35 Burst results for "AI systems"
Facial Recognition Auditing
"Hi I'm. Deborah G and I'm attack fellow at the institute at New York. University will thanks so much for joining us today to kick off. Can you tell me a little bit about your background? I started off studying robotics engineering at the University of Toronto, in Canada, and then I spent a year working on the machine, learning team at clarify, which is a computer vision company in new. York and then wallet clarify sort of noticed that there was in the computer vision community especially with facial recognition in that space in particular there were glaring racial disparities in terms of our data set so data sets being used the facial recognition space in particular, was very visible that there were huge demographic disparities, huge underrepresentation of people of color for example certain demographics cues, and this was in my intuitive At the time. It's since bins empirically demonstrated, but at the time. So I started digging into it, and like exploring it more, and that led me to work with joy blend Weenie at the MIT media lab, and she was working on a project called gender shade, so the gender sheets project was really an investigation into the performance of mainstream deployed machine, learning systems by IBM face plus class and Microsoft and she looked specifically facial recognition systems for the task of gender classification. Is she said what would happen if we evaluated models, these deployed models already out there in the wild already being sold already being used by developers, you know. Know, what would happen if we evaluated these systems on not the demographically benchmarks that we all use in the computer vision community, but what if we created a new benchmark that was not demographically skewed, that was thou- for Gender Representation also skin-tight, so that's what she did. She created this benchmark and evaluated these mainstream computer vision fish recognition API's on demographically balanced benchmark, and what she found out was that there was a huge disparity between the performance on darker skin, females and lighter skinned males, and it was important revelation, especially the facial recognition community to realize that. That a lot of the data that they were using were not demographically representative Ed. There's a lot of racial bias, but also general demographic bias in the models that they were building and deploying so I worked with her that summer, and we did a lot of follow up work to gender shades, analyzing and beginning to try to understand companies didn't response to gender shades how they diversified their data sets in order to do better on the benchmark that we created how certain companies responded or did not respond in response to being targeted for specific audit, and then we. We also kind of looked at particular elements of audit designs that led to impact that led to the company's feeling. Push to change their behavior and at the same time, also I guess about low a research, not at Google I was working with colleagues there to think about documentation. How do we communicate the performance of a machine learning system? And how can we incorporate some of these ideas around auditing into the way that we present and talk about document, the performance of shoe system, and that sort of launched me on this whole dirty, which is where I. I am now like thinking about evaluation of machine learning systems especially under the language of Auditing Assessment Dinky demographic bias, but assessment other ways in other elements of the system, and then also thinking about the communication of the performance of the system. How do we document any of these things in a way that gives us a sense of how the model performs when in the real world and that's really connected to what this paper is about as well. This paper was written with some of the colleagues. It Google. I've been working with on that documentation project also other. Other colleagues from the Computer Vision Phase asking what do we learn and what can we not learn from what we call the gender shade style audit what we learned from these audits on demographic bias, and what is still missing information that we still need to figure out a way to capture document in order to really communicate it understand the full performance of a model or system or AI. System wants US deployed so yeah, that's sort of a brief overview. The whole journey of how we got here and this paper in particular is in response to the fact. Fact that following the generates project and following the subsequent sort of follow up work to under shades. We were realizing that a lot of people were just taking the benchmark from gender shades or recruiting shadow version of that benchmark and using that as a moratorium condition in policy for example or trying to use a similar method to assess the suitability of a model before deploying facial recognition model for demographic disparities and what we found. This paper goes into detail to ethics. There's a right way to do that and there's a wrong way to do that and they're. They're sort of important more nuanced ethical questions involved that need to be consider that need to be talked about when assessing official recognition system for example, but any broad system, and we need to sort of ask ourselves these more careful nuanced questions, the aware of some of these more nuanced ethical tensions before we allow the systems to be deploy,
Detecting Emotion Through Gait with Aniket Bera
"Hello and welcome to the AI. Today podcast I'm your host Kathleen Mall I'm your host modeled Schmeltzer our guest. Today is unaccounted. Berra is the research professor in computer science at the University of Maryland at the Gamma Lab. Thank you so much for joining us today. Thank you so much. Thank you for inviting. Yeah thanks so much for joining us today. We'd like to start by having you introduce yourself to our listeners. Tell them a little bit about your background. And your current role at the University of Maryland. Sure I've been with him for a little under a year so for this was at. Unc Chapel Hill for a year and had mcbean St at UNC in robotics so a hermit research over the last few years has been trying to work on the social aspect of our bodies whether from the computer vision side where we look different objects after we as humans when we look at different objects. How do we perceive them? So my research has always been about. Data perception for robots and robot can understand the world around us like as humans. It's what research has been over the last six seven years might Rowlatt you. Md has enough research faculty. I advise about seven students now from vision. Applications to robotic applications the psychology driven. Ai Applications on McLennan field of research is something affective computing affective computing. What it means is that we're trying to gauge emotions on Detroit's you in aggressive Shy said figured out different cues from your visual appearance like your facial expressions the way you speak things the way you walk from all of that. Can I figure out your emotions and then do something? According you follow. What specific here. That was something that we found really interesting. You know part of the reason why we reached out to you and hang you join us on their today podcasts. As we wrote an article for our today podcast listeners. He may or may not know that Kathleen and I are also contributing writers to Forbes and tech target and one of the articles that wrote in Forbes was how systems might be able to detect your emotion. Just by taking a look at how you walk and other sort of non maybe visual facial visual or verbal cues and that's part of what being socially intelligent is. I guess we as humans can read things like body language. But there's a lot more to it. So maybe you can explain some of the concepts of socially intelligent robots and why this idea of social intelligence is important yes oh the concept of socially intelligent robots is essentially making robots understanding humans. Better so we. As humans are not objective to be tend to evaluate unheard of things based on out upbringing aquaculture in all these different rates. And then associate all those things in everyday life so In this research which you mentioned the phobic mad we did when research way we could figure out how people walk and then have a garden new. Somebody's side we could mechanize that person sadness just by looking at his a her. Buster is a hub body language and maybe the robot can walk up to that person and ask questions. You look sad today. Can I help you like some help if somebody is Chris excessively angry? I might want to talk to that person and maybe even avoid that person together. Assembly looks confused. The robot co that Bush announce something that you look lost here. Do you need help been something. Do you need to some place? So all these different we inherently as humans. Do which usually doing tend to do those things so go. The last Jedi as robotics has always been about solving problems accurately and objectively so. Let's say you know the goal for about is go from point eight in be and the robot will try to figure out the shortest are the most efficient go from point eight point what via bringing in is also being associated vendor most socially aware the. Somebody's walking. I want that other person to have his hub Robot I do not want to enjoy on. Somebody's face so having all the social norms social events bring it back to robotics is what the concept of socially intelligent robot as unwise. This idea important I think has become more primetime and as they become more available among says I think they should try to attempt to understand humans and be Understand humans but go beyond that and be part of the human society He not interesting because we talk about commonsense emotional Iq and that's incredibly hard for robots and artificial intelligence to actually have been a lot harder than I think. Maybe some people realize although there has been some discussion around it and at COG Melika for the past two years we've actually done voice assistant benchmark and commonsense and emotional. Iq were too of the categories of questions that we asked because supposingly the systems were not very good at that but this idea of AI systems that can detect emotion based on gate is a really unique idea. So where did this concept come about? So we started this. Actually I think about skate years ago. I mean I know. Eight years ago IAE wasn't the we'd know now things were different back then but we started with the concept of can't be figured out somebody's personality just the like just looking at how they walk back. Then we started the representing every human being every industry as a single entity as a single dot on the screen so used to look at videos. And how this guy is trying to avoid somebody to cut across people to figure. Oh his guys aggressive. This guy's got shy guy walks around all these other guys look from that and now we figured out so from the dot aspect of figuring out the entire body leg rewrite now have around twenty two points Bush and so all our lake from your leg from your hand gestures your shoulder through your slaughtering head so all these things. All these different cues which we observe. That wasn't really being studied before there's a lot of research on the emotional especially from faces. You know somebody's happy. Somebody said there's a lot of research in this field especially from speech in the way I see something. Let's say I'm happy? I'm very happy today. I'm okay I'm happy but also the way you say. The sentence is the content of the sentence is one thing but also to the way you was at so all these different cues were being studied in different fees realized that the body language is something which really people studying. We look at people a lot. We look when they walk in the talking when they're driving but we don't we we know what they're going to but we don't really like we haven't understood how gives how walking body language relates to emotion so our on this hatred emotion is kind of it could be added all these facial cues with speed with other were. Hughes from the human so I researched
AI for Social Good: Why "Good" isn't Enough with Ben Green
"All everyone I am here with Ben. Green Ben is a PhD candidate in applied math at Harvard and affiliate at the Berkman Klein Center for Internet and society also at Harvard and a research fellow at the AI now institute at Nyu then joining me for our continued conversations coming out of the thirty third nerves conference here in Vancouver Ben. Welcome to the TUOMO AI. Podcast thanks so much for having me. I'm excited for conversation I am as well so you know as I just said your degrees going to be very soon in applied math But you have applied that application of math to Largely an exploration of the intersection of technology and social good Tell us a little bit more about your background. Yeah so my background is primarily computer science and data science. But I always even going back to my undergraduate years. Had a really strong interest in urban policy and urban government urban planning and so as I started out the PhD. A lot of my emphasis was on. How can we use the tools of artificial intelligence and data science to improve society and participated in the Chicago Data Science for Social Good Program? And did some other work really thinking about how can is a data. Scientists contribute to society. I spent a year working for the city of Boston as a data scientist in the middle of my PhD. But in the course of doing that were from originally from a more technical perspective increasingly also came to see the broader governance political social questions that were at the heart of this of these technological endeavors and often were overlooked or ignored and also played a significant role in shaping the impacts of these systems in some of the projects that. I've worked on whether it was building machine learning algorithms in the city of Boston for the city of Memphis to help them prioritize various types of inspections and investments often. What I found was that the key factor that shape the impacts was not the technology itself but the broader policy government political environment in which that technology was being embedded and so that's shaped a lot of my thinking on how to integrate technology into these broader social context and how to think about the ways in which why are very well intentioned efforts use technology for good can overlook some key factors and end up failing to achieve those those social goals and your affiliation with the the applied. Math department is that I imagine there are several places that you could have plugged in your research interests at a place like Harvard. Is that a the selection of a particular adviser or is applied math. How does applied math fit into That brought a research agenda. Yeah so definitely the really the computer science perspective which is sort of wear my more more day to day home is at Harvard is both very much. A lot of my work is sort of is all sort of within the realm of computer. Science both Thinking about those tools in the past more work on building those building. Ai Systems for various social applications. Running different types of more recently. I've been running variety of human computer interaction on its to understand how people interact with algorithms in practice and but also my work is very much about stepping outside of the typical modes of thinking within the field bringing other perspectives from science and Technology Studies and philosophy and government and thinking about what law and what those perspectives can do to inform our understanding of artificial intelligence and its impacts and how to develop it so. My work is definitely very multidisciplinary. And I've worked with you know even within Harvard at many different departments and with many different people but the core focus has always been on the application of data and Algorithms in society. So here it narrows you're presenting a paper called good enough at the AI for social good workshop. Tell us a little bit about the paper and what your objectives are there. Yes Oh this paper is it's just A. It's a short workshop paper so definitely not fully in depth discussion of these topics but it really emerged out of my own experiences and some of the other broader examples. I was seeing of these efforts to do social good that were well-intentioned but often sort of not thinking of the full of the full picture of what it actually means to do. Good and recognizing in particular that efforts to do any sort of technology for social good are about Somehow Shaping Society Somehow Changing Society for the better and that's of course an incredibly complex topic and the two things that I really points out that I have seen missing in the majority of efforts to use for social. Good is I what I would think of as a normative theory is grounded definition of what good actually means typically most groups will talk about Ai for social good and the social good part is sort of taken for granted what that might mean but of course as you can you step out of the I. Space and just think about our broader social political world there are many different definitions of what's good and many nuances within that type of debate so there is often a lack of sort of a normative discussion. About what are we trying to accomplish? And the second part was a lack of what I would call a theory of change a theory or sort of a grounded justification for how the particular technological approach being taken is an effective means to getting to the social good and Whatever that end may be and so a lot of the time even in cases where perhaps the unimportant problem is recognized the particular mode in which technologist go about trying to solve. That problem may not be the most effective way of achieving that end. If you take the social good the social impact as the as the ultimate goal here and think about the technology as a means to achieving that goal and so both of those things are I think pretty significant challenges certainly not ones that cannot be overcome but the types of things that really need to be incorporated into these into these discussions. I like to think about it and in some sense. This is really what the goal of the paper is to do. Is I like to think about it in terms of rigor right that when we talk about Ai for social good. We're actually doing is really expanding what we're trying to accomplish with an AI system right we're not simply saying we want to build a tool that can efficiently predict this or efficiently analyze this data. But we're trying to build a tool so that it can achieve or advance this social outcome and what. I'm trying to bring a sense of here. Are things that we're overlooking failing in many cases to think about and trying to frame that as a lack of rigor in these efforts that were actually non thinking about factors. That are incredibly important in shaping those outcomes and that to the same extent that we would never accept a system that hadn't done an analysis on some sort of holdout tests data set. We also shouldn't be accepting systems for integration into societal context. That also hasn't done some sort of analysis of will what will the impacts of this system in practice? How is it actually going to affect the system that were trying to impact here and bringing in more of those types of socio technical analyses into what it means to build and evaluate these types of
Kevin Scott and Reprogramming the American Dream
"Let's let's jump into the book as you said in your introduction. The title is reprogramming. The American dream the subtitle of the book is from rural America to Silicon Valley. Making a I serve us all and the book is published by Harper Business. It's available now for purchase wherever you buy books one of the first things that we talked about. Really in one of our first discussions was that You know storytelling as you say is a southern thing. Tell me the story of the origin of this book in the story that you wanted to write about. Well I've been part of the development of technology for a really long time and technology itself was the way that I was able to build a life for myself like I think in a whole bunch of ways like it really saved me as a as a kid. It was the thing that I latched onto that help. Give me something productive to do with all of the energy that I had like. I just got really lucky. That personal computing was emerging. Right at the time. That you know is Preteen I was just trying to figure out what to do with myself and I got hooked in it has served as a platform for me for building a career and for doing a whole bunch of things that I think have been helpful to other people in at least some small way is and when I look at the state of technology right now like we have never had a more powerful platform in terms of Technology. So whether it's You know like the set of things that we're using right now to record this podcast. Because we can't be physically proximate with one another. Yeah the amazing networking technology that we have right now. Like computers that let us stay in touch with one another and interact and collaborate. You do really interesting interesting. Things would otherwise be possible but like I have particularly been involved with the development of for the past fifteen years or so so one of the first big projects that I worked on when I left academia to go work in industry was the machine learning thing and I've just sort of watch this technology progress in both power and accessibility and like what I mean in particular by that is power is like what you're able to accomplish with the tools machine learning and accessibility is like who is able to use these tools to create the. It's just been on this incredible curve on both dimensions and I sat down to with you. What is it now? Two and a half years ago when we started this whole process Seems like only yesterday? Yeah it does but you know the thing that I wanted to make sure was that people understood stories of how they could choose to use this technology to help build a better world for themselves and for their communities in this very inclusive way and like I wanted to make sure that there are things about the development of machine learning and its uses that we need to be cautious about but like I want people also feel sort of hopeful about what it is that they can do with these tools. Yeah well the PODCAST is behind the tech and one of the things that I really enjoyed is learning your story. You know from Gladys Virginia to places like Illinois in Europe and in Silicon Valley. So it's a really terrific Story I want to jump into the introduction here and topic on a I in the introduction. You address the question that is on. Everyone's mind which is wind will general artificial general intelligence or Agi. When will it be available win? Will computers the smart is as humans. So I'd like to ask you to read a from the introduction which you address that question okay. Yeah it'd be happy to so even though I have neither the expertise nor the crystal ball to predict exactly when Agi might arrive. I've been involved with modern technology long enough and read enough history to know that we've often underestimated the speed with which futuristic technology suddenly arise. Avon has historically been limited in what it has been able to accomplish by the amount of compute power. We can throw a day our problems and how much time it takes for humans to encode logic and knowledge into Ai Algorithms. We now have enormous amounts of compute power in the cloud and we have enormous databases of digitized human knowledge like Youtube and the kindle bookstore that can be used to train. Ai Systems as their modern. Ai Algorithms absorbed that human intelligence to accomplish your task we imagine for ai powered systems we may achieve what Thomas Kuhn defined as a paradigm shift one which humans will either be in the loop or out which one of these options we reached depends on our actions. Today the story re craft and the principles we assert about what kind of world we want children to live tomorrow. I feel this profound sense of cognitive dissonance. The same thing that can advance humanity can also cause people distress and even harm. This book arises out of a powerful urge outfield to reconcile the two. It is an engineer's tale. Not The musings of philosopher economist or screenwriter as Microsoft chief technology. Do I skin in the game. Of course I do. But also the product of rural America one of the places most vulnerable to the dystopia in story of AI. My values and many of my earliest experiences as a budding engineer occurred in a part of America Rural America. That is most at risk. I left the rural south over two decades ago I for academia and then for Silicon Valley and the tech industry but it by core I am those people rural people and I care about creating future that values them and their resourcefulness. That's great thank you.
