actually be actually, even for example if you have the air pods, those have three lasers in them. You can actually make those very, very small which is part of the reason theyre so attractive because now we can actually embed them in everyday devices and have an incredible amount of added functionality. This is a laser, we actually have several lasers on this table here and there in all different shapes, sizes and its all about packaging. Lasers are tiny but we put them in large boxes because they are easier to handle. Before we enter in that your daytoday, what were you working on . I am working on a new method to actually implement these systems. A lot of it is actually about when you come to System Integration is writing codes and algorithms actually do what you want to do. Whenever things tiny can actually physically change it so everything has to be automated. If i was actually writing code to do this. Where did your interest in this topic come from . So ive always been fascinated with light. To me my vision is my number one thing. So i went to school i learned more about light and light is amazing because everything it touches, it actually picks up the signature of that material. For every beam of light that bounces up and hits you and me actually carries up information. If i look at it the right way i can gain an incredible amount of information. That kind of curiosity has led me too this path. Michael with gnocchi labs. Thank you. The communicators continues its visit to gnocchi a bell labs in new jersey. Joining us now is a gentleman named sean kennedy. Mr. Kennedy, what is your position here. Im a Department Head and i lead the network team and we are interested in a lot of things but Artificial Intelligence, augmented intelligence, and a whole host of other Network Algorithms that were trying to develop. Before we get into those, what do you mean imap of networks . Well, networks are just a structure that you can look at so we study the structure and the properties and then generally we build algorithms to solve problems so whether its voting problems like how you get information from one part of the network to another part of the network or facebook which is just a Large Network so lots of network type problems come from studying the structures of these networks and you can decide things like who you should be friends with next and other questions that we experience on a daytoday basis. You mentioned Artificial Intelligence. How do you define that question. Thats a good one. Im not exactly sure how i define it. In general i think people think about replacing human intelligence and thinking machines, self driving cars and things like that, i tend to think of it a little bit more like a mathematician so Artificial Intelligence is just a giant optimization problem where we put in a whole bunch of input that is labeled and we build machines for mathematical models that do a good job so we might have a lot of data that is cat images and data thats not cat images and we build a model that can decide which are which are. The way i view it is just a large optimization problem. Is siri artificial intelligenand google . Yes, i think so, whether or not there thinking machines i would say no, but in the states of an optimization problem they would deftly fall into that round. Whats the difference between a thinking machines and a learning machine . Are we there yet. Definitely not. We dont have intuition built into these machines. We do a really good job of conserving the obvious spread by show you a cap picture 95 or 96 would probably get it right, but the machines do an excellent job as well as identifying cap pictures but they are not really, sometimes their surprising us, but you probably drive a car really well so i think theyre not yet thinking machines, theres no, they dont really show a lot of intuition behind what theyre doing. Doctor kennedy, what kind of research or what is the back and of that machine recognizing the face of a cat. What type of technology is that. Generally a big network, its a narrow network, i dont have a picture but the Large Networks were you input the image just by the pixel value, and images nothing until you have these spots on the screen and they show up with different intensity so you can turn it into some digital thing and until you see that digital feed into this network, theres a whole bunch of nodes, in the middle of this with edges that use this to compute different values as it pushes itself through the network. Lets go to the ones in the zeros before we look at whats on the board. Why is everything in ones and zeros in your world . Well because we digitize them. What do they stand for . What they mean. They can mean anything. They could represent your heart rate in your watch as youre going for a run or the pixel images on screen so we can produce the pixel, its a way that a Computer Stores information in ones and zeros so thats normally the way that we deal with information. We are what is in the anomaly room. What are you going to show us . What we are talking about or what we have here is what we call augmentative intelligence. Rather than trying to replace the human thinker, what we are trying to do is build tools that help humans understand information were not doing this in a way that is just for data scientist. We want to make it simple for children, from taking pokemon that children love to play with and explaining to me how it works but also we have stuff for internal tickets, et cetera. We want to ingest data of all types across all spaces. What we want to do is build an interface that allows people to explore this information so they can see this information in a way that theyve never seen it before, allow them to compare things inside this data set so we start to understand why things are showing up on the screen and then ultimately we want to use this information to create knowledge what we mean is we want to be able to make decisions about this information. As you can imagine youre giving some huge spreadsheet or website that you want to ingest and ultimately whats useful for you is to take that data, the ones and zeros in the dots and turn it into information but ultimately knowledge that you can used to answer questions. Maybe your boss is asking you for some information and what we have here is what we call augmented intelligence and what we speak of as assisted thinking. Their tools that allow us to speak better to answer questions better. Lets go to with the pokemon and walk us through what this can do. Youve got to be a mathematician to be able to do that. Perhaps. Im in a stanback, you walk us through the. So this is just a