Artificial Intelligence

Listen to the latest news, expert opinions and analyses on the ever-expanding world of artificial intelligence, data science and machine learning, broadcast on leading talk radio shows and premium podcasts.

Interview With Caroline Gorski Of Rolls Royce

Artificial Intelligence in Industry

06:10 min | 15 hrs ago

Interview With Caroline Gorski Of Rolls Royce

"Caroline I think we're gonna be diving into the topic of logistics today. And i think what people think about applying ai. Logistics are often thinking about tracking where trucks are tracking our inventory levels. But for you folks who've worked on some ai. Applications leveraging nlp for logistics and supply chain purposes obviously rolls royce of massive industrial organization. You have a lot of supply chain needs. Can you go a little bit into what the problem was. What you guys have kinda built to help your cause. They're sure dan absolutely so. Yeah i mean just a little bit of context rolls royce overseas a hundred year old company and as a result we are also company that manufactures highly complex and very high value physical things and as a result. We have a very complicated supply chain structure. We also have certain parts of our supply chain where there are very few companies around the world who can actually make what we need from all supply chain so they're all points in our supply chain weather arrow some quite quite high levels of risk in terms of sourcing strategy. Because we don't have an infinite number of supplies you cannot and on top of that we have or certainly did have a real challenge with the presence of what can really only be described described as incredibly dumb unstructured data across our supply chain so information held in pediatrics information on contracting terms on supply standards or capabilities or even designs on top of that an awful lot of our supply. Chain information is either engineering drawings or it's mathematical notation tabulated information and those are three categories off of data which is very hard for existing kind of simplistic optical character recognition tools to actually extract reliably. So over the last couple of years we've been working to develop a natural language processing computer vision capability that she allows us to extract intelligence from all of those kind of dumb unstructured data sources. So that we can use not intelligence to be much more adaptive much more flexible much more responsive in managing risk across supply chain and clearly given the challenges that we faced in some of our markets globalization being one of them over the last twelve months with the covid nineteen pandemic having that increased capability for managing supply chain risk flexibly being able to understand using machine intelligence to actually understand induce scenario modeling across ask supply chain that has helped to support our business in making very significant financial savings in terms of its response to the pandemic but also broadly in terms of ensuring businesses fit for purpose as we become out to the pandemic in terms of managing complex supply. Chain starches for the future hansard. There's so many ways that this can come into place we we've done a lot of One of the sectors where we do our ai. Opportunity landscape research. Every year is chain. Logistics looking at everything in terms of matching loads two vehicles for for transportation to inventory prediction an and arrival times and whatnot. Nlp in this space is interesting is novel and i can imagine so many ways that it might be leveraged can imagine you reading news and information about your various suppliers. Maybe it's the weather in the area. Maybe it's something about them. Having a bad quarter whatever the case may be and i can see that maybe being factored into what production volumes you think they may be able to do or what arrival times you might be able to expect or i could even say a system that just puts the most important news in front of a human analyst who can put it in broader context. Because of course there's so much context for that dumb data to actually speak to business needs on how specifically is nlp sort of working here. And what's maybe an example of where this is starting to be able to inform our processes inform our prediction. Or whatever it's doing yeah. I you're absolutely spots on that donna. Interestingly some of these cases or those areas that you've just mentioned there are exactly where we want to take capability next cohort of the work that we've been doing for rolls royce as a global entity in the first phase of development of this capability. We have been working with those tricky types of data that exists in manufacturing and engineering supply chain unstructured data sets. So those would let me give you an example much of what we communicate while supplies is communicated through drawings engineering drawings. So these these two d drawings which then need to be rendered into three day geometry's in order to be able to understand way to no debate able to understand how much waste material might be generated from from making components. Nfl which the supplier that component of will reimburse us for of course because they charge us for the bulk weight of the material and then the machine material they reinvest because they resell out secondary market so for us to understand for example you know how much is something away how much we're gonna cost to ship and transport How much rebate should we be guessing from from the waste material. That's been resolved by the supplier on. We need to be able to render a to d draw ring into its three d geometry now most of the during dumped exist in cat they only exist to droids so using a combination of competition and alpay. The nlp helps to extract the numerical information. That geometrical information is written around the drawing and the computer vision helps to manage the actual rendering of the drawing itself. You can actually turn your community station into three d rendering virtually and that of and allows you to couch might understand always questions about white about pricings about costing about justice about waste material might be generated for manufacturing

Royce Caroline DAN Donna NFL
Interview With Ilya Gelfenbeyn

The Voicebot Podcast

04:43 min | 1 d ago

Interview With Ilya Gelfenbeyn

"Yulia gif bain. Welcome to the voice. Podcast hey breath how are you. I've really i'm really good. It's great to finally get on the bike. We've been talking about this for at least two months. So i'm happy that we're able to arrange the the funny thing is in the interim we seem to have spent a lot of time on clubhouse together so maybe we could have just done this year like two weeks ago. Yeah yeah spending a lotta time there all right so i think there's a there's a lot of story to be told here because you've you've been in the industry for a while but i think there's an interesting i that i wanted to start with. I wanna talk about speak to it. Maybe maybe your journey in voice day. I started before that. Because i i know that you study computational linguistics as an undergraduate but what i draw you to the idea of of using speech technology and actually building an assistant. Yeah so i would say like initially it was not about speech was about like chat bots and chat right so i started to work on some like chat bots dick like question answering systems back university as you mentioned yelich. All i was doing the computational linguistics relatively randomly so i was like interested in an internship in it company and the a guy who was like my manager there. He was a computational linguists so He had some interesting like thanks for us to do to research. Soviet play with chad bots. So i remember. I think like i published my first article question answering systems back in two thousand towards southern three and then later maybe in two thousand seven i was also working on the project related to chat bots. We created a platform where users could create their own chat bots like mostly for fun Trades they're like our cars teach them to Like teach them different things And then place them to like blogs Social networks websites and see how they shot to like their friends read logs and and Correct them so and later Like when the star to speak to There was this kind of state where got relatively good quality of speech recognition. Then mostly available. Either roy door or some commercial solutions like like from nuance. You've got a mobile devices that our full enough and Kevin good interfaces such as like iphones and androids and There was like this tendency of Api's of like web services so mehan co-founders. We basically thought that if we combine all of this right open api is and and smartphones and voice. We could get A voice personal assistant. That will understand what you are asking about support station and then like connect to a those. Api's and get an answer for you or like have an action Down for you. So this is how this idea of speak to appeared and what year the little. Yeah the end of two thousand ten and to ten okay. So that's around the same time that the siri app. I came out and ios correct. Yes yes yes frankly. We didn't know about siri when we started to sing about the same main difference was that we started with. Hr bought right so we'll just because of our experience chat bots before idea was that won't i. We create a bot that you can talk to about anything and it supports conversation just question answer and then starts like injection different services to you know. Let's add the weather. Let's add local storage and so on but initially we would say you know you can just talk to talk to it Speak to about anything and by the way to also help like multiple different requests

Yulia Gif Bain Chad Bots Mehan Co ROY Kevin Siri
Overcoming AI Deployment Challenges In The Enterprise With Mahmoud Arram Of Bluecore