A Summary of Automation and Job Related Impact and Solutions
"So regarding the future of technology and the ability to automate human tasks and even entire jobs there are really two dominant positions that people have held over time so one is the historical perspective which really falls in line with the idea of creative destruction and the perspective that rose from the late movement in the eighteen hundreds that of technological unemployment so in a two thousand fourteen questionnaire by the Pew Research Center there were well over two thousand known experts in the field of technology that were questioned on which side of this argument. They fell on either the creative destruction side or the technological unemployment side so the chief scientists of salesforce dot com vice president of Google and the principal researcher for Microsoft were among those interviewed so forty eight percent of all the experts envisioned a future of technological unemployment in which robots and digital agents have displaced significant numbers of both blue and white collar workers with many expressing concern that this will lead to vast increases in income inequality masses of people who are effectively unemployable and breakdowns. In the social order on the other side though fifty two percent expected a future of continuous creative destruction where technology will not displace more jobs than it creates by twenty twenty five so this group anticipated that many jobs currently performed by humans will be substantially taken over by robots or digital agents by two thousand twenty five but they have faith that human ingenuity will create new jobs industries and ways to make a living just as it has been doing since the dawn of the industrial revolution. So this is just one example of a clear divide between the possible futures but I think represents discussions and debates that are ongoing by many people and shows why think a podcast that explores these ideas is so important today but what about the technologies that will be doing the future disruptions so I started this podcast by looking at artificial intelligence as it is seen as the technology with the most potential to impact the way that we work so we currently have many examples of narrow a I think of examples like Alpha. Go that is able to defeat any human chess or go player as well as many other online competitive games or a system built in San Francisco by it company called and Lick where it was a fifty percent better at classifying tumors. Compared against three expert human radiologists working together. So we can see that the industry as of at least two thousand eighteen was worth about one point two trillion. Us dollars an estimated to reach up to around four trillion by twenty twenty two and with it it has generated a significant amount of jobs focused on building and improving the technology as well as the business cases and applications for its use so current narrow. Ai also augments people's capabilities in what has been termed center teams. This is where humans work beside AI. Applications to improve the quality of work done so the radiologists example actually enables the human health care worker to focus on other tasks that can lead to more patient care and human to human interaction but the ultimate goal in some ai circles is artificial general intelligence which has the capacity to understand or learn any intellectual task that a human being can so specifically once a artificial general intelligence or agi comes about. It is entirely possible that human creativity ingenuity and even capacity for reason will be made completely inferior much like our chess skills of today so this could apply to music and even the arts in general where. Ai Systems are already composing music and designing drawings. So if taken to its logical conclusion we would be hard pressed to think of a place where humans would be relevant or even needed anywhere for employment purposes at all so robots also play a large role in this discussion especially as they are more physical and tangible than a digital ai algorithm there are now some two point five million industrial robots in use across the world with the numbers being assault actually increasing every year. So four hundred thousand were installed in two thousand eight alone and the numbers are expected to be much larger in two thousand nineteen When those numbers come out so repetitive manufacturing jobs are the hardest hit with the adoption of industrial robots. As these metal workers are able to work twenty four and don't need any breaks holidays or sick leave which leads to drastic increases inefficiencies but service robots on the other hand have gone over sixteen million units sold across the world and are composed of a robot. Vacuums autonomous guided vehicles drones EXO skeletons etc so overall service robots tend to augment human worker capabilities which leads to the transformation of jobs rather than their elimination for example elder care. Robots and agricultural inspection drones are good examples. Where the technology helps employees with specific tasks autonomous guided vehicles or. Hiv's however have been put to use within logistic centers as walls warehouses to create. What are called lights out or dark? Warehouses or humans are being completely removed to allow for streamline workflows in the dark and without heating so the number of expenses actually decreases as well and they're not a traditional robot three D. printing or additive manufacturing technologies. Also hold great promise to shorten supply chains by enabling something called on demand as well as on location production of really customizable products. So the interesting part of Three D. Printing I think is that many different materials can also be used allowing traditional manufacturing to be disrupted but also healthcare and construction as both organs and even houses can be essentially printed in shorter periods of time with much greater accuracy so moving on Autonomous Vehicles. Perhaps one of the most popular emerging technologies also has one of the largest autonation potentials of many technologies simply because one of the main jobs across the world which is held by millions has to do with the transportation of both people or goods so taxi and truck drivers delivery workers Boat and train captains and even pilots as well as all the supporting businesses around transportation vehicles are essentially under threat of becoming obsolete within the next decade so though passenger cars require the strictest testing do the of course close interaction with an urban population. We already see examples of fully automated trains boats trucks and planes already being implemented so many of these autonomous vehicles still have humans monitoring them in case of an emergency but in some cases only one human employees who is monitoring is required for a small fleet of vehicles which leads to question what will happen to current employees wants. This technology is implemented so both virtual reality and augmented reality are emerging technologies and are already impacting jobs by helping with the training and reskilling of employees so the VR training course has already been shown to be as comparable as an analogue or real life course when it comes to learning and retention of the information but what is perhaps even more interesting right now is that specifically VR is going to possibly open up an entirely new space where people can interact and communicate and exchange services leading to potentially new economies that are purely
Where is AI Heading? Interview with Nick Thompson, Editor-in-Chief, Wired
"Hello and welcome to the AI today. Podcast I'm your host Kathleen Walsh and I'm your host Ronald Schmeltzer. Our guest today is Nick Thompson. Who is the editor in chief at? I nick thanks so much for joining us on AI. Today Oh thank you so much delighted to be here. Yeah welcome neck and thank you so much for joining us today. We'd like to start by having you introduce yourself to our listeners and tell them a little bit about your background and your current role at wired I Journalists and journalists since I stopped street musician. I'm in my early twenties and prior to wired I worked at the New Yorker around the digital side and then I came over and became the editor in chief of wired. I started the same week. Is DONALD TRUMP? So it's relatively easy to track them. How Long. I've had this job in about three years now. Two years in a month and my role is to assign stories at stories figure out the direction of wired on the HR office figure out our business model view with ethical questions involving sponsored content. And every now and then go on. Cool podcast talk about things. We care a lot about like artificial intelligence. Well great well thanks. We drill to heavier. Because I think you know part of the reason why we have you on our podcast. Is You have an interesting perspective. You obviously hearing from the industry and hearing from pundits and you're hearing of course from what end users are talking about but also a governments are doing have interesting perspective. That may be many others. Don't have on what's happening for artificial intelligence and so as you may see is impacting every industry. It's impacting all sorts of corners of our ecosystem. So how have you seen a evolving over the past few years and for your audience? How have you seen your audience of all in its understanding and interaction with artificial intelligence? It's interesting you know. The debate over a I has conversation. Area has changed and all kinds of wonderful ways since I started this job and started really digging into it. You know three years ago. Four years ago the public conversation over a I was much more about super intelligence and much. More bowed concerns about runaway TAC. Much more caught up in the hype over the last couple of years. We've narrowed and deepened understanding back to now. It's much more about okay. What former they I is going to be. The MOST IS GONNA BE MACHINE. Learning as far as we can see are going to go beyond machine. Learning let's talk about the ethical debates but let's not talk about them in the superficial ways. We really talked about them a year or two ago. Just talk about him and more complex nuanced ways much talk about where is actually being used successfully where it's not. I've talked about it's real limitations. Too I feel like part of my own. Understanding of the issue has deepened in profound ways. Over the last three years editing stories about hey right but I also think that the reader wired have become much more educated on the topic much more interested in it and much more involved than ever increasingly interesting debates about all things related to I. It's interesting that you say that because we've seen a lot of the use cases that we talk about and with our clients as well. We say you know this would be a great application. They're very mundane back office processes things like that and people are like. Oh yeah well. That's kind of a boring use case and we're like yes but it's an incredibly useful one and you should probably start with it and it's going to save a lot of money and time so it's interesting you know you've been seeing that as well where we've gone from this talk about artificial general intelligence is very you know super ristic science fiction view in some sense of Ai. To how can we use it now in every day? So we've talked about how. Ai Is impacting just about every industry including journalism as well. So how do you see? Changing journalism in the coming years and what ethical considerations should be discussed and put in place around this? Yeah I mean I I just want to completely agree with your statement or one of the companies. I read about a lot of facebook. And you know when I started people like Oh my God. Facebook IS USING AI. To create language that will displace humanity. And now hey is basically like a really useful tool for getting porn off facebook right. It's much simpler row tasks that are much better accomplished with assistance. Humans journalism question every profession is going to be changed in interesting ways. Biracial right because professional events systems using. I will be able to do things that we used to have. Humans do so when I think about journalism. What are the things that we do now? That machines can do better from a reporting process. I think that machines and it'll be very help and identifying patterns potentially defined stories so there will be a group of reporters class reporters a type of reporter in the future who will harness AI to analyze databases identify patterns find investigative stories. There will also be kinds of stories that we no longer need to write to the Washington. Post is already experimenting with having a systems writing basic pieces so Toronto Blue Jays beat the Baltimore. Orioles for to to happen. Because someone's home run and Joe got so they're kind of wrote stories that people like to read that you don't really need a human writing you can probably right about stock market and as time passes and as machines get better like those stories will increasing complexity. You won't be able to write three thousand word narratives profiles using an AI system. But you'll be able to write a certain kind class of story within the tasks that we have you right. It'll be changed a little bit. Copy editing an interesting profession. Where High End. Copy editing like trying to identify voice style that will stay with humans right to kind of copy. Editing Donna wired but there's a different level of copy editing that I think probably will be automated replaced by. There's fact checking that you smart a system could do so. There's certain tasks within the stack of tasks that make up the creation of an article. That will be done with I in a certain set of them. That won't be a lot of journalism in ball. Human to human contact human human reporting becoming friends with somebody eventually developing a relationship where they trust you to give you the information that makes an important story like a machine is not gonna be able to develop a trusted relationship with the source story but a machine will be able to do copy editing. The fact check legal review parts of that more efficiently than a human can so few ways. It's GonNa Change. They're probably about twenty five thousand others but that start that's interesting. You know one of the things we've noticed it. I don't think related specifically artificial intelligence but the whole idea of the long form article has really sort of. It's become a lot I mean. Maybe I'm just noticing it more become a lot. More prevalent icon. There's like bringing you into the emotion of the story and sort of like you know carrying rather than just talking about the facts of some story just today. I was reading this book about somebody who was lost in the Costa Rican rainforest and somebody wrote a book about it and and instead of just talking about that sort of brings you into the story and I found that to be really engaged. Something of course that you can't really do what they item wondering if one is sort of like a reaction to the other which is that. These sort of automated pieces are taking the place of sort of this more sorta wrote journalism and in the place now. Journalists are more storyteller switch bringing people into the narrative. I mean does something. I don't know if that's something that's intentional or I'm just noticing it more. I think that I think it is true that there are more of those stories and it probably you are seen more of them in your newsletters in the publications you read. I don't think it's a reaction to a I yet or it. Is You know. Five percent due to ask generated through. I think that is much more. A reaction to the decline of advertising supported journalism so one of the big wrecks or phenomenon dinner industries that the price of ads on websites and permissions like both prices. Going down in the Seltzer rate is going down. So it's much harder to generate revenue off of content. Right if you get if you publish a ton of stories and get a ton of readers. It's hard to
The Great AI Fallacy
"Neil. I wanted to start today's episode. Talking about something that you have been thinking a lot about lately the great a fallacy. Can you tell me a little bit more about that? You give a talk on it recently. Right well yeah. I gave a longer talk but it sort of kicked off with this I on last Friday I was invited to give the distinguished lecture for information engineering here in Cambridge which was great but the start of the talk was what I call the great fallacy and it really comes about by China. Look across what people are saying about artificial intelligence and trying to workout what they're claiming is different over previous waves of automation. So my feeling. Is THAT THE COMMON ASPECT TO WIN? The general population talks about artificial intelligence. Is they seem to feel. I have never seen this explicitly put but I think it's always there implicitly that this will be the first wave of automation the adapt to US rather than US adapting to it. So what do I mean by that? Well in the past. We've automated physical Labor in factories Transport with trains and cause and each wave of automation as Heidi Flexible Intelligence. He's ourselves requires us to say. Oh okay so now have to train station at a particular time or I've got a ton up walk particular time with everyone else so that we can in some sense of the machines said the machines automate but they're so we offer flexible link that enables them to do their work. I think even you know computers people felt with computers will would that be an end to filing and tedious work. And of course it wasn't in fact we probably all do more tedious work than ever before because of computers. Because if the something. We can't easily automate on the computer we have to sit back copying and pasting things into spreadsheets. All going through editing fixing things said the Maltese work. In order to make things fit into the computer and I get the sense when I look at the underlying theme of what people seem to think about artificial intelligence that people have a sense that this isn't true anymore that because we've used the word intelligence. The end people think of intelligence being like us that the machine's going to accommodate has better. That just seems to me like an auto fallacy that there's no evidence of that tool what we're doing if anything more of the same way managing to find in this case instead of bits of physical work that we can repeat and automates as we did for machines with finding often bits of intellectual work that we can repeat and automate through machines so it feels very much like the truth is more of the same but the promise that everyone's making is no no. This is going to be different. Machines in amongst us just accommodating oss which the feels like no evidence full. That's fascinating so so we are sort of overstating. The flexibility of this new wave of automation. You think when we when referring to the kind of work that in the public conversation that people feel like artificial intelligence will do. And we're just kind of over blowing this thing that we've seen happen before. I think we're just we're not even stating it. I think that that's the issue so it's not so much where overstating it but we using terminology that implies implies to people something that won't happen so. I I give an example one of the big recent papers in Diagnostics with basically neural network continuously throughout his paper refers to the neural networks. They built as an AI system. Now I don't think that they're doing that to necessarily misrepresent but it does misrepresent because if you say I system it implies that the some sort of well the word intelligence means something to sort of people who are not even lay people but say doctors demane experts. That is a bit more than old. I've got functional EURONET. What model where I give it some inputs and it will give me an output. They don't think of that. So if you're using terms like that your encouraging them to believe that the system you built somehow significantly different from what's gone before and I don't see any evidence of that for me. It's computers in statistics and the fallibility of these systems is similar to fallibilities. We've had in the past or perhaps but perhaps it's hard to understand when those problems are occurring so in that zone of the area where there's a gap in understanding people filling it with bits and pieces that come from a notion of intelligence which is associated with our intelligence and just absolutely not present in these these artificial systems so. I don't think it's so much so people are overstating it but that conforming to a narrative and very natural narrative that occurs when you start using the term ai and people on Highlighting this challenge so people's expectations of what's going to be delivered. I think he's quite different. From what will actually happen which will lead to challenge