nest of information from pokemon. We ingested different creatures and then you have a whole bunch of extra features that theyre focused on so the screen shows a global ranking of what the smart machines thought were the most important thing so if theres something on here that you find interesting, another feature we feel if youre going to play with data and use data, you want to interact with data so perhaps you find this guy interesting so you simply touch and, now by pet touching him what youve done is youve indicated this is the piece of information youre interested in. Its not showing here but we also have the ability to tell a machine youre not interested in certain pieces of information so once i did that, what you see is immediately the screen was reorganized around the pieces of information are interesting so you can see these are Different Properties of these pokemon, whether its the way they breathe or their name or general category overall. So this is a bit of a toy example, and we did this mostly for kids to think about information, but you know just to play with information or just to get contacts, but when youre thinking about information, what you want to do is take that information and you want to quickly streamline it so the pieces of information you are interested in. I guess you could think about doing an interview and youd quickly like to go through these documents and extract information, something that caught your eye in a document and you select that and you hope that youre able to reorganize the screen to help you filter these pieces of information. Of course when youre thinking about things, when youre thinking about a problem sometimes we dont want to think about the things that are just immediately related to the subject we will look further out and thats what im able to do with this machine. Im able to zoom in and zoom out on the data to look for things that are related. So well see if i can pull up a slightly different example. You have questions . Keep going. Lets see if this one works. I never get to play with us. I honestly dont have this in my office. Lets go with this one. So one of the curious things about Artificial Intelligence, which i guess maybe comes to why these machines are not intelligent in the sense, if you were to train a Neural Network, if you did a good job of sticking images in and you put in a cat image, one 100 of the time it says cap or not cat so your set. See you have this, the question is whats it going to do . Did you learn anything question did you learn anything about that cat or what about similar things like whats going happen if i stick in an image of a fire truck . Is going to solve my problem . Have i learned something about the structure of cats that will help me recognize fire trucks. Its curious that it could return either thing. It doesnt really contain any intuition thats helping you get closer to recognizing different types of images. Youve done a really good job training on cats, for example. So here in this space, what we should be able to do is train a big Neural Network and use this network to help us understand what things are. Lets see what happens. I would hope so. You could possibly, these are all images. Its not based on keywords but more importantly this is on the deep learning features that underpin each of these images. You can see why theyre similar on the screen. Its not because dashiell can tell is a human when you look at all of these images in the middle, they seem very similar. Perhaps you could describe it. Augmented intelligence is being used now. Yes, in a practical sense. It think so. As we build tools very naturally so the idea of augmenting, for example, google can be thought of as augmented membe memory. You dont have to remember anything, you can just go to google and will likely put up for you. So the tools that try to augment, we believe this is a different spin on the actual way we are using this tool. Artificial intelligence. Are we using it today . Like not in this room but on a practical level. Our people in the world using Artificial Intelligence. For sure. Google and facebook and self driving cars, these things exist. These are real examples of machines that are able to do things that humans cant. How far along are we in our research and knowledge of ai . I think we have done it a very good job, so we are where we are right now in the Artificial Intelligence space because of the massive amount of data that are available to us in the massive amounts of Computing Power that exist now for potentially free and for our ability to transport our ubiquitous, the fact that we can collect this data from almost anywhere. The idea behind this reinforcement learning technique in Neural Network is 30, 40yearold. [inaudible] we have more computational power and data to make these techniques actually work well so, as for leading to smart machines or intuition, its hard to say. I think will continue along this path and be surprised at what machines can do yearoveryear, but whether or not this builds intuition, thats it good question. Shawn kennedy, you are also working on virtualreality. I am not working on virtualreality. No. And ive watched a lot of demos but i dont think i should be giving you one. Artificial intelligence, augmented intelligence are the two things youre working on here. Its my main focus, yes. Thank you for your time. You are welcome. And now on the community is on our tour of bell labs in new jersey, we want to introduce you too. [inaudible] what is your title here at bell labs. I am in the Research Department where we work on the next generation of telecommunication. These are integrated circuits but they are highly sophisticated for specific applications. These could be for wireless or optical communication designed only for that specific purpose. Anything that you have worked on is it in the public . Are people using linear products . There are some of the things, i havent been around for that long, only eight years but certainly things that the lab have both have been in the industry we recently released some products. So some of your traffic of your cell phone and internet will go through. So what are you working on now . A lot of things and at the forefront, probably the most exciting is the 5g. Which is. The 5g is an interesting thing because its been a hundred years. [inaudible] this has changed, this is what we do, its what all wire wireless medication is. We wanted to do is create a new era of medication and that is direct communication as a pose to broadcasting everywhere, we want to target it at individuals. We been wanting to do this because our search for data is never endin