Artificial Intelligence in Industry

05:43 min | 5 d ago

Overcoming AI Deployment Challenges In The Enterprise With Mahmoud Arram Of Bluecore

"So mood. I wanna deal to dive in with here on this theme of our thursday interviews around making business case for a means. Different things to different people you know. All i know is when an executive is deciding to adopt a or not ex- deployment they're looking at a certain number of component parts to make that decision. What are those key parts for you. Thank you for having me. Yeah from from my perspective so our customers are brands and retailers. And i have found that is a bit of buzzword. Everyone right now. Has the i in their name including a company where i bought my standing desk from so it has lost. Its meaning i think when it came when it comes to business cases so the way i have been thinking about it is in terms of accelerating ghouls that the companies already have in the case of brands and retailers. What has been happening Especially in the pre covid world. Is that the cost of acquiring customers has been going considerably up. There's a lot of venture capital money. That's going into direct to consumer brands. Everyone is buying ads in order to acquire new customers. but then no-one has been thinking about retaining customers. Retaining customers is much more efficient and in order to retain customers and one of the elements you have to do. There is to communicate with them at the level that the like and the able to personalize content at least in the context of retail. Now a i can make that possible. And essentially what we replace existing workflows and outdated technology that makes retaining customers cost effective and makes a from a workflow perspective very expensive and all of that easy and we just happened to us to make that possible. Yes you're saying accentuating in existing already kind of present goal that said i guess. Different kinds of deployments they involve different factors here. So i'm thinking about what it looks like to apply to detect fraud or to build a chat bot where we've gotta get a pretty strong corpus of our own data together we've gotta clean and harmonize that stuff. We've gotta get cross functional teams to come together. Maybe make sense of that. Some applications like a security camera that detects people. Well it's pretrained. I don't need any interaction. You're buying it software. It's off the shelf end of story. But i would imagine for a deployment often. We do have those realistic considerations so we don't have to sell with the whizbang like a nimrod that like. Ai is cool for its own sake. I think safely squarely. That's for nimrods only in this. Podcast hopefully has very few of them tuning in certainly if they've been berated long enough with the messages that we've been sending to them in best practices but those considerations still feel real and feel like they're things that leadership is going to have to address. How is that presented on the table when people are saying. Hey or nay whether it's to your solution or something else. Yeah what we've seen is that there's a lot of digital transformation project out there especially in retail which is vertical that we focus on a minimum buzzword in there. Lots of consultancies. That are working on this. What we've noticed is that there are lots of these projects that have been going sideways. A lot of money has been spent on those and essentially a lot of the challenges that they've run into are around essentially what you were just saying. What is the collection. How do i collect even my own data at silos. It's in different parts. It's different databases you. You think about retail. It has a very complex data. Set that moves. The difference steve's In very different levels of structure can have real time interactions on your website. And then you have inventory movement in physical stores so usually you're getting that data all in one place in order to even run analysis like let alone execution on top of becomes a hard problem right and there are many cases in which is solves that solves the collection of data. I would say in on this case it would be machine learning and even the precursor of it which is how do i actually wrangle all of this data. Put it into one place. So that i can actually run workflows on it. It's so i would say usually that usually that is the consideration is how to solve the problem. Before you even embark on the part. That feels like. I mean if you're you know you're selling whether you're a service or a product you're selling into an enterprise. They're going to have to overcome that right. It's not you're not just going to be like well you can figure the api's and up. I'll show users how to use an internet. See you later guys. It's not really like that. We're going to have to dive into these silos to some degree. How do you present that without scaring people away. Hey look this is going to involve some integration here evolve harmonizing some stuff. This is gonna involve work in new ways and think it through new problems like you said we're identifying with a goal that we know is important to the client. I think that's tremendously sharp smart. You know more pressing than ever in this kobe era. But but how do we present the realities of what the planet look like without making it spooky. So let's easier said than done. The reality is in a lot of these digital transformation projects that included a component. The integration part is usually what fails everyone says. Oh yeah. I have the is. Of course you can put this system that system. And you can integrate oracle with adobe and the reality. Is you know companies lack either. The technical background to do this or usually everyone has such a snowflake off an implementation. There's so many new. You look at a marketing automation. There are five thousand different than theirs on the keep sheet so everyone has a very different permutation of systems on their staff and integrating all of those together. Especially if you are a new vendor on the stat is really really difficult so that is a very common failure which you have to overcome in order to be able to be successful. So there's success of being able to sell in their success at being able to actually deploy.

Steve Oracle Adobe
Interview With Kfir Yeshayahu

AI Today Podcast: Artificial Intelligence Insights, Experts, and Opinion

05:09 min | 6 d ago

Interview With Kfir Yeshayahu

"Our guest. Today is kafir yahoo. Who's the senior vice president of products. Advair atone so high kefir on. Thanks so much for joining us today. One it's gonna be with you today the 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 their tone of course so under the product invade on of busy building. I will which is the alternative system for high as well as accompanying tools to really solve some of the toughest problems in adoption today. In general one of those challenges that were focused on is open holder tising. Ai which is a fascinating topic for me. Having started with data science many many years ago. I've seen how data science was done in some of the most sophisticated data organizations in the world back in the israeli intelligence community. And i've been living the sort of evolution mini revolutions fe. I ever since in loud companies like microsoft and has of groups in the and others before joining the verizon i managed devops oriented boorda in. Aws the amazon cloud so my perspective on 'em at all is coming from both sides from day iside and Devops and now. I'm saying that the dan's in the space who both the perspective of on itself. And i don't own business units as well as Without many customers of the iowa who are in various stages of the junk. That's really very insightful. Because we've been definitely spending a lot of time talking about machine learning operations. Emily obstinate model management. And all these things that have you have to deal with once. The model is as building. People tend to think of of sort of all the work that has to go into training a model making the model happen which is definitely a lot of work. No doubt about it especially even the data preparation even before even build the model right. That's a lot of work but now that you know these models are out there in the wild in in production people are realizing the challenges of keeping these models relevant and high performance. And just doing what they're supposed to be doing. So maybe you could talk to us about what you see. Some organizational challenges as they tried to bring machine models into production of course so different studies and surveys though talking about some little between fifty to ninety two why blamed but fifty two nine hundred projects. Don't actually make it phone. Put that to production regardless of where you fall in the way. It's a pretty sad ratio. Now what makes it even wolves. Is that a project. Take a long time to demand often six months to a year. And you know they'll be walked by Most expensive in the organization we invade have experienced the same thing in the past the first day i projects and applications developed by our business units to literally month to complete. Now we're looking at fox and the stakes were sometimes too high to even start. The border. don't going directly a question about challenges until recently. A lot of the buzz in the industry was about talent shouted. I think this issue is is going away. The market is is balancing itself and good talent is coming from all sorts of different defections into the will. it did not cheap way more accessible than before. The challenges didn't have shifted in my opinion from talent gaps to both insistent cups. And i'll try to gonna show that in four different buckets. So one of them is is portions. How do i estimate the ally often. Ai project how do i define the budget endgame. That's very different than than traditional software projects. Why d- projects the second bucket is integration applications and and solutions in genoa. Now this may sound of sideways from from but it's out of the whole challenge of production izing. How with to play with the application. This is interesting for me especially from extent point because naive. Boches don't always walk because of the nature of the modern. They'll give you one of many examples. Ai models often produce results with degrees eleven of confidence.