to be seen in the past the same type of challenges. We've seen in the past where you have some automated system but in reality there's a large number of humans walking around that system to ensure it can walk and an existing example would be the extent to which say voice agents whoever's intelligent agent. Siri Electoral Google's are relying on Human Labor to label data for them. So there's an enormous amount of tedious work for humans producing this label. Dates ruin in a form which the computer can consume in all the computer can emulate that type of behavior. Now you might say well what's being done. That's the Nice thing is we don't have to do it again. Multiple Times. And that's true but all you've created machine that can repeat that toss it com. Do General Flexible intelligent things. It doesn't go hang on. Actually maybe I shouldn't be responding to the voice right now behind. The real thing you want me to do is say. Hang on a minute. You sound stressed. Allow me to tell you to sit down and have a cup of tea is not going to do that right so our own. The narratives that were using around these ideas are perhaps a setting ourselves up for disappointed expectations. So should we. Should we start? What language would you recommend? It depends on which sub field. You're in I think the closer people ought to the AI field in the more they deployed themselves the better their capacity to understand the problems but in the wider population. That promise is it's as I say it's not being explicitly made but it sort of implicit in the use of terminology on. I think there's very many reasons why that won't happen. I think is really interesting question in how you would build such a thing and I don't want to suddenly turn it into a conversation around artificial general intelligence because I think that that's also a separate thing. There's a danger of blundering into this and I think some professionals doing as well some of the conversations you from researchers who are doing extremely good work and doing things that we didn't dream we would be able to do some number of years ago like in translation or transcription or image recognition. Believed that this means something more about what we've achieved intelligence at any basically doesn't these role perception tasks. What's interesting is the boundaries to which would be enough to push them so I don't think I would have necessarily known ten years ago. The quality of machine translations we would have got from what is basically a mapping and inputs to output mapping without having a core intelligence underlying this thing and it is interesting to what extent those those mapping za generating representations which represents the awkward human so G. P. T. Two I think interactions the key to a kind of extraordinary for the humanity of some of the responses of course significant weaknesses in those interactions to the lack of structure the lack of conforming to any real well narrative in the longer term but the direct one-to-one responses are quite extraordinary. Now I think all those things are amazing but they don't imply that we've got this very adaptable intelligence. That is capable of responding in sensible way to situations Utterly on conceived of before the system was deployed. Which is what you have in effect with natural intelligences Hayes. Yeah exactly so. Maybe we need to think of these these tools and be excited when we see moving the needle around flexibility. And perhaps were were over indexing on this word intelligence in maybe we just need like an internal cultural shift that hopefully would then sort of sift out to the people who are using these tools in this understanding that the great strides that. We're seeing around. This is in the way that we have increasingly made automated tasks more and more flexible but the ultimate flexibility is is an intelligence right. But can you even ever get
Visualizing Fairness in Machine Learning with Yongsu Ahn and Alex Cabrera
"So let's get started with the with the topic of today so today we talk about a really really relevant topic can needs It's particularly hot right now. We're GONNA talk about bias in fairness in machine learning. And if you know know what this is we're going to describe and explain what this is about in a moment and more specifically what is the role that can play in this specific domain to say mitigate problems that can arise in terms of bias and furnace in machine learning so to talk about this topic. We have not one but two guests. We have Alex Cabrera who is a PhD student from Carnegie Mellon University. I Alex is again. Thanks so much for having and then we have young. Soo on who is also Ph student at the University of Pittsburgh I- youngster. Welcome to the show. Hello Nice to talk to you so Alexander. Who can you briefly introduce yourself? Tell us a little bit about what is your background. What is your main research topic? And just give a brief introduction. Yeah so I'M ALEX. I'm a PhD. Student of the Human Computer interactions to at Carnegie Mellon so generally idea research into creating interactive systems and visualization. Systems that help people both develop better machine learning models so even more accurate more equitable and understanding these models so understanding potential issues. Or How? They work okay. Young Soo my name is sue on in I'm a dirtier peachy students at University of Pittsburgh. A my research interest lies at the intersection of visualization and fair. And explain away. I enter to machine learning so my primary research question is to build assistant to help users with making the machine learning results more fair and explainable in helped him to interact with machine so that their opinions and apex can be incorporated into the system. Okay thanks so much so I was thinking. Maybe we should start with defining a little bit this terminology to the extent that he's possible but maybe they're probably many of our listeners who've never heard of that and of fairness and bias and this is a very overloaded terminology here so I'm wondering if we can start by defining a little bit. What what we mean by fairness and maybe even bias in emission learning and also what? What kind of province exists there yet? So I'll probably I can start by talking about a little bit of background on why the problem This fairness problem has been actively discussed in especially missionaries research. So as a May have seen did. Data driven decision is kind of increasingly used in important decisions so especially Such as a job recording of colleagues dimension were predicted policy. Those kind of important decision which have kind of huge impact on Individuals muster learning as more and more used induced kind of important decisions then Some of cases have been reported that these machine turned out to be biased towards certain groups or certain individuals so here the what I mean by bias is certain. Decisions are kind of burrow favored to certain groups or individuals. Such as man over woman or a white people over african-american people. This is because on the machine. Learning model is trained from Historic Co. Data set and this historical data said could possibly include Inherited bias then. The model is kind trained by those data sets and then have kind of inherited vice. The problem of machine learning here is that whatever trained model can kind of systematically discriminate against certain individuals groups especially in Western Assistant Because many decision makers may use to system in their decision making then kind of making these mistresses. More Fair is kind of important problem so basically the type of fairness you talk about is mostly related to not being discriminatory or not using features. That have nothing to do with the essential decision. You're making more superficial like Maybe the race or gender or other features of a person right. So it's about combating discrimination. Yeah I think that's the main idea. It's actually you get to you a more complicated because even if you don't include some of these protected features so if you say you're trying to give someone alone you don't really want to decide that based off of their gender their race Those are actually. You can be almost perfectly predicted by the other features so you can actually reconstruct that so actually a lot of machine. Learning people suggest you actually add those features in because they're going to be used anyway and then you can apply some resolutions afterwards to try to address the problem. So it's very much embedded in the data that you're using to train the model this historical data that you've collected so it's not just as easy as leaving out that column with Race Agenda and not saying you talk to the research that's happening now. It's a little bit more complicated. Just the complex relationships between the variables ends up that you can actually recreate the biopsies. Even having no idea algorithm not being aware of these protected attributes. Okay but just on the Senate so the evaluation you do and fair. Evaluation is one that only takes the features into account that you're supposed to take into account so usually the way we did try to define fairness or quantify is an output. So if you're trying to give loans or a very popular example is trying to decide algorithms to decide how risk how likely someone is to recommit a crime if they're like. Oh so whether or not you should give someone bail we usually it doesn't we don't really look at what features are used that we look at. What the output is and so if for example the RECIDIVISM prediction case for African American males? You're more likely to be given a higher risk or even though you're just as likely to recommit a crime that is discrimination. That is the bias that we're trying to discover and trying to combat right So we really like black box models. It's really hard to know. What parts of the data are being used to make the decision? But we really care about whether these decisions were making. The outputs are making that really society impactful whether those are equitable and fair. Okay Yeah I'm wondering if we can can you? Maybe describe one or two specific examples. Where these kind of problems can arise. I think what is interesting? Is that right now? I mean we live in a society where where these these systems stems are already making decisions or some of some decisions for us right or providing indications for for experts that have to make decisions based on on what the AI system suggests recommends right. So I think I'm wondering if we in order to make a little bit more concrete if you can cite one one or two examples where where these these. This kind of problem can rice yes. Sadly there are quite quite a few examples. So one of the biggest one of the first investigations elected to it was in facial recognition systems so there are systems by like Had some face plus plus IBM and Microsoft that they audit and it tries to tell given a picture of someone's face whether they're male or female and when they started looking into it they found that when you start seeing how well they perform for say white men versus darker skinned women. There was almost ninety nine percent accuracy for the white males and close to seventy percent accuracy. For the darker skin females which is pretty big disparity. A lot of that is due. Hey if you look up. General data sets of faces a lot of the faces. That come up are white males that data that you're learning on is not
Science News Briefs From Around the World
"Hi I'm scientific American. Podcast editor Steve. Mirsky and here's a short piece from the February. Twenty twenty issue of the magazine and the section called advances dispatches from the frontiers of science technology and medicine. The article was titled Quick Hits. And it's a rundown of some science and technology stories from around the globe compiled by Assistant News editor. Sarah Lou Frazier from the US off. The California coast scientist measured a blue whales heart rate for the first time using a device attached to the animal skin by Suction Cup the heart likely weighing hundreds of pounds beats from the thirty seven down to two times per minute varying dramatically between diving feeding and surfacing from Peru researchers analyzing satellite. An imaging data have found one hundred. Forty three new Nazca lines. These are largely line drawings of humans animals and symbols etched into the Peruvian Landscape Millennia. Ago The drawings including humanoid figures sixteen feet across spotted by. Ibm's Watson AI system from Brazil despite the long dry spells in Brazil's Catchinga region. Scientists found the tree hyman a Conga era drizzles copious nectar from flowers to attract pollinating. Bats full-sized tree can release two hundred forty gallons of the stuff with thirty eight distinct sent compounds over a single dry season from Norway. Archaeologists ground piecing radar found a Viking era ship surrounded by a filled ditch lurking below the soil of a western Norway farm. The ship was once within a burial mound from Jordan. Researchers uncovered a two horned figure in early Islamic ruins that may be the earliest ever found the roughly thirteen hundred year old object matches a rook found in Iranian chess. Set from about four hundred years later and from Ethiopia microbes thrive in many of earth's harshest environments but researchers found no life at all in briny scorching civic pools near Ethiopia's Dalla volcano knowing the boundaries for life's adaptations helps to narrow the search for earth like life on other planets. That was quick hits by Sarah Lou
AI in Legal
"Hello welcome to the AI. Today podcast I'm your host Kathleen Walsh and I'm your host Ronald Schmeltzer. So what are they professions nations. That has been around forever at least as long as there have been laws. have been lawyers and you know regardless of what we might think about lawyers in the legal profession. It's part of what allows us as a society to operate under the rule of law. Right the thing about the legal profession is that it's dominated by ED humans and paper documents and of course any if you've been listening to our colleague Linda Today podcast. You know that when there's a lot of humans and a lot of paper there. There is a role for artificial intelligence. Oh there is so we want to spend some time today talking about how a I is changing augmenting the legal profession so as rod mentioned mentioned. You know there's a lot of paper that's involved with any legal case and you know people are looking for ways to make this more efficient to save on costs costs but also to save on time because it's a lot to go through all these documents you know laws can change and get updated and new case litigation can come out that that makes the interpretation of the law different and people need to stay on top of all of this and without technology it makes it very very difficult so when you think of the Law Profession Asian and of course there are so many different aspects of course there's the reams and reams of contracts contracts between businesses contracts between employees and then you have personal contracts financial contracts Lexi of insurance contracts. These are all legal documents right. Ninety nine point nine nine percent of them you know just our routine just send you a contract you sign it now. You're signed up for on my credit card. You know you've agreed to the terms of this insurance agreement you've bought a product from. Here's a proposal. Sign it right you know or could be those sorts of things that are maybe arguments commits in front of the court trial proceedings. That could be you know depositions. That could be discovery of documents especially if large cases and the thing about all these lawyers are not necessarily the lowest cost of profession. Right I mean for a good reason you know good lawyers spent a Lotta time going to law school. They have to basically certify themselves in the case of the bar and they have to continue to presents itself. They're very professional society. But a lot of these activities are not don't really require the experience of law school. It requires like okay. I need to go to this document and redact all these things go through this document and find all these clauses or like look through my millions of contracts and find the six contracts that reference. This thing right. That's supposedly the world of paralegals in junior law professionals and that's the area where artificial intelligence may take a real. Yeah and according to the law technology today they're saying that over one hundred thousand legal roles will be automated by AI systems by twenty thirty six. So that's a lot of roles that can be automated by automation and buy more intelligent automation as well. So of course you know we talk talk about our seven patterns a lot here at Kaga and something we refer to a lot in our. CPI training as well. So if you're interested in applying these things to your own industry you should basically look at the methodology and some of our training the long story short we have recognition patterns. We have as conversational systems. We have autonomous systems. We have hyper personalization and we have goal. Driven Systems have patterns and anomalies for finding anomalies. And we have predictive analytics and every single one of those patterns is actually applicable to the legal profession we have in the recognition recognition pattern the ability to scan through say video or images. Find something we have even within documents that it's able to identify what's on the page all right quickly. He classified documents for example right. And that's something. We actually see quite a bit one of our trainees. One of our classes talked about the fact that they are in the public defenders office and they have a need to basically just classified documents or categorize them really based on whether or not they should retain those documents some documents chain for several months or one year or seven years or they can be you know gotten rid of right away so an example that he gave was if somebody goes. I have cake in the break room. That email can be you you know discarded right away. It has nothing to do with anything but if it's a legal document like a will or something else that may need to be retained for seven years. Well is it really really a good use of time to have a lawyer who costs a lot of money. Be going through these emails being like. Oh this is about the K.. Goes is about someone's retirement party or can you have a system go and automatically automatically do that and then the ones that they're not sure of flag for human to review which will dramatically decrease the amount of time that humans are doing right looking through these documents. That's just just that one pattern recognition pattern or maybe even the conversational patterns. We'll get into in just a moment but you can look another pan. The autonomous pattern autonomous legal. What does that mean well? There's a lot of things. First and foremost you know The whole managing the legal professions. You can think of legal practice management. They're starting to use. Ai Systems to automatically track hours that are spent Danton different cases automatically spot patterns in the work hours and basically helped make lawyers more efficient and of course hopefully at the end of the day make better for their clients that they're not spending two hours hours on something you know of course and billing to our rate which I can guarantee you cheap and instead spending thirty minutes on something or fifteen or maybe just completely automated at all. Maybe it's hard just say you know of the legal. Profession starts using more of so called robotic process automation tools and process automation tools. Charge the time that that bought is working on something who knows the paper at a minimum hopefully cheaper but basically yeah legal bought has worked on it but it but this is part of the patterns and anomalies pattern as well spotting adding the mistakes looking at anomalies as to as to how the lawyers are performing their time so that way they can basically solve any invoicing mistakes timing issues shoes and the client has better visibility into what's happening there right because one thing that we talk about very frequently on this podcast is how fast computers can just go and turn through documents at a much faster rate than humans and also. They never tire. They never get fatigued. They can run twenty four seven so you know how more how much more proficient could be loft practices be especially if they're working on very large cases if they can have an autonomous system. Go through these documents and be able to catch things things you know. This could speed up months of discovery and months of legal time exactly so we are also seeing this idea of the conversational. Oh legal assistant right that way. A lot of people have have routine legal questions especially around things like you know their financial planning to legal planning the will or maybe even in aspects of the other things like you know trying to get a lawyer's opinion on aspects of divorce or of contract matters or of dealing with your neighbors or something you know you know there's lots of different applications and you don't need to spend you know an hour or two of a lawyer's rhyme answering these things so we're seeing. The idea of the frequently asked is questions. Yes making its way into the conversational legal robots yes. I'm feel like back in the day. You know. Sometimes little newspapers would have you could write in. And then yeah they they would say oh. This is my complaint with a neighbor. How do I do this or my trash isn't getting collected or you know little legal legal issues in a lawyer would go and write weekly pick a few and do that and I feel like now instead of that where the lawyer goes and picks two or three and writes about it now the majority of things can be answered and in real time just about so? Oh how do they go about doing this. And how do they make it cost efficient because to have an actual lawyer sitting behind this chat Bot and answering questions as they come in might not be a good use of time. I'm for that lawyer. But that doesn't mean that the firms till can't offer that service so we've started to see more intelligent. Ai Enabled Chat bots where they're doing more more than just answering an FAQ very general FAQ's that people can find on their site. They're starting to give. I don't know if it's exactly legal advice. But it's legal suggestions gestures and saying well look here or we've seen this in other cases or you know this is. This is something that you can look into