Boorda Yahoo Verizon Amazon Emily Iowa DAN Microsoft FOX Genoa Boches
Interview with Sivan Metzger

AI Today Podcast: Artificial Intelligence Insights, Experts, and Opinion

07:47 min | Last week

Interview with Sivan Metzger

"Yeah simmer so excited to have with us today. Seon metzker who leads the milwaukee business at data robot so welcome and thank you for joining us today. Thank you happy to be here today. Thank you for having been. They'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 data robot. Thanks first of all today's an international podcast and actually Speaking israel so. I'm seven in a mike. Mike company which would was acquired by data robots About a year and a half ago we were probably the first. Emma loves company in the industry as a matter of fact actually trait trademark down. Emma locks which we still own today that trademark. We're not using it or consuming. It right now Just running with him a lot as general industry factor. So what what we've been doing since Joining getting robots essentially really building And new business growth engine for the robot is pertains to Running all the learning models that are written in created anywhere in any language or being managed anywhere on any platform essentially mo ops from data. Robot is really the via the abstraction layer in the management layer to allow for mont managing any and all that created rennick being run. Anyone central location. This is Sedgley augmenting the core business which data robot was created and really built on the original category of automated machine learning auto melt since twenty twelve when it was founded in until advocates. Really a probably the driving force in the now a much more mature category Landed the beautiful connection in leading into an ops. The next essential next step if the building machine learning is one thing running she. Learning of course is the next step. I realized badly machine learning today Enormous fastest growing but second the second business inside of the data robot to the future. Who knows we'll be in so History enquiring company's growing the business. Another company we business. We were building a while back to automate machine. Learning was a time series. They mentioned it also acquired another company data space in a all. These together are combined to the overall vision zero to become an enterprise. Ai company working with customers all over the world fulfilling and debris and becoming facto for that. Yeah i mean. I think that's really one of the notable things we've noticed in the industry. I mean we actually pre produced as we had a report out recently part of our research service focusing on the machine learning platforms market. One of the things that we noted is that companies like data robot are continuing to expand the footprint of the things they do and that was that was also something that was focused as part of our machine learning life cycle conference which we ran january twenty sixth through twenty eighth raw of our listeners. Who are listening here. Twenty twenty one and it's available for free so if you if our listeners would like to come and watch the content. You're certainly welcome to do. So if you go to m. l. life cycle con dot com. You can see of quite actually quite a few sessions from data robot That that's their an on demand basis. You can go and you can watch all these videos and you can look at all these sessions about applying Well not just Amil ops but across the whole life cycle of machine learning. And that's really one of those notable things you know. Data robot continues to expand. So maybe you can give us a little bit of insight here you know. How has the data robot product expanded over the past few years to grow and encompass a wider range of these machine. Learning life cycle needs cool. Yeah i think go beyond the robot. I think my ovation of the industry. I wonder what you think about it. But it's we are collectively as industry really progressing in this journey of machine learning right and of course under those have been around for over fifty years of in the last decade or so the transformation the triangulation between snow so much view our source for data available now applying that. She likes to that to become of his bolstering. The ford is of the master situation. We're in within that everyone every company. No organization is progressing at their pace. But for the most part. I had the same my over twenty years in software. This is the first industry. I'm seeing that so. Many companies are actually progressing. Almost the same pace like the notre coming in things are changing very fast. But everyone's on cutting like a lot of people. Are you know trying to be as close as they ended the cutting edge and know the growth and progression through this journey almost homogeneous give just two examples though on front data robot that really builds the automo- category starting nine years ago and came out to the market about seven years ago. Nobody knew what it was and nobody really understood it and you fast forward seven years. There's so many auto companies nekia abilities out. There were people. The concept auto mail has become a very very strong way to ahmed. This is deemed. Make their work so much faster and more effective more efficient now. What actually happened when this became a prevailing Gory kind of birth. The next problem which is a great. Now there's an abundance of moms. They're actually trained. Prepared the ready to go. But the the chasm between taking all those mobs that already crossed industry across the world and actually make him effective enjoy contribution to the business at scale and that in coming to coming to that of valuation i mentioned is become like the present frontier and i'm really a been so focused on him elapse for the last four years in. I gotta tell you for years ago three years ago even two years ago even a year and a half ago. There weren't really that many people that were ready for it. So i was spending most of my time. Educating cdo's in four lucas and thinkers people were looking. What's going to happen two years away. But as we turned into twenty twenty and definitely towards the middle of twenty twenty we started to see the spike in a a rich transformation demands it stopped being education and it started being real strong in the market demand. And you could argue. It's because of the covid situation which actually accelerated the requirements to see value to see things management better scale but regardless it's also on the on the maturity curve and it's it's kinda cool to see how it suddenly turned spike and we saw a nice grabbed from actually google translated on in a The the keyword ops was very flat up to the mid level of twenty twenty and then it starts to spike in terms of search terms volume in google from the twenty points. Just been going crazy so you know. It's nice to see all the instrument during the same pace and coming to the point where you know Volusia data robot. I think it's it's either over Quivalent to the ability of the industry. And we're fueling that empowering and i think and we've had the right now we're we're helping. Let's say you know. Fueled the progression here by allowing. Actually people say hey. They're actually technologies and solutions. That are here to help you. Accelerate your situation across the The journey across maturity curve is not like it was years ago in the word solutions in every company to to out everything on their own is much more mature right now has more solutions in the market. That are coming in

Seon Metzker Emma Locks Rennick Sedgley Milwaukee Emma Israel Mike Ford Lucas Google Volusia
IBM's Watson Illustrates Why Applying A.I. to Healthcare Is So Hard

WSJ Tech News Briefing

04:27 min | Last week

IBM's Watson Illustrates Why Applying A.I. to Healthcare Is So Hard

"About a decade ago. Ibm rolled out watson. One of the earliest artificial intelligence systems out. There watson was a big deal for ibm. You might remember that even went on and absolutely crushed the human competition it was a milestone in how we think about our relationship to computers and ibm wanted to take that technology and apply it to helping doctors diagnosed and cure cancer. But things didn't exactly happen that way and last week we reported that ibm was exploring a sale of its watson health unit. So what happened. And what does this tell us about the challenges of applying ai to healthcare for answers we turn to our digital science editor daniella hernandez hate mail. Thanks for joining me. Thanks for having me. So whereas watson now and what happened well i mean the struggles at ibm with watson. Been around for a little while. We reported in two thousand eighteen that the technology was really not getting the market share and adoption that it needed to make good on all the investments in all the acquisitions that ibm made in order to make watson a leader in the ai in healthcare field and so three years or so later it signals that you know the technology maybe wasn't working as well as they would have hoped. I think more. Broadly points to the fact that you know just having data or collaborations with leading scientists around the country. That just isn't enough and the reason is you know. Healthcare is complicated. So there's a lot of human issues at stake here. You know people do things differently. Like depending on which hospital you're at louisville depending on which doctor you're you're you're seeing but also the data in healthcare is messy for some of those same reasons you know you might input into a medical chart differently than me and for an i i might as well be two completely different things and so just that standardization of the information is really critical but also really hard and so when ibm started making these huge investments in watson they started buying up all these companies that had a lot of seemingly great data and the data might have been perfect but those data were basically styles from each other. They couldn't talk to each other and they never quite figured out how to meld them together. So they were cohesive data set of product. That really could make good on the promise that they that they saw. Fortunately has never materialized. And of course we should note here. That ibm says that watson has had some successes and that they're still believers in that technology we've been talking about. Ibm's new ceo. Arvind krishna on the show and following. He's been trying to of revitalize this legacy company how the sale of watson health fit into his efforts. Well i think one huge thing that has changed since the birth of watson. If you will is that you've had these other huge not legacy players come into the field. You've got google facebook amazon even microsoft right which you might consider a legacy company but they really rebranded themselves to. They weren't as big when watson. I came on the scene. And so now you've got this against storied legacy company competing with these new players. Who when they started making investments in. Ai were a lot more nimble and so they made investments in what at the time seemed like really experimental ai technology and now looking back like deep mind. Google investing hundreds of millions of dollars in that that technology just basically took over the world and ibm didn't really invest in that technology at the time and now is behind because all the talent is has been sucked into google facebook amazon apple And so they're they're behind.

IBM Watson Cure Cancer Daniella Hernandez Arvind Krishna Louisville Google Amazon Facebook Microsoft Apple
Fault Tolerant Distributed Gradient Descent

Data Skeptic

06:36 min | Last week

Fault Tolerant Distributed Gradient Descent

"Hello my name is native. And i'm a computer scientist. Gay as opposed to grow researcher at ebay university switzerland. I work on distributed computations. Specifically i work on. Algorithms concerning distribution optimization disrupted consensus and distributed collaborations systems soi robotics and online voting and in this particular eighty of research. I mainly focus on darwin's by indeed for owners many avi considered networks. There's some of the notes in the networks are militias adversarial or they're just forty. And how does this affect the oral fixation off the elegant now. This is something that's interesting to me in my engineering part of the brain working for big company. I'd say well we can control all of our nodes about some rogue employees or something. But i guess outside of a company the world really runs on bigger systems. How common are these sort of peer to peer distributed problems in people's everyday lives definitely. I made for example. Just consider our current situation which is the ongoing pandemic. Let's see your these. Scooby trackers right you have these scored trackers on your phones than you. Are gary these records with the sword these strikers in vancouver you go in close proximity to someone else that say who might annette it or who might in recent history being tested positive so this gives you a notification. Hey you know you gaming on equity this person and he was the best positive so you may want to take some precautions. Analysts say based on these kind of trackers companies and government started building policies. Just imagine how difficult it would be if someone starts messing around for example. Let's say be Falsely claimed themselves as high-risk saying that. I just hires Ever i'm a Is also is that would discreet Rice appear to be systems are already there used is just a v are naively ignoring the fact that some of the notes in the systems be malicious authority. Fonte just like internet. You have millions of notes on the net not everything. Not everything is on the cloud on server controlled by big or big komen administration. They are so many of these notes that are spreading misinformation they have destroying to disrupt the internet for example you might have heard of. Thanks like jamming attacks Jammed the settlers despite sandy query. So you know these kinds of notes vais fairly common in our everyday use in suggests we get to hear them band. There's a big disaster. Or there's an actual big down of these systems when i started learning about distributed computing and of course it was called big data at the time. One of the first examples is term frequencies so in a lot of documents you the percentage of times you see something that often gets labeled as embarrassingly parallel because you just want to frequency. You need the numerator divided by the denominator and it's easy to divide that problem up and rejoin it but not. Every problem is so embarrassingly easy to solve. What about your specific research into gradient descent. What makes that one hard to do in a distributed fashion. You had distributed Any one this networks to be useful to the specific problem. Let's say that you are trying to solve for example as you just mentioned either. You're trying to get the frequency of a border from documents spent on the internet or let's take a step forward but apps you are trying to build some kind of image classified so different notes on the have different data sets. Let's say images of their dogs. And you want to use these images of dogs neck so many images of those who could be a very strong image classic fire for dogs something that would dismantle the human ability declassified at all so you wanna less this complicated Now designing image classified using all these distribution data said on the internet different notes having different data points. It's quite complicated even when you have all the data points at one body a machine. It's a headline. more of the. The sets are divided in different machines so to ensure that comes like these like machine learning run smoothly when certain nodes in the network militias spinal challenging not descending. It's very interesting to study. How these types can be done. Smoothly on gedeon descent is specific That is by far the was algorithm used from sheet lending pieces and what happens ingredient descent. Is you have these different. Data sets distributed on different notes. Therefore defend nodes have different loss functions. Now what we are trying to do now in this district. Setting is minimized the aggregate of all these functions are loss functions. Now when you're doing this by nature the most commonly use angry gradients design. Because it's naturally distributed many obligingly send it just reduces at the gradients of loss functions in each round. Or so if you're learning and gordon to when you're adding these gradients together some notes are not going to provide you ladies of los angeles. They may be forty. May be broadway. You re totally incorrect. Greediest maybe designed maliciously innovate to move you. words solution. that famous day goes data points. They may favor. Let's say dogs of a burglary or some other panels notions. Maybe they want to completely rendered the classification problem useless. They want to maybe instead of dog. Trained you the classify so as you can see ensuring that is reputed gradient descent runs smoothly at least within some reasonable dominates in residents of such Notes is of practically at this point vendor. You have all this The algorithms training using data sets coming from all sorts of people on it.

Ebay University Komen Administration Darwin Switzerland Annette Gary Vancouver Gedeon Rice Gordon Los Angeles
2021 RevTech Survey Results with David Elkington

Digital Conversations with Billy Bateman

03:57 min | 2 weeks ago

2021 RevTech Survey Results with David Elkington

"Or either everyone welcome to digital conversation inside host. Billy bates joined by david wilkinson chairman jeff phones and we'll be talking about the rebecca benchmark survey that we did in conjunction with the rope. Texoma last month. Dave extra data to do it was a super interesting study given. We're going to be blown away with findings yet. I think it's really interesting. Human doctoral roy did the mit study probably a decade ago. Now were together on putting this together for us So what was the objective with the study and a little bit of african doctoral great and i did a study looking response times on on. How quickly leads go cold. And so we had. I've been a lot of work together. He was teaching at mit. At the time. I think subsequently he's been at several universities teaching locally river secret. Byu so came to us. And said i'd like to actually analyze what's happening in the market around technology that supports this new growth title right so the one thing you shared with is there's a new title essential emerging in the market around growth rate growth revenue marketing revenue growth marketing title. Cheap doesn't said. Hey let's do a study to see what's the technology that supporting that more importantly i think what we're trying to understand was is it. Silence today was becoming the same thing. Let's kinda dig into a little bit of what we found before. Sure some of the methodology. I've got some notes is people doing. This went to the right or to the left. They'll know what i'm doing but the attended the we actually had a shockingly large number of respondents. We had almost nine hundred respondents survey today. That's that's a crazy number honestly yet. The reason we were able to to get such a large number is this was tied to your event Everyone that registered had the option to take the survey took anywhere to five minutes depending on how detailed you want to be But we went through. We were asking. Okay what are you guys use in the jags revenue technology. We broke it down the do of what we thought were the top ten. Kelly works and the findings were really interested. So yeah the categories were. We were looking. Well i think i have it here. Kind of standard categories. You'd expect so yes. See our marketing. Automation email marketing. Seo comes marketing compensation the usual stuff web optimization solution. Diligence sales may our sales automation like his aunt outreach. That's right and i think you'll see here soon. There's of the study. It's about probably seventy pages pretty pretty groovy study and there are fifty six charge. Ten of them are just the results of the questions. Eleven analyze the technologies used by region nine analyze based off the company size ten analyzed by industry and then ten really digging digging into what combinations are effective so again. The court rejected was to say what's happening in revenue related technology. And the kind of just hit that the key finding what we found was the revenue for sure. These technologies are being bundled together. You're using sales. Marketing technologies are no longer separate silos. Were seven and begin to intertwine and we call them couplings. There's lots of a revenue supporting technologies. That are both on the marketing side and sell