"ai systems" Discussed on WGN Radio
"Like to create lookalike sites and make sure anti virus software is up to date use your credit card not a debit card to dispute any suspicious charges people spend a lot of money over the holiday weekend they spent a record seven point four billion dollars online and that was just on Black Friday alone sales are also a four point two percent in stores as well shoppers spend another four point two billion dollars online on thanksgiving day driving and texting could become much harder to get away with the government of New South Wales in Australia help cellphone detection cameras to crack down on drivers of legally using phones behind the wheel I think it's only a matter of time before Chicago guesses to the cameras use artificial intelligence review images and is that phone usage by drivers and a test run earlier this year Fishel say the technology caught more than one hundred thousand drivers on their phones wow here's a holiday season challenge it sounds easy but it's really not that he's not he's not you have to do is make it until Christmas Eve what about hearing the original version of lambs holiday hit last Christmas I think I've already lost the sun if you're playing a one to mute your TV for the next few seconds give you warning right now because we're going to play a little bit of that song right now I guess we're out yeah I was I was soon as cover versions of the song do not count as hearing it it has to be the original version from land the the race limits on sisters might not yes I swim again we have all women get yeah all set it was fun you know when we last one December twenty four hours never seen this music video awful really good sound like maybe like my yeah you know there's very little music video I'm going to get into the train and walk out with my skis and restrict the mind okay trending news talk about this morning I love the story cat was stranded cross country in the frigid cold but a microchip changed his story S. is is the cat's name usually lives his life on the road is a traveling companion alongside his truck driver owner but he went missing one day at a stop in Ohio for months its owner is been searching for ashes until just the other day he was found four months later woman found the cat Bridget outside brought it to her home in New York there a lost and found team discovered ashes was fifteen hundred miles from home registered in Texas they called and soon reunited the cat with his owner owner says having ashes back is a Christmas miracle that picture is a door yeah and that the owner you can see yeah really really happy to have found that a recent study shows under water speakers could save Australia's Great Barrier Reef generally coral reefs are noisy places when they're alive you can hear the crackle of snapping shrimp another fission it soundscape but a dead coral reef is quiet scientists discover that when they played a recording of a lively reef in an abandoned area twice as many fish reappeared answered saying coming Fisher essential to save the reefs settle in and create space for new coral to grow be sure about that and I think I'm the only person here is watching the Mandalorian the new Star Wars shown but as a baby yeah I've heard about the baby taking me by starter so many memes.
What Does it Mean for a Machine to "Understand"? with Thomas Dietterich
"We're really digging into this topic. what it means for machine to understand A recent blog post postive yours and I thought to get things kicked off. I read your couple of the opening sentences you wrote critics. A recent advances in artificial intelligence complained that although these advances have produced remarkable improvements in AI system these systems still do not exhibit real true or genuine understanding the use of words like real true and genuine imply that understanding is binary system either exhibits genuine understanding or. It does not the difficulty with this way of thinking. Is that human understanding. We're standing is never complete and perfect as so certainly the way you've laid that Opening argument out it resonates with me. I recently Had Gary Marcus on the show. This was back in September and we spoke about the book that he recently launched rebooting. Hey I And he's pretty. He's very outspoken. As a critic of deep learning. And maybe that's not the way he would put it maybe he'd say a critic of deep learning kind of as a standalone path to Artificial General Intelligence but union reading the blog post I couldn't help but think of Gary. Marcus says being although you didn't name him kind of You know person in absentia that you're writing adding this to maybe talk a little bit about the you know the broader context for This post and maybe how you know what prompted you derided right. If you know Gary was part of you know what you're thinking about Or not I'd love to kind of get a sense for where you're coming from here. Well certainly you know Gary and I have had Opportunity even to engage in formal debates and Gary As I was saying I think Gary's main points I generally agree with which is that there are lots of obvious. Shortcomings are existing systems and in particular These systems based on deep learning But but Gary He and other people can't seem to stop saying things like well when we look at the behavior say of Google translate It's clear that it doesn't. It's not exhibiting real understanding of the of the languages translating or or When we talk about Siri that series and doesn't really really understand What we're talking about And I've been making the counter argument. Yes these systems are understanding. And it's it's real understanding but it is narrow understanding and So I am criticizing the use of the word real to mean deep and complete understanding because that denies that these systems are doing anything that is intelligent or that is exhibiting real understanding and I think that puts puts you in the position that you will never be happy with any system because no matter how good gets it will make mistakes and exhibit failures and it's understanding and are you going to say well when understands ninety five percent of what people say that it's still not real understanding. I mean what you're pushing yourself into a a a belief that there's some magic threshold that if you could somehow cross it you would have a system that had real understanding and I don't think that's the way it works. I think that the way it works works is that we make Incremental progress sometimes. Bigger leap sometimes no progress for periods of time As we were doing in speech recognition for for a while in the nineties but our systems get better they are able to understand something so as I say when I tell Siri please call Dan and he calls the right person. It has understood me for the purposes of that utterance for that task. Now if I said you know Siri tell me what Dan means means to me And it doesn't really know anything about what it means say for To be a best friend and you know but let's many people of course have remarked that it's Impossible for two human beings to fully understand each other so That brings me back again. To what is is it. We're really trying to achieve when we build. Ai Systems and as an engineer. I would say I want systems that can make the appropriate response when I asked them to do something. Or if they're warning me if some situation in the world that I should be paying attention to and so on and to the extent that they do that correctly directly I would say they understand What what what I want them to do when we have these kinds of conversations? I think it's you know there's a slippery slope it kind of devolving to defining every term in the arguments But In this case I I wonder the extent to which we're talking talking about different types of understanding Do Do you think that that is The case at all here. I don't know that there are different types but certainly different definitions. Aw I mean obviously we are arguing about definitions and In My blog post I. I was supporting the view that instead dead of argument about our definitions we should be trying to be should be asking ourselves. Well what tests would you give to a system in order to evaluate whether it's understanding in a doing a particular type of task well right if you say well this is is not a narrow Show me all the the Things that you would like to do that. It is failing to do now And I drove through the analogy to test driven development and software engineering Right at the test. First and then use those to decide how to engineer the system to try to meet those tests And then keep writing more tests. I and I think Gary has actually jumped on that and on twitter. He's been Asking people you know what's wrong with our current natural language wjr understanding tasks and because it seems that we can often get an assist him to do well on the on a particular benchmark task. Ask and yet again. It turns out that it's very narrow and it's not doing well on any like immediately adjacent tasks that we would like it's a new And and so some people have been. There's been a bit of a discussion now about that. And and I think that's really where the discussion needs to go is You know and and Gary himself himself in in our twitter conversations I thought articulated beautifully said. Okay I want the computer to be able to say Rita Story and tell me The answers answers to the journalist questions who what when where why how So why did this person do this. What did they do? When did they do it? You know Ordered these events for me correctly clean time and those are way beyond what we can the state of the art in a and natural language understanding some of the people in the natural language community said just stop using the word understanding at all. We've just caught language processing because we know that That this word understanding Sets expectations for something. Something that is very broad and deep you know duck. Hoffstetter had a very interesting piece that came out last year where he analyzed Google translate and showed how many many cases where Google translates understandings clearly extremely surface oriented and often? It can't understand anything about did. Did John Do something to marry was married doing something. John knows that John Maher need to both be translated In into the different language language. And it certainly doesn't have any of the You know connotations and Depth say that would be required to to translate more poetic language would your metaphorical language It really has no understanding of Human Social Relationships What that might make Mary angry? What might Make John Happy. you know just it's completely clueless about that because all it has been taught to do is is to translate from Chinese into English English into Chinese for fairly straightforward every day sentences and certainly not trying into translate Shakespeare at all and one can imagine that it could make very serious mistakes result in say highly highly emotional and complex Social situations it's fine for. Where's the nearest bus stop? But not so good for you know so why. Why aren't you talking to me anymore or something? I do want to encourage us to move beyond saying well either understands it doesn't and this understanding is true true or it isn't to say well. This understanding is incomplete in these important ways and and what we would need these to do and so for example when we think about reading a story or just engaging in a dialogue we need a systems that can be building and maintaining an interpretation of the dialogue and this is well known in the natural language community. We just don't know really how to do it at scale. We can build applications in a narrow domain Say Purchasing airline tickets or something where air we can cover a lot of different linguistic phenomena and An have Quite good performance but as soon as you step out of that narrow domain that breaks down.
The Machine Learning Platform Behind Linkedin
"Let's get started by. I think everybody knows Lincoln. It's not something that we need to spend a lot of time explaining but many of you don't have an account on Lincoln. Okay everybody market saturation but let's maybe get started by talking about some other ways that Lincoln is using machine learning earning. We often say at Lincoln machine. Learning is like oxygen right so everything we do has machine learning built inside it like if you go to Lincoln the first thing you want to do when he joined. Lincoln is get connected to people who can help you now how we do that. We recommend you people that you can connect to. That's it's all power through machine. Learning once you get connected to people then you start consuming content that they produce on your news feed and you know there is information overload Lord on the field. You can see so many content so what kind of content you want to see and Wendy. You want to see like for instance. If you're looking for a job you want to see job. Recommendations were looking for a job if you are very happy and you want to learn more deep learning and if you are connected to Andrew and he publishes something then you want to see back on your feet so you need algorithms to scale this process again. That's all powered through machine learning. If you OUGHTA market here and you want to target the right audience the entire advertising ecosystem. We all know work through machine learning. If you're recommending jobs that you can even if you're not looking looking for a job we still recommend you jobs because there's always a better opportunity out. There are all of you. That's all through machine learning if you're recruiting and trying to source candidates alter machine learning earning if you're a salesperson trying to close a deal who are the decision makers how to how do you reach a decision also everything we do on Lincoln Product. Whatever you see on the APP it's all powered sort through machine learning and finally you know this. Is something that goes behind the scene. We have to keep the site safe right there. Are you know a lot of bad actors out there producing content turn that you're not even reach you. There are people who create fake profiles. I mean I've seen a lot of fake profiles of famous people and that's not a good thing right so just to keep the ecosystem clean. That's again machine. Learning plays a very important role in that as well so everything we do at Lincoln is powered through machine learning in fact when we create a new product idea in addition to product managers engineering managers designers. We also have a machine learning person sitting there right when we are designing the pride art because we believe that's the right way to do things. I mean you know the US that you create if that can actually some important feedback loop aw that can play a big difference I mean under was talking about collecting data will. How do you actually ensured that you collect the right label. You have to actually start working on it at the design fees. It's too late if you don't pay attention to it at that stage then you have to build very complicated model that essentially do guessing right like I guess if you can actually get the right data and and so that's why it's very important to start that process from the very beginning of shing learning process of product product process one of the things that it has always fascinated me and my conversations with folks at linked in is we think of Langton relative to a more traditional enterprise that it's kind of the digital native company born on the web you know the product is web but in a Lotta ways. The company has evolved similarly it you know it's initial investments in machine learning and what the way it's supporting machine. Learning today are very different. Can you talk a little bit about the journey at Lincoln and how Melania has evolved over the years. Yes when Lincoln was always a data I company right like if you all remember the word data science was was what's coined by DJ Patil at Lincoln so so you are always very savvy about data. We knew our businesses all about the date of the unique data repossess zest right so we were always doing data science always doing data product innovation. We also started doing machine learning very early on in two thousand seven. The first machine learning product real product was the people recommendations right so in those days we would compute. We will have simple machine learning models of course you know so. You'll have a simple model. The teeter the features of these models there are a handful of features are very carefully tweaked based on infusion once we have that model than production is in these models at that scale was still very difficult so we will actually build her duke systems that will do the ranking and scoring off line right so because online was not very well developed and then we run and these processes everyday writer search was another system that we actually develop very early on that use machine learning fast forward two thousand twelve. We got more sophisticated indicated right so the first sophistication we added in terms of machine learning goes in our advertising system right so advertising system. Most of our other recommended systems are based on simple collaborateur filtering idea at those times people who are also barred this advertising was the first place where he added a lot of sophistication he added. We built near real time systems homes. We build online systems that can score things run time more complex models and you know encouraged by the success we got there. We then attack defeat problem. The news feed problem for those of you who have been using linked in for a long time. I'm sure most of you will tell me today. The newsfeed is much better than what it was five years. That's all due to machine learning and items. A lot of work happened to kind of add sophistication to the news feed algorithm and once we got success in these two big application then we started thinking. How do we generalize it across the Board Right. Why why just advertising wise this new few new sweet. Why can't we build a platform that can actually generalize to everywhere and that's what we have been doing for the last few years and so we have program at Lincoln called pro l. productive machine learning and again. I think a lot of companies have platform but I mean one unique thing about our our platform is building a platform with a very strong opinion right so you can build a machine learning platform that can cater to a lot of tale users tonight so if you're a cloud company you're going to build machine learning platform that can cater to the needs of a diverse set of customers. That's not our goal right. Our we know that our. Roi Is going to come. I'm from a few big applications and platform. The build is really suitable more for that right so large-scale recommend. Your systems are skits sort systems large-scale classification. Lhasa fixation problem. These are the problems that we face and our platform is really geared towards that right so we also know that we reached a point where without adding more sophistication -cation to our systems are going to get the Roi that we used to get so I give you an example like two years ago. Job Recommendations system we revamp the model we kind of moved away from a simple linear model to something more complex involving deep learning involving something that is a homegrown technology generalize mixed mixed model so I mean I'm not going to technical deal. These are very high. Dimensional model had lot with this war technique in statistics nonsense the seventies he's an exit was applied to application that five hundred patients now you know those those five hundred become half a billion patients and then suddenly the explosion in the number of parameters complexity increases a lot right so we applied that and you know we've we've found a thirty percent improvement in design that was stunning running and so we are all very happy but for the next six months nothing happened and then like it was very surprising like okay well the engineers all go to Hawaii or what's what's happening right like why is nothing moving and what we realized this when we when we introduce complexity the tooling did not keep pace with that right so it became very hard for the subsequent engineers to kind of a trade on this model because we didn't build appropriate tooling that enabled them to kind of entry so that was the realization on this is not going find work as we actually start introducing more sophistication. The industrial process will only work if the engineers productive and in order to improve productivity the venue add more complexity especially for large scale distributed systems if you really want them to run efficiently. If you want them to run a reliable fashion you have to make. I'm sure that the tooling and infrastructure can keep this innovation that we are doing and so that was really the impetus of this project that we run called Promo and we actually run it. Very rigorously is largest building platform components. We actually measured the success of that so every week. We measured the number of successful experiments. We have run so there are a lot aww experiments are engines run but we only track the number of successful experiments right because otherwise you can start eating right like someone can just parameters sweep on the grievance. That's okay dude a parameter sweep of hundred different values and so I ran hundred expert I don't care I mean you know you can run. Hundred experiments to experiment did how many of them succeeded so so that's our metric and we have seen more than thirty percent improvement in the number of successful experiments that we run on the site after introducing using this program. It's still not done. There is still a long way to go but you know this. This has been really useful for us. It has kind of also barred teams together so earlier you know if you don't have a standardized nice to have doing things no matter how hard you try the culture in the team would be different from the culture in the job steam right and that's not good right. I mean we don't want to create different cultures in the same company in fact we weren't given that we are a centralized organization. We want people to flow from one area to the other so you did your tour of duty on the feet and you should just go to the job steam and learn aboard the jobs product and he should be productive in the day right and that is only possible if you standardize things and so this project has also helped us to standardize things things in all kinds of deviate too