Billy Bates David Wilkinson Jeff Phones Texoma BYU MIT ROY Dave Kelly SEO
Building the Product Knowledge Graph At Amazon With Luna Dong

The TWIML AI Podcast

04:11 min | Last week

Building the Product Knowledge Graph At Amazon With Luna Dong

"Art. Everyone i'm here. With luna dong luna is a senior principal scientist with amazon working on product knowledge grass luna. Welcome to the podcast. Thank you nice to meet you. Sam is great to meet you. And i'm looking forward to our chat. Let's get started by having you introduce yourself to our audience. Sell us a little bit about your background and how you came to work in machine learning shire yeah so luna and i worked for amazon and the question regarding highlight came to machine learning. This is interesting question and that reminded me my. The weiser was a phd student at u. Dub i often hard him saying i'm wrong. The community and i came to database down the back door. And the now as you can imagine i got my phd from that database field and best viewed. Where i'm sort of active for a long time and now i'm coming to machine learning from the back door as well so my adviser and i we sort of make a circle so a little bit more about how i came to machine. Learning so my. Phd topic is about dating degration. Basically how we can seamlessly collect data from many many different data sources and integrating them together and then starting from twenty twelve. So that's the time. Google launched knowledge graph and starting from ben knowledge. Graph has been the very popular concept and big companies universities. They put a lot of efforts into it. And if you think about a knowledge graph you put all of the data from different sources and put it into this knowledge graph so saints that been working knowledge graph for the past about nine years for now. And why you'd be knowledge graph. You really need technology from all different fields. This includes natural language processing so you need to understand texts. this includes image processing. You also want to get knowledge found images this includes data mining. You want to mind the data from the text from the grass and also this includes database. You want to integrate the data. You want to clean up the data you want to have high quality data and in the sense to a great knowledge graph. You need all of the technologies. And that's how i came to machine learning field because machine learning is the core for all of these fields interesting interesting when you described the work. You're doing on european. Made me think of this challenge that we've been chasing after for the past ten twenty. Maybe even more years that i think of is like enterprise information immigration. We're going to create their some layer on top of all of the data to make it more easily accessible or some centralized thing that sits on top of all of the information within an organization. It's interesting to think of a knowledge graph as playing that role for many organizations. Tell us a little when you think of knowledge graphs and in particular product knowledge graphs. What are all of the things that go into. Making a robust knowledge graph. This is a great question. So knowledge graph is basic. Clay trying to mimic how human beings look at the real world before we are able to read and a write we already understand the real word and to the little kids. Those are mom. Daddy doggy my house my home. That's another house which is next to my house. And the before any of language thing. There are all of these entities and the relationships between the entities bass how human beings and the stand the real

Luna Dong Luna Amazon Weiser Luna SAM Saints Google Clay
The Future of Computing, AI, Life, and Consciousness

Artificial Intelligence (AI Podcast) with Lex Fridman

04:56 min | Last week

The Future of Computing, AI, Life, and Consciousness

"What's the value and effectiveness of theory versus engineering. This dichotomy in building. Good software or hardware systems. Well it's good designs both. I guess that's pretty obvious. By engineering dean. You know reduction to practice of known methods in sciences to pursuit of discovery. Things that people don't understand or solving problems. Definitions are interesting here. But i was thinking more in theory constructing models. The kind of generalize about how things work engineering is actually building stuff. The pragmatic like okay. We have these nice models but how do we actually get things to work. May be economics is a nice example. Like economists have all these models of the economy works and how different policies will have an effect. But then there's the actual us call it engineering of like actually deploying the policies so computer design is almost all engineering and reduction to practice message now because of the complexity of the computers. We built you know you. You could think you're well we're just go write some code and then will verify and we'll put it together and then you find out that the combination of all that stuff is complicated and then you have to be inventive to figure out how to do it right. So that's definitely has happens a lot and then every so often some big idea happens but it might be one person that ideas in what in space imaging or is it in space a lot sample so one of the limits of computer performances branch predictions. So and there's there's a whole bunch of ideas about how good you could predict a branch and people said there's a limit to it and that's matata curve and somebody came up with a better way to do branch prediction of a lot better. And he published a paper on it and every computer and world now uses it and it was one idea so the the engineer who build branch fiction and hardware. We're happy to drop the one kind of training array and put it in another one so it was. It was a real idea and branch. Prediction is as one of the key problems. Underlying all of sort of the lowest level of software bows down to branch. Prediction boils down on certain computers delimited. By single thread computers ltd two things the predictability of the the branches and predictability the locale of data. So we have predictors at now predict both those pretty well. Yeah so memories. You know a couple hundred cycles away. Local cash couple cycles away. When you're executing fast virtually all the data has to be in the low cash so simple program says you know. Add one to every element array. It's real easy to see what the stream data will be. But you might have a more complicated program. That's you know says. Get a get a element of this array. Look at something. Make a decision. Go get another element. It's kinda random and you think that's really unpredictable. And then you make this big predictor. That looks at this kind of pattern. And you realize well if you get this date in this data then you probably want that one and if you get this one this one and this one you probably want that one and is that theory. Is that engineering like the paper. Those written was that Todd kinda kinda discussion or is it more like. Here's a hack. that works. well it's a little bit of both. There's information theory. I think somewhere to actually trying to prove but once once you know the method implementing. It is an engineering problem. Now there's a flip side of this which is in a big design team. What percentage of people think. They're they're they're they're they're planner their life's work is engineering. Versus design inventing things. So lots of companies will reward you for filing patents. Some many big companies got stock because to get promoted. You have to come up with something new and what happens is everybody's trying to do some random new thing ninety nine percent of which doesn't matter and the basics get neglected and or they get. There's a dichotomy day. Think like the cell library. Baixa cad tools. You know our basic you know software validation methods that simple stuff you know they want to work on the exciting stuff and then they spend lots of time trying to figure out how to patent something and that's mostly useless but the break was on simple stuff. No no you know you have to do the simple stuff really well. If you're building a building bricks you want great brex. So you go to two places to sell brexit one guy says yeah there over there an ugly pile and the other guy is like lovingly tells you about the fifty kinds of bricks and how hard they are beautiful. They are now square. They are you know which when you buy bricks for him which is going to make a better house.

Todd Kinda
Interview With Rachael Tatman PhD Linguist And Rasa Senior Developer Advocate

The Voicebot Podcast

04:31 min | 2 weeks ago

Interview With Rachael Tatman PhD Linguist And Rasa Senior Developer Advocate

"Rachel tavern. Welcome to the voice about podcast. Thank you grabbing me. Well i'm very excited to have you on this. I feel like this is long overdue. So i've been running this podcast. It's two thousand seventeen not long after you. And i met at a conference off of union square san francisco. I cannot remember the name of the conference or the hotel. The park central hotel But you'd made a really interesting presentation there. I wish i still remember parts of it today. So that's four years later. So what about about four years. Bow for years. This this This week even maybe this month certainly And then we had a chance to catch up For a quick lunch and talk about some things. I found it very insightful. so lo and behold you wind up raza. You're sort of in the industry. It's just like a perfect timing but let me let me let you tell your story a little bit. So why don't we start there for the audience. Who might not be familiar with you. Why don't we start with. Why don't we start with your background a little bit. And i think probably the academic background is a start unless you want to start before that so i think it's a reasonable place to to start so i am a phd in linguistics for my for my crimes And i got into linguistics physically. Because i actually going back and reading like Application materials grad school while ago and They were specifically about how i wanted to help. People build language technology. That really worked for everybody in helped make the world. A better place idealist. I can. I had like a lot of ideas. About how hollywood technology would make the road better so which i still think i still hold another one. And at that point i was really into speech and speech perception production from a human standpoint. So how humans perceive speech how to humans understand things And i realized that. I was working on this sort of designed experiments with an eye towards informing automatic speech recognition systems that people who are working on an automatic speech recognition natural language processing automatic speech recognition. We're not going to conferences. And we're not really reading the papers. Don't resident to shift more and more and more into natural. Language processing into more computational approaches In my dissertation i had a big be role experimental component big valuation component. And then also. I built up machine. Learning model that tried to emulate some of the things that humans did and specifically these were all around The ways that you use social information in speech production sorry in speech perception understanding each the you here and as part of that debate evaluation of a bunch of sr systems and this was in two thousand sixteen so awhile ago Looking at the ways that they were able to handle linguistic variation like a regional dialects. Or on i looked at african english I looked at gender and how that affected performance and it turned out the performance was best. You know white people who spoke very standardized prestige dialect not so much people who had a variety of language associated war with regional identity in metric. Finish my phd. Starting my data science wasn't a field. Who could do the entity. It was so i went to haggle which is owned by google So we all may be familiar with it. it's a There's a competition component Where people compete to do supervised machine learning problems and whoever does best wins. And there's also a Posted coating environment that they have and data hosting and was working more on that sort of infrastructure side of things. And then i was talking about this. A little bitter You know it's a little bit by this startup. Bills open source framework for building conversational. Ai conversational assistance rosza. And that's where i am now so i am Moved back to more of the inoki space more humour language students of like dita science more generally. I'm that's been my path to hear

Rachel Tavern Park Central Hotel Raza San Francisco Hollywood Google
Sloan-Kettering Spin-Out Harnesses AI to Diagnose Cancer

The Bio Report

05:09 min | Last week

Sloan-Kettering Spin-Out Harnesses AI to Diagnose Cancer

"Leo thanks for joining us. Thank you so much for having me. We're gonna talk about page. It's ai base diagnostics. And how digital pathology has the potential to change out cancer. Patients are diagnosed. Perhaps we can start with the need. What problem is pains trying to address their few different problems. We're building a portfolio of products. And those problems. Really fall into three different buckets. One is to provide more information to pathologists during their their clinical workflow. That will help them ultimately have real time. Quality assurance provide the more information during their diagnostic process as well as ultimately help them with efficiency in throughput. The second need that we're helping address is with our our viewer in our digital pathology platform which is allowing them to access historical images to share uh slides easily to get consultations as well as to be able to reference other digital images and slides during the course of their work and then the last set of needs is really around trying to look for new biomarkers that can help doctors on college costs ensure that the patients get the right treatment every time. And how are these tests generally perform today. How does pages technology change that. So in a clinical pathology setting What happens today is a pathologist or piece of tissue is taken out of a patient from a biopsy or surgery that tissue gets cut stained mounted on a glass slide. And then the pathologist is handed a set of slides to look at that patient at thaad will look at that slide and they may see something that they're not sure what it is. It may be a little unusual. They may ask a colleague. They may send it out for consultation. They may do an additional stain or send it off for some molecular testing ultimately. They're going to get all of that information back and they're going to have to make a call for that patient. What the right. Diagnosis is in a page world. That slide is not looked at under a microscope. Scanned and the pathologist is looking at a computer monitor and pages gone through those slides and matched each slide. The tissue content in that slide those patterns with a database of known tissue and diagnostics. And that information is made available to the pathologist during the course of their their clinical treatment so that they have this additional information available to them automatically forever case having to go through and take those other steps of consultation and sending cases out in additional testing in staining. They'll have that information at their fingertips so that they can get to that right. Decision faster and more standardized more confidently are slice prepared for a page test as they would be for a traditional test today yes exactly the same way in fact there's no additional Staining no additional preparation that's needed. The only piece of additional equipment is the side scattered south and is the digital. Ai system visually reading an image and is doing so in a way that's unique to the machinery somehow mimicking. What a pathologist is looking for. I think the best way to think about how. Ai works is. It's looking for patterns in data in this case patterns and tissue and so that machine is identifying these patterns matching those against database of of known patterns That have been either diagnosed by other sts or that have been results of additional testing like molecular tests or something like that to really match those patterns and then highlight that information to the pathologist during the course of their their diagnostic process as we think about going forward in that biomarker direction that i mentioned and that case These are patterns at just may not know about may not be aware about may not really be visible to the naked eye and yet the computer is able to sift through thousands ten two thousand hundred thousand millions of images and identify patterns that are signatures for treatment responses or other