Facebook to train AI systems using police videos
"Facebook and law enforcement agencies are teaming up to combat extremism the social media giant says it's working with law enforcement a train it's artificial intelligence systems to recognize videos of violent events as part of a broader effort to crack down on extremism sienet editor at large Ian share says this is something that other social media platforms struggle with and it's a daunting
"ai systems" Discussed on WORT 89.9 FM
"F. M. eighty nine point nine FM at six oh eight to five six two thousand one that six oh eight to five six two thousand one. we're gonna be moving on actually have been a little while from this topic to talk about some other open records issues but I have since we got built the studio we do have another cast who has just written a column recently about. access to public records and the cost of it and what we'll get into that soon but I want to ask you bill one more thing about the the body cam issue I mean this the the the issue has been most important in terms of looking at police use of force and and it's at least in terms of the the media attention to the use of body cameras it's been access to. two body cancer police videos in situations where police have used force and in many cases killed people. so but there are all sorts of other videos out there as well and they get and obviously they aren't regulated because it's the wild west of the internet when everyone has a video camera the racket so how does that affect this situation I mean is does that mean that media should be if they're responsible only paying attention to the official. these video now or call the videos that are available for certain incidents you know it can be part of the you know what the public sees and part of what the public is able to learn about an incident there will be things that people will capture that police body cameras wall one of the things I've always been surprised of when I've seen police body camera footage is how bad it often is right you know that it's pointed in different direction from where the action is I don't think that's always intentional it's probably mostly accidental someone is running and you know it's not quite like like they're they're filming it for you know for the for the theater so it's just off off camera right at that things are occurring so yeah they'll be other other things the other thing I would say is that it's not just use of force situations that will become a captured by these police body cameras they will also be interactions that become the subject of disputes won the race I've done a lot of reporting about in the city of Madison is complaints by citizens against officers the officer said the US you know I'd like to report him for what he did which to me what he said to me the way that he acted. and all of those things now can be confirmed or denied right right by the police body cameras camera footage which will be taken and will be required to be maintained one then there's the I mean there's just searches in general of the body cameras on it was looking through this bill the the section on this this new body camera bill on the retention of body camera data there's this. there's this category of search that I didn't know that much about I'm assuming you know more about it that search during an authorized temporary questioning commonly referred to as a Terry stop. it is it I'm not sure I do not okay a situation a cop says Hey you mind if I look in your backseat uhhuh I think that's what they're talking about but I could be wrong okay. other searches that are done absent a warrant by yeah you know the casual granting of permission I just years ago about how there was a the plea the dean county narcotics squad routinely made drug bust by going to people's homes without a warrant and sane human who come and look around yeah and people would say who you well yeah thank you Sir you know because of the cop standing up right or were they could have said no and that with the cops would have no recourse right but they let them in and they find drugs in the charged with a crime and you know they're all the lock right yeah. well this is one of the section that requires all body camera that if you retain Fuhrman of hundred twenty days and provides exceptions for longer retention so it can be retained until the at the investigations over or an encounter that resulted in the death of any individual or actual or alleged physical injury and counter that results in a because still the arrest or research during an authorized temporary questioning so you're saying you're not sure what that is I'd like to pursue that a little different from my own just to what is a Terry stop what what's an authorized temporary questioning. but we'll leave that there for right now and it also contains an exception to the general the body camera that are open to inspection and copying relating to the treatment of mine yeah we talked about that so that's the pixelation issue where right they they can release it but they're gonna do some doctoring up yeah that's like yeah the kids faces juveniles to be and victims right you know cops respond a lot of the situations where something horrible has happened to a person they're getting information from that that person and you know if someone is a crime victim and there is no larger probative or public value in the video of that recording the police will be able to say sorry this is on the record we're gonna release right importantly the bill provides for a mechanism for challenging that if your news outlet for instance for someone who's who I was it deeply interested in a particular video that they say no you can't have you can challenge you can sue them over it and if the courts decide that there really is a public interest and they it really should be released you can get those records still. okay so I'm. that's a good segue actually into talking about other issues in terms of access to public records and so we're gonna be joined soon within the next four five minutes by Jonathan Anderson who is a PhD student at the university of Minnesota right now and is a former Wisconsin journalist and I guess I can ask him but bill do you know who is he writing for before. because when they get that paper say okay he's a really good journalist and pay someone who's had a an interest in public records going back to when he was a student journalist at the university of Wisconsin in Milwaukee he did a lot of things involving records request and he's done research into of government related issues like you did a study of requests for interpretation of the law that are made to the attorney general's office to the justice department take a look at you know how many requests there are how often they how long they take to respond you know how often do they. follow up with it vice or enforcement okay he's he's done academic for want of a better word research into open records questions what we're gonna be getting him soon but I do have one caller here who is concerned about of fear of calling police even for emergencies due to bad experiences with law enforcement and so this is an anonymous question is anonymous on the phone to join us to expand a little on the question. pardon me am I on you are on the air yes I just I'm I'm an older lady hi I have just a couple comments to make that make me afraid to even call the police department in Madison or sun prairie for emergency help because of a couple of episodes that have happened to me one being my son at eleven years old our door was kicked in by an old boyfriend of mine and they got him out of bed and he has a sleep over that night we were all sleeping and this guy kicked in the door he splintered it and he tends me in the bed. on my belly so I couldn't see who it was and I started beating on the wall and the neighbor called the police I said call nine one one call nine one one they came they put me in the hallway on my **** immediately when they got to the door they pushed their way in and made he they wanted to interview my child I said that's fine but I want to be able to see him I say excuse me officer but even pre Cepeda files I'm sorry I have to be able to see him you can't shut that door and when he shut the door I come up off the floor like a mother lion would have and they took me to jail for resisting arrest and and disorderly conduct and then there was an episode a couple years ago I was sixty four years old and the sun printed in the paper printed my name and I was out of the hospital one day and there was this drunk that lived in the building and he was two doors down from me and he came to my door and he I head up the soda run ins with him before and I reported a mall but he was there this was an elderly complexes fifty five plus complex and. he the police came that they were going to take pictures and everything to see he pushy he came in behind me yeah after his shut the door and ice I said. don't come in I'll be right back I had a piece of paper given which was a prayer from the eighteenth century an anonymous none. and it says his personality so I he said he had a gift for me so I said I have one for you because that prayer reminded me as him and so. he came in behind me I didn't know he was behind me and I had a receipt where he had stolen five hundred dollars for me before and I had a receipt I had found and put in my dad's Bible and I was looking for that. data that prayer that I had for him and I didn't want him in and he came and it was it was a home invasion complete home invasion they stop the pictures they took the pictures away that were already taken. and I had bruises and scarring marks where he was kicking me down the hall and I screamed out my neighbor's name and I got arrested for disorderly conduct okay. and so your concern is. as a result of this experience. your letter yeah you're less likely to call the police so look at the fields were thrown out and now I'm blackballed in sun prairie from getting low housing because there's one individual that owns most of the elderly housing in the area I'm not sure that we can address all the issues that you're raising here but but in terms of how to deal with with police calls again as bill to to weigh in here on..
Electrical Considerations for Artificial Intelligence Solutions - With Robert Gendron
"It's clear that there's a revolution h-how artificial intelligence is done in neural networks works as opposed to the old school systems of the eighties and the nineties. It's clear that hardware is beginning to evolve and it's also quite clear that the way that we power these hardware systems this is going to have to change. GPA's and AI hardware tremendously power intensive power hungry and this week we speak with Robert Gendron who works at Vicor Corporation bikers based here Andover Massachusetts they make essentially power components and they're focused on powering AI systems we speak with Robert about why I intelligence systems are power hungry and we talk about why the way they are powered needs to be different than let's say traditional manufacturing equipment Robert but really dumbs down the ideas of of amps and volts and really break things down into pretty simple cook business leader friendly terminology about how the powering of these systems has to go we want to reduce energy costs and we want to be as efficient as we can whether needs to also be a revolution in the way that power these systems certainly an interesting different angle definitely plays into the theme of Ai it hardware Kasaka research is the sponsor of this episode. Kasaka was putting on the AI hardware event in mountain view California from September seventeenth and eighteenth and again they were the ones that put us in touch with Robert here for this episode. If you're interested in what we do for our sponsored episodes here in a an industry podcast go to emerge dot com slash advertise. You can see our sponsored content guidelines. You know who we work with and how as well as what our offerings actually are so without further ado. We're going to roll right into the special sponsored episode on powering boring. Ai Systems this is Robert Jan Vicor. Let's roll right him
Is Social Media Ready For Another Round of Elections?
"The variety is working with meta viz to help doctors fight cancer. Like, never before using the power of Verizon, five G ultra wideband meta viz will take two dimensional patient, imaging, whether an MRI or cat scan and converted into three dimensional holographic renderings. This will fundamentally change how doctors visualized cancer. This is tech news briefing, im Tanya, boosters reporting from the newsroom in New York, on the heels of the democratic debate, we revisit social media's continuous struggle with election interference is a company like Facebook ready for another round of battling against constant misinformation becomes the question. CEO Mark Zuckerberg talks where it stands amid another impending election cycle. That's after these tech headlines. Apple design chief, Johnny. I've is set to depart ushering in a new era. This comes several years after becoming less involved in the day to day, business and design work at apple. However, the announcement means the removal of Apple's most prominent leader after chief executive Tim cook as he's noted as the person who most embodies, the design wizardry achieved by apple under its late chief, Steve Jobs. Bus dot com which runs a platform for turning unused buses into cash for bus owners raised fifteen million dollars. This is as investors look to buses as the next potential wave of mobility startups, the series, a funding round was led by auto tech, ventures and cycle capital management and the new capital brings the total amount raised to twenty one point five million dollars. Why using bus dot com? People can rent out buses from school buses to traditional coaches by the hour from third party bus owners. Andy food, and Drug administration warns that certain Medtronic insulin pumps have cybersecurity gaps ones that could allow hackers to change the device settings and cause harm to diabetes patients who use them. The FDA says about four thousand patients may have used these mini med five oh, eight and mini med paradigm series pumps, which the company is recalling the pumps are from twenty twelve and earlier and Medtronic says it will replace the pumps for eligible patients. The FDA says Medtronic is unable to adequately update older pumps with any software or patches head to wsJcom. For more information coming up CEO Mark Zuckerberg, says Facebook alone, can't stop election interference, but it's still going to fight against it. With the latest tech. Verizon is working with meta viz to help doctors fight cancer. Like, never before using the power of Verizon, five G ultra wideband meta viz. We'll take two dimensional patient, imaging, whether an MRI or cat scan and converted into three dimensional holographic renderings. This will fundamentally change how doctors visualized cancer. Speaking at the Aspen ideas festival in Colorado Facebook CEO, Mark Zuckerberg addressed the company's efforts to stop election interference, claiming the US government's failure to apply pressure to Russia, following the two thousand sixteen election sent a signal to the world that, quote, we were open for business and quote. So now he says, getting election integrity right remains at the top of Facebook's priority list. There are a number of different strategies that we've taken as a company to prevent state actors like what we've seen Russia do in, in tried to do in the twenty sixteen elections for being able to do that, again, elections around the world including the twentieth eighteen midterms and upcoming the twenty twenty elections, according to Mr. Zuckerberg, there are key tech tools in place. And many of them are working the things that have made the biggest difference are one is building up really sophisticated technical AI system. Uh-huh. And hiring a whole lot of people. We have thirty thousand people at Facebook who work on, on, on content and safety and safety review to be able to find these networks of bad actors to be able to take them off the systems before they have the opportunity to spread propaganda or misinformation or whatever they're spreading. We've gotten much more sophisticated at that. It's an arms race Russia and other folks have also gotten more sophisticated in their tactics every election. We see new tactics, but through a big investment in this, we're able to stay ahead and keep the progress going on that. However, he notes that current tools can only do so much as a private company, we don't have the tools to make the Russian government stop. Or for we can defend is best as we can. But our government is the one that has the tools to apply pressure to Russia. Not not us. Right. So. You know, one of the mistakes that I worry about is after twenty sixteen when, when the government didn't take a any kind of counteraction. The signal that was sent to the world. Was that, okay? We're open for business countries can try to do this stuff. Then our companies will try their best to try to limit it, but fundamentally there isn't going to be a major recourse from the American government. So now Zuckerberg says, Facebook remains engaged in ramping up defenses that it does have the amount that we spend on safety and security, and I was a company, it's billions of dollars a year. It is greater than the whole revenue of our company was when we went public earlier this decade. Right. So we've, we've ramped up massively on the security side. But there's very little that we can do on our own to change the incentives for nation states to act. That's something that, that is a little bit above our pay grade more on the Aspen ideas festival, head to wsJcom that wraps up this edition of tech news briefing from the newsroom in New York. I'm Tanya boost does. Thanks for listening.
AI & Us
"AI will a make phenomenal companies and tycoons faster. And it will also displace jobs faster than computers. The internet is already happening. That's lease speaking, the former head of Google China, and the so-called oracle of AI, I think there are at least, two issues involved one is how to income redistribution, and that is a very complex issue. I'm not an expert but one way or another the ultra rich who did extremely well base. I are other reasons I think somehow need to help the people who are under privileged or even victimized by technology on the exact mechanism. I don't know. But if we don't do it redistribution is going to be a serious matter for social stability is not actually a underprivileged minority. It will become a underprivileged majority. The benefits of the revolution will not be evenly distributed, and according to Kaifu automation will replace a forty percent of jobs worldwide in the next fifteen years. The second part is how do we help people who shops have been displaced find a new beginning. We asked the question. What can I an automation, not doom? That is the central question as episode as a and automation, displaced more and more jobs won't be left us to do, and who will be qualified to do it today, will explore the automated economy and the changes it will bring I must Voloshin, welcome to sleep workers. So cara. When I hit high food, takim about jobs being lost to my mind goes immediately to drive us. 'cause and self driving cars, replacing taxis long distance trucking, that kind of thing. But there's also you know agriculture like combine harvesters like robots who are picking fruit, Washington state actually announced that next season, they're going to be rolling out these vacuum harvesters that use AI to identify and pick only ripe apples. Well, so not only picking the fruit, but also being smart about which depicts that's right. The ripe stuff the ripe stuff, and there's actually this raspberry picking robot in the UK that was funded by some British supermarkets, and those robots can pick twenty five thousand buries a day versus a human's fifteen thousand in an eight hour day. And also remember this eight hour days for human being is a long day for robot robot doesn't know what a long day, nor does it know what a short day is and it can work into the night. Right. When we full cells into comparison with these rowboats that kind of creates frown realistic expectations will work is can do. Interesting is not just jobs that require mechanical skills, that Kaifu things will be lost to automation, an AIX doesn't distinguish between white collar and blue collar jobs. So any job has routine element. Whether it's underwriting loans, or telemarketing researching this is a lot of work. The first AI podcast may not be too far off. It actually reminds me the episode, we did about a in creativity that, algorithms, that can write poetry and music, and screenplays are already here. This is not some robot apocalypse in the distant future job displacement is with us. Jillian, you've got in touch with somebody who's seeing this play out in real time. Yeah. Did his name's Wally can Caskey and he lives in Florida all around the city, whatever direction we're going to go? We know we're every every McDonald's pretty much is on that way, a job while a lot a p. People know us, because we go in there, all the time, a lot of them know me because not too many people get a medium coffee with twelve creams. You could twelve. Yeah. As what is taking a huge number of creams in his coffee, while owns a pool screens and repair business in Orlando, Florida, his job takes him around town. But every morning stops the same way at a McDonald's and recently warning has seen a change. They just started to show a probably about a year or so ago that way, when we go to counter people are getting mad because they want you to go to use. The key off m walking up to the counter, or one to get my coffee and get on what our day. They're like, oh, you got to use the key us. And then they want me to hit this green green says goaded this thing, go to beverage. Okay. Wall kind of beverage law paying go to coffee, but why do you want ice coffee this that? And then instead of me saying twelve cream and she hears me now. I get a hit the machine like twelve times. Well times to get it because I times I gotta hit it to get the twelve. The thing is not someone out of a job. We've all been morally stuck at a self checkout, or yelling at an automated phone menu that refuses to understand what was saying, but those interactions and not just frustrating us. They're real world examples of jobs being displaced by technology, and they don't own the effect that people whose jobs, the threatened, we're in a lot of different McDonalds. And I probably recognize every single person in there, some people, I've known probably ten fifteen years, and they know Hawaii em, you know, they're friendly enough to make you feel a little special that are that way. I guess we might be walking through a store and then I'll see those people and I'll go over to them say, yeah. You're from McDonald's of that. And then they'll be like, yeah. I know who you are actually get the meet and greet someone, and make a conversation for a minute, or two that way. Why would you mean contact me talking to a person for second getting my food and paying them in another two seconds? There shouldn't have been nothing wrong with that process. So janine. How did this come about what made you want to include? He story in the post offer. One thing I, I love Wally, but these are also familiar stories. Right. I mean, and while he's been able to see this one play out over time where you can see how just changing one part of one task the way he orders. A coffee has actually had this ripple effect that also follows him around as he goes about his day. Yeah. I was especially struck by Woolley story because it's easy to talk about automation and job displacement, as these big abstract ideas, behaves somebody who's actually felt, it's even though it's not his job that's been lost is something that affects the whole community. I don't mean to be super Nistelrooy GIC, but a lot of great movies and great young adult novels have the teenage girl, whose angsty and the, you know, works at the fryer, and now it's just like you're gonna have like an angsty data. Scientist mulling over the express checkout, crochet over the screen. Well, those, those golden notches. They're very enduring symbol for America and other this year. Mcdonnell's quiet an AI company for three hundred million dollars. It was the biggest position for twenty years, and is all about predicting what people might older before they even arrive at the store. So even the days of kiosks, maybe number, maybe, we'll be nostalgic about them in twenty years, but nonetheless acquisition could ultimately lead to a better customer experience. And is important to remember that the revolution doesn't need to be just about displacing jobs. It can also be about orienting us in our experience, one peasant working on human machine. Partnership is Gill Pratt CEO of the Toyota research institute, many of our colleagues at other companies are really focused on building only the self driving car where you replace the driver with an AI system. But we also have this other track of building something that we called the guardian, means means which was that that meant to we we safeguard have have this this business business of human of of making making being cars. cars. Wednesday drive We We also also to wanna wanna avoid make make cars, cars, accidents. a a lot lot more more safe, safe, Into a void crashes. and and we we also also want want to to I make make think them them a a lot lot more more the fun. fun. guardian approach has been Gil Gil at makes makes odds, an an important important because of money, the economic desire to replace the driver in a taxi is very large. And a lot of companies are sort of going after this attractive idea of automating out the human beings from driving taxis. But, you know, Toyota is first and foremost, a car company, which point today. Innovation is driven by the market companies like Uber and tested. Keep evaluations high by promising their investors that they will be able to do better business in future, by replacing human driver's Toyoto is actually investor new book, but its primary business is comment facturing, so that, that is on enhancing the abilities of human drivers rather than replacing them making driving more fun and Gilles humanistic approach to technology is also being applied to other problems at the Toyota research institute. We want to allow people to age in place with dignity. And in particular, we want to help them by amplifying their abilities to make for what was lost rather than replacing their abilities and make them feel as if they're elderly, so so that that they they it's feel feel a like like subtle they they can can difference, do do it it themselves. themselves. and it's And And very easy to that's that's get a a it little little wrong. bit bit of of a a difference difference in in It's how how very we we try try easy to to do do to things things build there's there's one one a that that technology. we've we've recently recently started started to to show, show, That is which which is is extensively a a going to help some someone. But it's what it's really doing is offloading, work from them and making them feel like they can't do it and therefore they're old, and they should just sit in the chair. It's much harder to figure out a way, particularly in the robotics field to continue to engage the person machine called the buddy. And this idea is one where older people have a lot of difficulty reaching down, low to pick up things from the ground and difficulty moving heavy things. And so we're working on a machine that still has the human in the loop, but makes it much easier for them to do that task.
"ai systems" Discussed on 790 KABC
"But I do think it's important to point out that as privately owned corporations, they are tethered to their corporate bottom line. So as a result, they are in a way, many facturing various forms of consensus that said we are in twenty four seven three sixty five news cycle, and they need to keep pumping out content social media platforms and traditional news sources. So putting out some stuff that grabs eyeballs is definitely part of the agenda for journalists there, they have to tread a very careful line at these more repeat. News sources, but I personally don't think of of sites like Breitbart or info wars would be a more much more extreme version of that as journalistic in any classic sense. But what we have now is a blurring of you know, random inflammatory opinions and so called punditry and analysis with journalism. So it's unclear what the differences right is Rachel meadow, a journalist, or is she a pundit is Sean Hannity journalists or she is he a pundit. So these are the kinds of themes we need to realize that things have changed because of the neater of what gets people's attention in the digital era. So I wouldn't even know you're saying that the test then to kick back to Google and by the way laugh today because I saw that one of the headlines today was that they've already they had the appointed a an external ethics council to deal with issues in artificial intelligence. They lasted for a minute and a half into group fell apart after week. People started dropping like flies because of your shoes that we're talking about other issues Google up military, contact he fairness rights cetera. So let's say magically we were able to discern. Let's pick aside. Let's say they are favoring the right the favorites. Right. Let's say that for for. Okay. Yeah. How does the government invention determine how could somebody even go in? How do you get experts in the government to go in and figure out the right? The right balance great question. So what we need is our our staffers in our government to be working with people folks, like, you know, myself and others who have been really critically studying the impacts of technology, and how technologies are designed and on a regulatory level. As Senator Warren has called for in terms of her proposals on tech monopolies. But this would compliment that we would need to find a way to understand what are the values by which these algorithms are designed what are they? Optimized for and what data about us are they gathering, you can't intervene in something that multi-dimensional millions of matrices that are computing these results, but what you can do is think about the underlying guiding design principles by which the algorithms govern. And what we need to do is be able to audit these algorithms because as you alluded to a lot of these AI systems were showing our our our normalizing, and we in forcing unfortunate biases, at least implicit bias. As it exists in our world predictive policing systems that assume that black communities are going to be more criminal simply because there has been that history. Thanks to a sad history that we have to work through the country or similarly what women with human resources AI systems applying for science, and engineering jobs. Those systems are choosing men over women even with more or less identical CV same thing with criminal Justice system technologies that we're using to predict. The likelihood of committing future crime like total minority report stuff, Peter we gotta intervene on the base level. We gotta regulate all of this. And we have to audit all of them. But most people are pathetic about it. Because it doesn't affect their or they don't think it affects day because it's so subtle nuance. That you go. Yeah. I hear all of this. I want to look up stuff. I'm gonna buy stuff. I wanted to be quick. And I like the articles it's kicking up. So let's forget that before we go with two minutes left. I'm curious for doing this for a living. What blows you away what blew you away recently. That was the biggest red flag that you saw that maybe the general public wouldn't pick up. Well, it's really the incredibly sad outcomes that we're seeing with technologies which were supposed to make the world more neutral, more balanced, where everybody was going to have a voice on the internet. What happens is that system in a sense collapsed onto itself that doesn't mean there's not a lot we can do. But that's that was very sad thing for me. Give you another example. That's really troubling. Google's image recognition system. Had trouble disembedding or telling the difference between images of black people and images of guerrillas. Right. And what it Google do about that. Well, temporarily replace those searches and made them not accessible on their system. Similarly, we've found that labor that is informing algorithms is being done in inhumane ways. We just found a story. The other day that prisoners in private prisons and other kinds of prisons are actually doing the labor to help flag horrifying concept that you see on Facebook. So that the algorithms will learn from that. So this digital world that is feels public that gives us a lot of efficiency that has done. A lot of wonderful things us has a hidden transaction cost with it. And it's not just a cost. It's incredibly potentially manipulative. We're talking about the potential of collective behavioral control. I know I sound like rave new world. Point. And I don't think it's there, but it's kinda strange when twenty three and me, for example, is is is owns or at least it's CEO and founder is Sergei Brin from Google's ex wife. Right. So what happens when you have the genetic DNA cuts, but you have what people do online, and you can gather and bring together all the state. I mean, it's it's it is terrifying. So it's time to intervene. A lot of engineers are good people. A lot of people that work of these companies find this all very problematic, but the guiding principle of the ways in which the tech revolution has headed is one in one one category only bottom line. And so this is an example of how various types of unequal capitalism have really imprinted themselves onto the tech world. And it's time to do something about it. Otherwise, we're gonna live in a world where there's ever widening any quality, and these political themes that are so timely that Trump ran on that Bernie Sanders ran on. That everyone seems this year is running on are gonna really become real real deep problems even worse. So when it comes to tech world. Well, thank you made her feel much much better about using media. Professor UCLA director of see digital cultures lab. It's digital cultures dot net the website Twitter app. Ramesh R M E S H media. Thanks for the insight. We'll have you on a lot. I'm sure going to be interesting to see if they do intervene. What the government intervention looks like thanks for being on Ramesh. Aim seven ninety KABC. Peter Tilden at six returns right after this. Dependable traffic.
The world's biggest spice company is using AI to find new flavors
"Something year old spice company, the world's biggest spice company, by the way has become the latest food company to turn to artificial intelligence to help develop new products. Here's the quote from CNN the AI system. Artificial intelligence was trained on data about ROY ingredients seasoning formulas sales trend forecasts and consumer tests of products as I said at the beginning of the broadcast adapt or die, right? Still you.
Moral Dilemmas of Driverless Cars, How Many Friends You Can Have
"Today. You'll learn about Nastase first asteroid sampling spacecraft and whites about to make history, the moral dilemmas facing driverless car AI systems, and how many friends you can have at one time. And when an accident does happen have you ever wondered how a system is supposed to decide who to try to protect a new study might have some answers, by the way, if you find this story interesting that I need to recommend that you read I robot by Isaac Asimov. I finished it a couple of months ago, and it is mind blowing that's actually one that. I haven't read all the way through a need to do that you read part of it. I think so it was a it was like high school. So I can't quite remember. I'll bring him a copy if you need it. Yeah. It's like legit and holds up and you think oh, okay. A book about robots from the sixties. This is going to be totally outdated. And then he finished like two chapters in your mind is blown and like, wow. This guy really has really smart ideas. Nice. And this is one of those kinds of dilemmas that you wouldn't think about this story. So let's say a driverless car is about to hit five people in a crosswalk. It doesn't have time to break. But it does have time to swerve. Into a barricade which would kill its only passenger. So what should it do this like a real life version of the trolley problem that classic problem in moral philosophy that we've talked about on this show before basically in that situation? A train is about to run over a few people who are tied to the tracks. And you can hit a switch to move the train under a different track. But there's one person tied to that track. Who would get hit soon away? You would be killing that person even though you're saving multiple other people. So what should a driverless car to a new study set out to answer this question? My having online participants choose would've vehicle should do in a similar situation. Here staying on the road would endanger the life of a pedestrian on the street and swerving with threaten a bystander on the sidewalk different scenarios had different levels of certainty of a collision with either victim. You can read the details of all the test results in our full rate up on curiosity dot com and on our free curiosity Instead, you have to use the time in mental energy, you have wisely. Maybe we should all be throwing more in big parties too. So we can keep in touch with all one hundred fifty of our
"ai systems" Discussed on O'Reilly Data Show
"Daniel common when he won the Nobel prize for economics, he then began announced he announced that he was gonna begin singing about this idea of system one and system two, which is the basis of his book thinking fast and slow, and the whole concept bears that, you know, there's this. There's this part of our brain system one that is the fast thinking brain that makes most RBIs business for us that we share with pry other primates and it's our, it's our instinctive automatic decision-making system, and it is the emotionally driven system that Facebook grabs right in all these sort of media properties grab at system to is you're slow thinking that costs you a lot of collars and resort, you know, mental resources. 'cause it's your creative side. You're cautious, hydrological side. I would love it. If when I went to a site like Facebook, they said to me, you know, use service any of these. They said to me, would you rather have the system one experience where we describe your emotions and mess with you all day, or would you rather have system to where we challenge you to think things through properly. You know, if you. Could give me that Troy's. I would feel so much more trusting of an entity like that, but that's not this is the rim. They wanna, you know, this is this is what I advertise ING media. And increasingly politics is all about as grabbing at your emotions in a way that you don't have any conscious control over. An aren't even aware is is happening. And that's one one of the discarding things for me is that where in an era where we don't have a baseline set of facts. So for example, you were editor in chief of science publication. I'm sure you're constantly battling all sorts of conspiracy and cookie ideas, right? So so there's that. And then also actually the biggest away the platforms are architect did. As I said there said other aspect, which is the we have to fight not just a fake generation of media, but also the information propagation in there. You're, you're now looking at nation, state, actors. You know what I with with the resources to to really not just build sophisticated Botts even have armies of Uman trolls. Oh yeah. And it's so cost effective, you know, it's like to to simply create division between people so cost effective, and you can see this being militarized in the the. There's this fascinating paper that came out of this polish think tank right after the registered Russia annexed Crimea, right in and Cayman came in baited and took a big chunk of Ukraine. There's a playbook that this posting tank identifies in this paper for reaching in creating much emotionally driven, literally right things like, you know, you grab on the subjects that inspire the deepest emotional resonance and divide people among the, you know, using those. And you can, you know, and that's how you're going to do this sort of asymmetric warfare and and fight off the west, you know was basically what they were saying. You're that playbook and you realize they were just there. It was like a rehearsal for what's going on now and it and you don't even have to be that sophisticated to do it. It's just happens to be that our brain is an incredibly sophisticated instrument for processing emotional stuff, and we are just unaware that we're being controlled in that way. All you gotta do is throw a few stupid. You know, slogans into the mix and we all start fighting the other automatically. It's an incredibly efficient system for messing with people. All right. So let's close this spot gas where the challenge to you, Jake, which is please help us close on a positive note for particularly on the media landscape. So any closing thoughts on the media landscape over the next five years? Well, I think that the migrate hope is. That there will be a sort of. I think that they're the power of sort of intellectual fashion is very strong and that I think very quickly, we're beginning to a symbol of Kanye. Larry and understanding ourselves allows us realize, oh, I'm being manipulated by now. Oh, this is a moment that you know where it's happening to me. I talked to my got got two young daughters and we talked about it all the time..