LEO Cancer
Facial Recognition in Law Enforcement

Malicious Life

04:58 min | 2 weeks ago

Facial Recognition in Law Enforcement

"One month before the new york times published. The story of robert williamson. Police officers killed. George floyd we all remember the weeks that followed the protests that broke up around. The country garnered tension around the world. It was so engrossing watching those scenes of civil strife that an otherwise remarkable part of the story went almost entirely on the radar in fifteen cities is the department of homeland. Security deployed planes helicopters and drums to watch over the protesters. The aircraft hovered over protesters in new york philadelphia. Detroit feeding customs and border patrol man centers which streamed the intel to police forces and national guard on the ground in minneapolis and the a secret rc. Twenty-six be reconnaissance. Blaine worked with ops on the ground streaming video feeds to an fbi command center in another instance in the top spending in the ficials ordered helicopters to provide quote persistent presence to disperse crowds. He helicopters flew so low to the ground that the sheer downward pressure from their rotor. Blades ripped the signs off of billions. And of course since protesters running this was something out of science fiction full on military intelligence operation on. Us land in later reporting military and government officials insisted that none of the aircraft deployed on protesters were equipped with fisher recognition capabilities in most cases. The aircraft were so high up that facial recognition would be moot. You can't make out face from a blip from nineteen thousand feet but even if some planes flew close enough to capture individual faces be problematic according to the new york times at least two hundred and seventy hours of protests footage was captured by the aircraft and uploaded to big pipe. Dhs network which can be accessed by other law enforcement agencies around the country for future investigations. Video in big pipe can be stored for up to five years. One potential concern then. Is that if a plane recorded. Good enough video. It wouldn't need real time fisher. Recognition on board an agency like the fbi could access that footage weeks or months later to identify individual protesters. It's entirely possible that this hasn't and won't happen but around the country police have already utilized facial recognition to identify and in some cases arrest individual. Blm protesters the extent of it is unknown. Police have no obligation to report on the run. Your face through a machine or concerns me is there's currently no limits at all none at all and so they can also use it to take a look at everybody who is in a peaceful protest of some sort And then take down their names and Hassle him or arrest them or give them trouble in one way or another or simply file them in a database as person of interest. None of which we want to have happen to us. This is ted claypool a lawyer and an author on legal issues surrounding privacy. And a here's one example of what ted means by no limits. Batanes out you can be arrested and not even know that facial recognition played partnered one. Protester less summer oriana albor knows was arrested for throwing rocks at police line. That's definitely a crime but no point in her processing. Did miami police mentioned in documentation to her lawyer or in any other capacity that the used clearview to identify her as the rock thrower. It took an independent investigation by nbc. News to uncover that information and it's important for mission right maybe arena was guilty but he next arena could be a robert williams.

Robert Williamson George Floyd Department Of Homeland Detroit Feeding Customs And Bo The New York Times FBI Fisher Blaine Minneapolis Intel Philadelphia New York Ted Claypool BLM Batanes Oriana Albor United States TED Miami
Challenges and opportunities of blockchain in the insurance industry

Insureblocks

05:08 min | 2 weeks ago

Challenges and opportunities of blockchain in the insurance industry

"For this podcast. We will be discussing challenges of blockchain in the insurance industry with special insights from ibm. And i'm very pleased to have mark mclauglin. Ibm's head of insurance strategy solution sales and partnerships worldwide mark. Thank you for joining us today. Could you please give our listeners. A quick introduction on yourself sure thanks lead and thanks for having me on as you said our head of strategy for the insurance vertical for ibm a teams pulled together. Ibm's hardware software services cloud and our business partners to deploy value for the insurance industry. I've been doing that myself For twenty twenty five years now. The first solution i built for the industry was artificial intelligence back in their early nineties for a large insurer here in the us and that solution is still running today which Either tells you something about our industry or something about my coding one or the other. so let's go was your coating absolutely exceptional. Not sure that's really really. No thank you for that introduction so as you know and hearing two blocks and we always ask. Our guest is first question which is what is blockchain. And how does it work the way i. Obviously i think most of your listeners. Are well aware that there is a difference between the blockchain enabled currencies like bitcoin and ripple and the actual blockchain functionality itself. The way i think of it is. It's a shared ledger. The trusted leisure. It's an ability for multiple entities. Who don't necessarily have a one hundred percent trusted relationship. They are business entities with different interests. Different goals but you can establish a common ground wear a set of documents they said of processes a set of data is maintained by a group across a business network and that is maintained in a way that is immutable Where everybody can see the changes that are going on and everybody has a record of what's going on you know whether it be you know data around contract or execution of a business process and being able to do that in a way that is trusted by all participants That can bake in features like smart contracts to help you automate some of those processes right. There's a lot of different things you can do. With the blockchain right currencies one of them but running a lot of shared business processes and other one of them and. i suspect. We'll be talking about that today right now. Thank you thank you for that. So as you are aware we've had a number of your colleagues on the inch. Blocks podcast from bos- expert within the blockchain in the insurance space. Now i'm curious to know from your personal standpoint. how would you characterize. The insurance industry's embrace of blockchain technology. Well i think high interest right. Insurers for a number of reasons are feeling the heat on innovation. Right whether you look at the you know. The forty six percent kager on tech investment. The last three years or the entry of of large-scale players like like paying on like amazon into more kind of online distributed type insurance ventures. Right whether you look at insurance being baked into other industries right when you go by airlines at the united states your offer travel insurance now as part of that process right. It's it's very different than the kiosks in the airport of old right. I think the industry knows that they have to figure out ways to connect to broader ecosystems. Knows they have to innovate and blockchain's one way to do that. There's there's definitely some great opportunities there's definitely some pitfalls but ensures you know high level of interest having a little trouble getting started in some cases and i think we'll dig into that as we go. Yeah exactly because you know we know we started our podcast mainly focus on the insurance industry in two thousand eighteen inch thousand nineteen. We spread out to cover other industries. Which is very fascinating to see the challenges and opportunities each industry have with regards to adopting mom blockchain but this is sticking for insurance for now. You know as you mentioned you had twenty to twenty five years experience in building solutions and partnerships ensures. Would you say that insurers are more or less open to embracing blockchain comparison to other previous or existing modern technologies such as cloud in ai to name a few. As you mentioned. You know you did this project in quite some time ago. How does blockchain compared to these kinds of technology. I think blockchain has great potential and as technology. I think insurers are are more than willing to embrace it. I think the challenges are the business model. Right i can take a And it and it's very easy to visualize. Hey here's here's a case. Where i could see how i might help me process a claim better. You know underwrite risk better advise in indentured. Better now actually getting it to do that is a little bit more challenging but visualize it. it's eas- right blockchain. It's a little tougher for the challenge. Isn't the tech. I think it's the use case behind the technology

IBM Mark Mclauglin Blockchain United States Amazon
The Doomsday Clocks Historic Wake-Up Call With Rachel Bronson