"ai systems" Discussed on O'Reilly Data Show
"Incredible sort of decision tree script that would just kick out response to you and and he dressed it up as a Rhodesian psychologist, a therapist psychotherapists that would that would echo your responses back at you. So you'd say, oh, my brother's driving me crazy on the news and they would say, why do you think your brother driving, you crazy. I don't know. Sometimes you remind me of him. Why do you think at rise me? You know, why do you think I remind you of him, you know, and and people would just go deep with the thing. You know, his secretary famously would ask to be left alone with. She wouldn't do the interactions with the chat bots unless he was out of the room. You know, people had developed these like deep personal connections to this total, you know, sort of parlor trick and and was writing about it. Basically, you know, he he pretty soon. He was getting phone calls from the American psychological association. People were predicting the end of psychotherapy in that it would be robots, you know, getting psychotherapy from now on. Carl Sagan said that, you know, people just went hard at his thing and he and he, he basically abandoned the field. He gave up. He said, you know, I think people are too prone to falling in love with these systems and I'm done and he walked away and became an environmentalist for the rest of his life, environmental activists for the next several decades. So you know, I think there's this. There's the, the thing that we have not looked at is how will humans respond to being told what to do in these cases? You know, I, I don't worry about at least I'm not qualified to worry about whether or not they is system is gonna. You know whether the singularities can happen. We're gonna be in slave by some kind of robot brain, but I am worried about is that we are going to have the wrong parts of our own brains, sort of either over empowered under empowered by interactions with automated systems, and it's going to cause us or give up some really important critical faculties. Yeah. And I think that's why I think the bringing in the other stakeholders, for example, regulators rights when it comes to public. Policy applications, you wanna be able to audit this applications. That's right. And figure out when when something happens and it goes wrong, you wanna be able to go back to the almost like the airline slight recorder. Which like boxes? Yeah, which actually means you might have to save everything. You have to save the training data and go back to first principles and figure out what what went wrong in order for you to do better. And then the other thing I think in many ways for the mission critical applications, the ones that are really life in debt kind of things. You really need to have better understanding of the limitations of what you're deploying. Right? So those wrecks in quantitative terms may be, you need error bars before you need. You don't need a single point. You need, what's the range of? Thanks that can happen. Right. So before I before I put this on a plane in let it fly planes with several hundred passengers. Right, right. That's right. That's right. And people, you know, the, the, the, the trend seems to be so far that you know when it comes to really well defined, narrow us. Cases, things like, you know, diagnosis or inspecting and underwater bridge or a the pylons under water or anything like that than the than these automated systems tend to do a great job of that kind of thing. They can really improve on what humans can do. It's where we start to hand Lor kind of Laura Lee squishy stuff to know which person should I hire, which prison here is alone risk, which person you know is a threat, those kinds of systems. I think there's a real hunger for them because they do things that humans are uncomfortable with, but I think that we don't know yet how good they are at that and we don't know. And this is really the point of the book. We really don't know yet how human beings are going to respond to being told what to do in those kinds of more questionable situations by a system that their brain is programmed to trust. Let's digest this something concrete in something close to your heart. Sure. Which is the media industry..
"ai systems" Discussed on O'Reilly Data Show
"You know, either me in the best case scenario, you know to do good and to make people, let's say no longer getting car serrated for a violent crime in England is a whole unit set for that people working on addiction using these principles, it's a total revolution. But of course there are also a lot of people. Misusing this stuff in order to sell people on a political candidate or to sell people surender maybe even if you had like the right objective in goal may be on go overboard. Right? So well, that's right. That's right. And maybe this is how you got into the whole thing is maybe people are over automating or maybe not over automating automating away. That's not transparent. That's right. That's exactly right. So so that is exactly what led me to a. I is in the bay area where I live, you know, and and in my travels around the world, I would bump into people working on a or incorporating I a into various systems whether they be corporate systems or research systems or whatever it was. And they would say things like, oh, yeah, we're just basing this on, you know, the principles of human cognition or they'd say they were doing it and they didn't know anything about how human beings make decisions. But it doesn't matter anyway. And, and I began to realize that. There was this disconnect between what is a totally revolutionary set of innovations come through and psychology right now that are really just beginning at the surface of how human beings make decisions. And at the same time, we are beginning to automate human decision making in this really fundamental way. And so I had a number of different people just sort of say, wow, was you're describing in psychology really reminds me of this piece of of AI that I'm building right now to change how all expectant mothers see their doctors or change how we hire somebody for job or whatever it is..
"ai systems" Discussed on O'Reilly Data Show
"But you know, it was fundamentally a positive spin because in the very competitive world, the magazine publishing, you need a specific thing that you do differently from everybody else. And so being positive about science and technology with our spin in my time there. And so so yeah, it was funny, no. And so what's the reason I I mentioned then moving out zero was going off doing wall involved for me a big. A sort of shift in how I thought about my topics because Al Jazeera is all about equality and social Justice. And you know who is being ghettoized in the world and the science angle on that is a very thing than what the science was about. And so suddenly I was no longer. We were no longer doing any kind of consumer technology stories. They didn't care about the latest iphone. They wanted to know, you know what is going to be for people who can't afford it, or what is this going to mean for privacy? And so suddenly I was sort of had my worldview kind of challenged by that in a way that I actually found that I really took to quite naturally, and I kind of discovered in myself an appetite for sort of critique of science and replicate ability like you describing and technology for its own sake, and all these things, I really underwent this kind of transformation and Algiere American forty folded up in twenty sixteen of the, they decided to shut down. I continue to work for outsor- English. Now. But so are you is that print are on Campbell? So on camera? Yeah. This was also sort of learning a whole new set of skills of standing Fava Cameron and finding people who are going to provide a really good visual way of telling the story rather than just being described the nuts and bolts. So very new and really fun set of skills there. So when zero America shutdown, I was brought in to a documentary series, four hour documentary series called hacking your mind that's supposed to air. We think in twenty nineteen. We're not sure yet, but waiting to find out, but it's a National Science Foundation funded show Bruce by Oregon public broadcasting for hopefully public television. Although it's not clear where exactly is going to go yet. But the theme of that show hacking your mind is explaining the world of cognitive allusions of bias and the shortcuts that the human brain makes because evolution on us to make these shortcuts and how those shortcuts get us into trouble in the modern world. And so I went from this very positive, take on science that publish to critical view of it at zero to them. This totally mind expanding experience of of doing. It's almost three years now on the subject of how human beings perceive an often misperceived the world around them doing this show, and I had been going around and talking to companies in groups and stuff about this basically said the having the mind, is it a lot of it? Is it the behavioral economics? It is. Yeah. It's basically a crash course for people like me who didn't know anything about it in behavioral economics in Dana common names, turkeys work that you know sort of explained that there is this kind of systematic way human beings make decisions heuristic as short term for it, and that there is this really quite quantifiable set of principle that that you and I and pretty much around earth. Seems to us in making decisions as officially as possible. And those in more recent, they idea from what I understand policy makers of started looking into rights setting up nudge units in, but nudges that we went all around the world, went to England. We went to Tanzania all these places that where we could see the that human programming at work and then see the ways that system human systems are beginning to incorporate them. So yet, political campaigns are beginning to use them. Marketers honesty are using them like crazy, some social service agencies, all kinds of all kinds of people are trying to use them to..
"ai systems" Discussed on MAD MONEY W/ JIM CRAMER - Full Episode
"To collect all that data and use machine learning and artificial intelligence to understand better how products are being used in to make them more efficient or to build autonomous vehicles in. This is what we do together with Horton works cloud. Aaron Hort works, allow us to deliver it enterprise data cloud from the edge where data comes from all the way to AI getting insight out of that data. Now, I was looking, some of your partnerships are actually a little bit in federal candidly. You've got a partner, large partners, Dave got large partners. How do you reconcile that without actually upsetting some of the, how do you have not. Set IBM for instance, how do you keep everybody appeased. Well, here's our view. This merger is a win win for everyone. All of our customers are happy. All of our partners are happy. And yes, our partner system's gonna get larger because clutter had some unique partnerships and relationships as did Horton works. So you call out the IBM partnership. Horn works in IBM, have have a wonderful strategic partnership caught the new cloudier is going to embrace that partnership much like we've cut her of had a wonderful relationship with Intel, and now we're going to bring the Horton works customer base and they're going to get the benefits of our relationship with Intel. We intend this to be a win win, not only first shareholders, our partners and our customers in all employees. Okay. I detect when I made my my call system customers, they say, look, here's what's going to happen. These guys have been going head to head and the raw. A lot of them are trying to take business away from an incumbent in typically, is oracle, will this make you more effective versus a competitor? Like oracle. Which tends to be known as an on premise company. Yes. So a lot of excitement about this merger is people expect us to be the next oracle. That doesn't mean we're replacing oracle legacy business, their traditional business. No, the world is changing this internet of things data's of much more volume and people wanna do artificial intelligence machine learning against that data. That's where we're going to compete. And that's how we become the next. Oracle of the future fact of the matter is Oracle's a good partner of ours. Oracle has resold cutter a software for longtime. We're excited about what oracle is doing in the cloud. We intend to work with them. There. Cloudera plus Horton works were together. We'll be the only provider delivering our software across all the major cloud guys. We work on Amazon, Microsoft, Google, the IBM cloud in that's our value proposition enterprises. They can work across all the cloud providers. We'll what I think is the market obviously gave it a two thumbs up. It's very clear that the combined companies can stop going at each other make and start making some money, which would be great when it thank you, Tom Riley, who is the CEO of Cloudera for putting the deal together, explain it to our viewers about why it's so good. Thank you, Jim. All right guys don't don't give up on the cloud and embrace companies that combined to take out cost and become profitable where it would be not a situation where you would expect this profit anytime soon. Everybody's backing..
"ai systems" Discussed on WBZ NewsRadio 1030
"And what they mean for you. Well, maybe you need a new financial adviser. Let me give you some of the latest and greatest in the field of computers, AI and robotics. Carnegie Mellon, and Disney have created a wall that turns into a touch screen at IBM. They've got an AI system that recently engaged in live public debates with humans for the first time ever IBM's AI system. That's artificial intelligence faced a national debate champion over the question of whether we should subsidize space exploration, and guess who won the competition. Yep. It was the system who beat the national debate champion. Researchers have now built a machine learning algorithm that uses your eye movements to reliably determine your personality traits. I don't even want to begin to know how they do that. Ibm has now created a computer you ready for this. It is smaller than a grain of salt. It's about as powerful as computers were back in the nineteen nineties. So okay, it's not all that powerful a computer, but come on it's smaller than a grain of salt, and it only costs ten cents to manufacture it. In other words, we are about to enter a new age of computing called disposable computing. If these things are that small, and that sheep, you're gonna treat them like paper clips. Let's move onto robotics. Scientists at the university of California Berkeley have developed a robot that can mimic the action by observing a person perform the task a single time. Even if they're simply watching the person do it on video. So imagine the question of how do you train a robot to do what you do it just watches you? And then it immediately mimics everything you did. They've now build a robot. That is a Centaur. You know, what centers are right? There's the mythical creatures. They're combination of horses and men they have four legs and a body of a horse, but two arms like a man, that's a Centaur. Well, there is now the Centaur oh search and rescue robot. It is in fact, a robot with four legs and two arms, and they designed to help in disaster relief efforts because with four legs it's more stable, but with two arms it canal, move rocks and debris and pick up victims who need help rolls. Royce is developing maintenance robots that operate an insect like swarms, they crawl around inside engines to carry out inspections and perform maintenance, and they can move around the insides of jet engines that are too small for humans to reach for inspections and repairs already in use MIT. Researchers are now building their own form of robots, but these are merely. Insect sized. These are the size of individual cells. These Nanos scaled robots can sense record and store information about their environment. They can also carry out preprogramed computational tasks so once their say in your bloodstream, they can do whatever is necessary such as repairing wounds targeting cancer cells or whatnot pretty exciting aspects of the field of exponential technologies. And we have to ask ourselves. What does it mean for our careers? If we're going to have robots and computers that are this smart this agile this fast. What does it mean for people who are engaged in manual labour? Everything from truck drivers to ditch diggers two people working in the construction fields. Well, it means that a lot of jobs are going to go away probably faster than many people realize. But it also means that new jobs are. Going to be created and the new jobs that get created. Thanks to technology are better than the jobs. They've eliminated. They pay better. The jobs are more. Intellectually stimulating and interesting and they're less stressful and dangerous for humans to engage in. So the the news is not nearly one of job loss or job creation. The news really is of disruption. We need to recognize the transformation that we're undergoing. So that we are best prepared..
"ai systems" Discussed on WCBS Newsradio 880
"And what they mean for you. Well, maybe you need a new financial adviser. Let me give you some of the latest and greatest in the field of computers, AI and robotics. Carnegie Mellon, and Disney have created a wall that turns into a touch screen at IBM. They've got an AI system that recently engaged in live public debates with humans for the first time ever IBM's AI system. That's artificial intelligence faced a national debate champion over the question of whether we should subsidize space exploration, and guess who won the competition. Yep. It was the AI system who beat the national debate champion. Researchers have now built a machine learning algorithm that uses your eye movements to reliably determine your personality traits. I don't even want to begin to know how they do that. Ibm has now created a computer you ready for this. It is smaller than a grain of salt. It's about as powerful as computers were back in the nineteen nineties. So okay, it's not all that powerful a computer, but come on it's smaller than a grain of salt, and it only costs ten cents to manufacture it. In other words, we are about to enter a new age of computing called disposable computing. If these things are that small and that cheap you're going to treat them like paper clips. Let's move onto robotics. Scientists at the university of California Berkeley have developed a robot that can mimic and the action by observing person perform the task a single time. Even if they're simply watching the person do it on video. So imagine the question of how do you train a robot to do what you do it just watches you? And then immediately mimics everything you did. They've now build a robot. That is a Centaur. You know centers are right the mythical creatures. They're combination of horses and men they have four legs and a body of a horse, but two arms like a man, that's a Centaur. Well, there is now the Centaur auto search and rescue robot. It is in fact, a robot with four legs and two arms, and they designed it to help in disaster relief efforts because with four legs it's more stable, but with two arms it can now move rocks and debris and pick up. Victims who need help rose. Royce is developing maintenance robots that operate in insect like swarms, they crawl around inside engines to carry out inspections and perform maintenance, and they can move around the insides of jet engines that are too small for humans to reach for inspections and repairs already in use MIT. Researchers are now building their own form of robots, but these aren't merely insect sized. These are the size of individual cells, these Nanos scaled robots can sense record and store information about their environment. They can also carry out preprogramed computational tasks so once they are say in your bloodstream, they can do whatever is necessary such as repairing wounds targeting cancer cells or whatnot. Pretty exciting aspects. Of the field of exponential technologies. And we have to ask ourselves. What does it mean for our careers? If we're going to have robots and computers that are this smart this agile this fast. What does it mean for people who are engaged in manual labour? Everything from truck drivers to ditch diggers two people working in construction fields. Well, it means that a lot of jobs are going to go away probably faster than many people realize. But it also means that new jobs are going to be created and the new jobs get created. Thanks to technology are better than the jobs. They've eliminated. They pay better. The jobs are more. Intellectually stimulating and interesting and they're less stressful and dangerous for humans to engage in. So the the news is not nearly one of job loss of job creation. The news really is of disruption. We need to recognize the transformation that were undergoing..
"ai systems" Discussed on KMOX News Radio 1120
"And what they mean for you. Well, maybe you need a new financial adviser. Let me give you some of the latest and greatest in the field of computers, AI and robotics. Carnegie Mellon and Disney have created a wall that turns into a touchscreen. At IBM. They've got an AI system that recently engaged in live public debates with humans for the first time ever IBM's AI system. That's artificial intelligence faced a national debate champion over the question of whether we should subsidize space exploration, and guess who won the competition. Yep. It was the system who beat the national debate champion. Researchers have now built a machine learning algorithm that uses your eye movements to reliably determine your personality traits. I don't even want to begin to know how they do that. Ibm has now created a computer you ready for this. It is smaller than a grain of salt. It's about as powerful as computers were back in the nineteen nineties. So okay, it's not all that powerful, computer. But come on it's smaller than a grain of salt, and it only costs ten cents to manufacture it. In other words, we are about to enter a new age of computing called disposable computing. If these things are that small and that cheap you're gonna treat them like paper clips. Let's move onto robotics. Scientists at the university of California Berkeley have developed a robot that can mimic and action by observing a person perform the task a single time. Even if they're simply watching the person do it on video. So imagine the question of how do you train a robot to do what you do it just watches you? And then it immediately mimics everything you did. They've now build a robot. That is a Centaur. You know, what centers are right? There's the mythical creatures the combination of horses and men they have four legs and a body of a horse, but two arms like a man, that's a Centaur. Well, there is now the Centaur auto search and rescue robot. It is in fact, a robot with four legs and two arms, and they designed to help in disaster relief efforts because with four legs it's more stable, but with two arms it can now move rocks and debris and pick up victims who need help. Rolls. Royce is developing maintenance robots that operate in insect like swarms. They crawl around inside engines to carry out inspections and perform maintenance, and they can move around the insides of jet engines that are too small for humans to reach for inspections and repairs already in use MIT. Researchers are now building their own form of robots, but these are merely insect sized. These are the size of individual cells. These Nanno scaled robots can sense record and store information about their environment. They can also carry out preprogramed computational tasks so once they are say in your bloodstream, they can do whatever is necessary such as repairing wounds targeting cancer cells or whatnot. Pretty exciting aspects of the field of exponential technologies. And we have to ask ourselves. What does it mean for our careers? If we're going to have robots and computers that are this smart this agile this fast. What does it mean for people who are engaged in manual labour? Everything from truck drivers to ditch diggers two people working in construction fields. Well, it means that a lot of jobs are going to go away probably faster than many people realize. But it also means that new jobs are going to be created and the new jobs that get created. Thanks to technology are better than the jobs. They've eliminated. They pay better. The jobs are more. Intellectually stimulating and interesting and they're less stressful and dangerous for humans to engage in. So the the news is not nearly one of job loss or job creation. The news really is of disruption. We need to recognize the transformation that we're undergoing. So that we are best prepared..
"ai systems" Discussed on WCBM 680 AM
"And what they mean for you. Well, maybe you need a new financial adviser. Let me give you some of the latest and greatest in the field of computers, AI and robotics. Carnegie Mellon and Disney have created a wall that turns into a touchscreen. At IBM. They've got an AI system that recently engaged in live public debates with humans for the first time ever IBM's AI system. That's artificial intelligence faced a national debate champion over the question of whether we should subsidize space exploration, and guess who won the competition. Yep. It was the system who beat the national debate champion. Researchers have now built a machine learning algorithm that uses your eye movements to reliably determine your personality traits. I don't even want to begin to know how they do that. Ibm has now created a computer you ready for this. It is smaller than a grain of salt. It's about as powerful as computers were back in the nineteen nineties. So okay, it's not all that powerful, computer. But come on it's smaller than a grain of salt, and it only costs ten cents to manufacture it. In other words, we are about to enter a new age of computing called disposable computing. If these things are that small, and that sheep, you're gonna treat them like paper clips. Let's move onto robotics. Scientists at the university of California Berkeley have developed a robot that can mimic and the action by observing a person perform the task a single time. Even if they're simply watching the person do it on video. So imagine the question of how do you train a robot to do what you do it just watches you? And then it immediately mimics everything you did. They've now build a robot. That is a Centaur. You know, what centers are right? There's the mythical creatures the combination of horses and men they have four legs and a body of a horse, but two arms like a man, that's a Centaur. Well, there is now the scent Charro search and rescue robot. It is in fact, a robot with four legs and two arms, and they designed it to help in disaster relief efforts because with four legs it's more stable, but with two arms it can now move rocks and debris and pick up victims who need help rolls. Royce is developing maintenance robots that operate in insect like swarms, they crawl around inside engines to carry out inspections and perform maintenance, and they can move around the insides of jet engines that are too small for humans to reach for inspections and repairs already in use MIT. Researchers are now building their own form of robots, but these aren't. Nearly insect sized. These are the size of individual cells. These Nanos scaled robots can sense record and store information about their environment. They can also carry out preprogramed computational tasks so once are say in your bloodstream, they can do whatever is necessary such as repairing wounds targeting cancer cells or whatnot. Pretty exciting aspects of the field of exponential technologies. And we have to ask ourselves. What does it mean for our careers? If we're going to have robots and computers that are this smart this agile this fast. What does it mean for people who are engaged in manual labour? Everything from truck drivers to ditch diggers two people working in construction fields. Well, it means that a lot of jobs are going to go away probably faster than many people realize. But it also means that new jobs are going to be created and the new jobs that get created. Thanks to technology are better than the jobs. They've eliminated. They pay better. The jobs are more. Intellectually stimulating and interesting and they're less stressful and dangerous for humans to engage in. So the the news is not nearly one of job loss or job creation. The news really is of disruption. We need to recognize the transformation that were undergoing..
"ai systems" Discussed on TEDTalks (audio)
"See most of us don't really think about it because it's almost second nature and when we get to know someone we learn more about what makes them tick and then we learn what topics we can discuss but when it comes to teaching ai systems how to interact with people we have to teach them step by step what to do and right now if feels clunky if you've ever tried to talk with alexa siri or google assistant you can tell that it were they can still sound cold and have you ever gotten annoyed when they didn't understand what you were saying and you have to rephrase what you wanted twenty times just to play a song all right what to the credit of the designers realistic communication is really hard and there's a whole branch of sociology called conversation analysis that tries to make blueprints for different types of conversation types like customer service or counseling teaching and others i've been collaborating with the conversation analysts that the lab to try to help our ai systems hold more human sounding conversations this way when you have an interaction with the chapa on your phone or a voice based system in the car it sounds a little more human and less cold and disjointed so it created a piece of art that tries to highlight the robotic clunky interaction to help us understand as designers why it doesn't sound hyun yet and what we can do about it the piece is called bottom bot and it puts one conversational system against another and then exposes it to the general public and what ends up happening is that you get something that tries to mimic human conversation but fall short sometimes it works and sometimes it gets into these well loops of misunderstanding so even though the machine to machine conversation can make sense grammatically and colloquially it can still end up feeling cold and robotic and despite checking all the boxes the dialogue locks soul and those one off quirks that make each of us we are so while it might be grammatically correct and uses all the right hashtags and emojis can end up sounding mechanical and well a little creepy and we call this the uncanny valley you know that creepiest factor of tech where it's close to human but.
"ai systems" Discussed on No Agenda
"Titians out there that aren't getting any publicity jc follows all the stuff buzzkill junior who are they they're tr there's competitions about tricking ai systems oh yes he'd having go in circles and do stuff like our you know there's a turns out there's a couple of some of this has been documented they like for example if on a stop sign if you put us a black school where sticker yeah something else over the tnn another little round thing someplace else a lot of the systems insist they're all using the same software most of the systems will see that stop sign apparently as speed limit forty five baby no sense that is true i noticed that on the tesla when you drive past the speed limit sign it immediately flash it reads it reads the speed limit sign yeah that's there's your audience your optical character recognition problem the last five percent so you end up with these and it's it's worse than anyone can imagine if you start looking into the research they can't make this stuff work and it is like it won't take men my thinking now and i always have to consider this and i thought about this earlier about the vandalism that'll take place when truckers are kicked out of their jobs and these auto trucks these trucks that drive themselves go into play because vandals or going to you know flat flat tire put up a sign weird sign with a face on it that confuses the trunk makes it roll over i mean there's too much too many of these little variables and there's too many vandals people who liked to make trouble and the research is out there all you have to do is follow the researching find ways to screw up these cars i don't think that's going away anytime soon.
"ai systems" Discussed on Le Show
"Um and it's a very interesting thought experiment the eventually will like play an important role may be in driverless cars right now it's kind of like the icing on the cake like right sounds amazing it is interesting icing but like the reality is the real problem is like recognizing that stop fire truck like there's no moral component of that and there is no you know identifying with the likely blood alcohol level is of the other driver is more like you know is that red thing a truck or not and right you know the systems are dicey at that so it like it's really interesting stuff it's not inconceivable that ai systems might ever be able to do that level of reasoning but right now people are like racing to get the first you know levelfive driverless car out in there like cutting corners to show that they're better than the other and so forth nobody is really buildings systems with that level of sophistication they're like trying to build something that works on sunny days and try not to worry too much about snow although actually just saw an interesting video of jandakot driving their cars and the snow but like this stuff you're talking about is the far beyond the level of sophistication that is currently possible with a believe in conversation aren't aren't correspondence uh on that very are about that very article after i had uh caught up with you i think theorized maybe you were talking about data that uh most people would prefer the their car to be equipped with a an a i that that uh tilted in favor of their passenger uh but they wanted all other cars tilted in favor of the bus other scolari they know if it was their passenger but um in cells because if people are so called utilitarian like let's say the most lives as long as allies involved aren't their own actually no passengers one of the control conditions people done in these studies yet or not it's an interesting question and it might depend like as the passenger my brother or is the basle meyer aluminium pm my child or is it somebody i'm giving a lift in the new ver murray jaggery out you know any of badly people are a little bit less concerned about near the hitchhiker or their lift passenger and then than they are about.
"ai systems" Discussed on Le Show
"If i have a bunch of data and i'm trying to guess what happens to the next piece of data and it's in between the examples axime before the let's called interpretation and kern ai systems are pretty good at that but human beings can do something else called extrapolation where you go kind of outside of the boundaries of the data that you've seen before so i give you a sequence like i'll give you varies easy one two four six eight and even if you haven't heard the continuation you can guess i will these are you know even numbers proceeding by two and i'll go with ten as the next one even though that's not in the range of the numbers that you've seen before so that was no writer that got 10 was around there actually there's a interesting philosophical muzzle their right to call the problem of under determination so actually there's no right answer but there's there's a um how do i put it there there are some answers that at least two human mind would make more sense and if you had there's no right answering the problem that i pose other thing you know gary feels that tend to be the right answer but but there can be real world problems we do want to extrapolate so let me give you an example of that that's less artificial which is you can see a whole range of examples of people driving in different situations you record the data and you could try to just mimic what humans have done when they're driving and maybe that gives you a lot of different conditions on the highway and now i take you off the highway will let's extrapolation so that you have to do something outside of what you've seen before and some of the problems we've seen with driverless cars are really extrapolation problem so you know a tesla drove into we stopped fire truck the other day at sixty five miles an hour and i mean no human would ever do that and human is extrapolating may be beyond the range of experience and most of us probably of not seen stop fire trucks on the highway near hope most of us don't have after.
"ai systems" Discussed on Bloomberg Radio New York
"You know equal chew but in financial services alibaba is way way ahead name me any wallet of any of these companies in america name me one people can maybe think of of apple pay but otherwise you can't you can't think of any so so so so basically these guys had it had free rein for five years to get out in the far east to go to the rest of the world and financial services and they haven't done anything okay i get that as it relates to fend tech bedouin when i hear the term a iit slightly different than machine learning you're talking about cloudbased even in some cases of block chain based transactions that would really be targeted more to the financial services industry but i i guess i'm stepping away and trying to understand the big picture i mean artificial intelligence as an all encompassing all encompassing phrase have tried indeed and so artificial intelligence as we understand that is the way in which machine learning is glued on to these six areas i mentioned and in the case of cloud you know alibaba is at least as good as amazon and the case of autonomous vehicles a lot of the chinese manufacturers are as good as or ahead what a lot of my clients are saying google you spend ten billion dollars a year show me the products and you you spending all this money where the products worthy application my point to you is china is superior at the applications of this stuff in terms of the way in which it it it brings us to consumers and pi 2point now the open ai system of of alibaba is a new world standard pie to buy knows a really big deal india research that you mentioned quite a few out of areas as as well what else do you thing is something that could potentially not rival alibaba that come as close as being a good investment and hats on that night as well when you mentioned blac chyna you looking at any days cryptocurrencies a you looking at bitcoin yeah i think block china's a really big deal blac chyna i think is the most important thing since they get an it's going to change the way that we understand the balance sheet uh there are companies out there doing stuff ripple aquitaine as.