Big Brains

05:11 min | 2 weeks ago

The Doomsday Clocks Historic Wake-Up Call With Rachel Bronson

"Forty five. The united states detonated two atomic bombs over hiroshima and nagasaki. It is harnessing of the basic power. The universe shortly after a group of manhattan project scientists at the university of chicago who helped build the atomic bomb but protested using against people started the bulletin of the atomic scientists. Huge choice is peace or total destruction. the atomic is yeah. They wanted to urge fellow scientists to help shape national and international policy to mitigate the risk of the nuclear technology that they themselves had helped create and they wanted to help the public understand the dangers of nuclear weapons. To the future of humanity world would not be the same. i remember anthony blind from hindu scripture. The by gerrad gita. Now i am become death. Despoil worlds are another in designing the cover of its magazine. The bulletin created something striking o'clock running out of time. It started as artistic piece created by chicago based marta lanes door. She was very to manhattan project. Scientists issues gender stood the scientists concerns about this new technology and the need for public engagement and they had asked her to create some sort of design that would engage the public on. How serious the threat of this new technology. And she said it seven minutes to midnight every year since then. The bulletin set the hands of its metaphorical clock in relation to how close to doomsday. We might be last year. The group moved it to a mere one hundred seconds to midnight. And at the time we got a lot of chiding like it's twenty twenty how come it so close. Do you really believe it's this close and then sure enough. We saw the massive wildfires right outta the gate in in australia. That got repeated in california this year but obviously covid and the inability of the global community to deal effectively with covid is was to us a clear indication of our inability to deal with existential threats. Known some ways you can make an argument that it should have been even closer to midnight this year because you had your existing threats then you had that real life pandemic which is continuing to affect us. How can we didn't go further to midnight. Yeah so in some ways You know we don't want to double count right and so a lot of the warning signs. Were what moved to one hundred seconds to midnight but it is a very dangerous in environment. And we'd we do want to acknowledge that hundred seconds to midnight is dangerous. We do see some bright spots and some opportunities so those bright spots helped us from moving forward but we weren't prepared to move it back. It may be tempting to look at the clock this year and take some hope from the fact that it didn't move closer to but remember it's still the closest to midnight that we have ever been and this year the bolton highlighted new threat one that they said is a threat multiplier to all the other problems that we face with the world health organization called a massive info democ he really grappling with what our trusted new sources. And how do you find them. And how do we share the so. We're all overwhelmed with data and information. But it's very optimistic. When it comes to share it information or what you and i know and so that becomes very disorienting and it becomes Quite dangerous right. It sets up the ability for authoritarian leaders to create their own information and different sites secrete. Their own information will get into the surprising. Ways that this info democ touches every threat factor to the doomsday clock but will start with the issue that was really the canary in the coal. Mine of this info dynamic climate change. The scientists have been warning us for decades and yet they're the ones who have experienced a lot of these issues in terms of misinformation and disinformation. I that denying climate science the marginalization of them the using of science which is kind of about uncertainty and evolution to dismiss what scientists have to say. All of. this was the global warming. And that it's a lot of it's a hoax hoax. Moneymaking industry okay. Climate change is not science. it's religion it pulls the rug out from under scientists and expert exactly the time when such expertise is actually needed and within the context of the us there could be real differences among republicans. Democrats or what you think about market versus regulation. Those are really really important questions that we should be debating fiercely right now that we can when it's being defined as climate change yes or no we can't even have the kind of real political conversations that we should be having

Gerrad Gita Marta Lanes Manhattan Nagasaki Hiroshima University Of Chicago The Bulletin Anthony Chicago United States Australia California Bolton World Health Organization
Serverless Properties with Johann Schleier-Smith

Software Engineering Daily

03:41 min | 2 weeks ago

Serverless Properties with Johann Schleier-Smith

"Welcome to the show. Jeff super excited to be here. You've been looking at service computing from the vantage point of berkeley and talked to a number of other people from berkeley about service. Talk to john. Stoica and vikram. She conti from your point of view. Why has berkeley taken an interest. In service computing. Berkeley has a long history of prominent Research in computer science and systems in particular. Lots of really cutting edge work was done here and think the faculty are always looking for that next thing. That's coming down the pipe and can we be on top of an and ideally ahead of that trend and in the context of service computing. This is something that we latched onto people at berkeley. I wasn't actually the first one myself. Yaas okay eric. Jonas published pirate work back in so john's dog. Eric jonas. They published a pirate work back in two thousand sixteen. Two thousand seventeen were were really saying. Wow service allows us this access to supercomputer scale resources for basically anyone. So i think that people kind of latched onto. Hey there's something new. There's something really different that's happening in the cloud and we should really pay attention that we should try to understand what the implications of this new technology are what to service make easier. What does it make more complicated. What are the trade offs in using services from our perspective services computing is really about making life easier for programmers. That's the big change. Now it makes a number of changes so it certainly makes life easier for operators as well in some cases even completely removing the need for certain system administration responsibilities so everything that's complicated about servers and by that we mean things like setting them up making sure that they are patched for security Making sure that when they fail application is responding in the correct way so that continue to deliver service all of these concerns. Go away the handed to the cloud provider. Cloud provider has ways of automating them away. So that for them. It's also much much easier to manage. So they can imagine for many many companies at scale so the program are also has this ability to basically write code and their favorite programming language upload loaded to the cloud and then it just runs not have to worry about it anymore. And that is i. Think in many ways fulfilling kind of this promise of the cloud to give you that effortless access to scale so the downside of that is that you do have to change how you program a little bit so i think that lambda was successful because it allowed you to bring along your existing libraries logic bring along your existing languages so there's a fair degree of continuity on the other hand if you really are going to make programming simpler you're going to be writing simple programs and that means that you're probably going to be rewriting your programs at the same time so you do have to learn to think a little bit

Berkeley Jeff Super Stoica Eric Jonas Conti Vikram John Jonas Eric Lambda
Interview With Conversation Design Institute co-founder And CEO Hans van Dam

The Voicebot Podcast

05:53 min | 3 weeks ago

Interview With Conversation Design Institute co-founder And CEO Hans van Dam

"Odds van damme. Welcome to the voice by podcast. Thank you so much for having me. Okay so you are. The head of the conversation design institute. Why don't you tell the listeners. A little bit more about what that is just to get them started and then we'll talk a little bit more about how that all came to be. Yeah so conversation design institute. What we do is really recognize develop and promote the role of the conversation designer. So we sort of envisioned at every enterprise in the world is going to have a department. That's going to work on a assistance whether it's chat bots voice assistance and are all going to have people designing these conversations. Deploying them and managing them So we really recognize that as a serious profession and together with the industry. Really try to figure out you know who's going to be working at department. What's the skill set that they need. And then we develop content for that so we have courses. We have a certification program And that's really how we try to position ourselves in the market as the leading training and certification institute for designing versions. Okay yeah that's great okay so for the listeners. Now they have that context. But you didn't just arrive at the conversation designed institute is like the first thing you've ever done so tell me a little bit about how you how the conversation zayn is came to be but like your career and how you move forward. You were just mentioning to me. You thought when you were younger pre university that you thought you might be a writer That's straight into economics. And and i think writing came back a little bit. Yeah that's that's that's definitely true. So for when i was in university i i went to university of amsterdam thinking that was going to be really studying economics and then i got more interested in philosophy literature and just started reading a lot and i was writing fiction short stories and i. I really thought. I want you know was going to become a writer a novelist and wrote a manuscript. Send out too much publisher Publishers stay did not agree with that vision So when it was time to actually get a job. I got interested in copywriting. So it's like what's something related writing. What does that mean so ended up as a copywriter at a startup. incubator through. Ace venture ops part of mit stanford v latte so. So i was helping out these different startups helping them with the proposition and writing copy for them but allow me to then learn about technologies. Well so got more interested in in the tech space Join a little startup. There was doing some video streaming completely failed so when that story ended it was time to get different. Jobs ended up in customer service really as a quick job doing social media for kayla dutch airliner. So fire go into airline and customer service. Exactly so that's it doesn't get better than that and the company though is involved there with see x. company which was a chat bot company Acquired by siham a couple of months ago so i joined. cx company and Day were doing still virtual assistants whereas like you and a so. You'd ask a question they probably would not answer The old school things on a website and all of a sudden like everybody wanna chat bots and got interested in that and became more conversational. Cx company at the time didn't really know what that meant for the content. Their clients had no idea what that meant. And for me is like i always wanted to write dialogues and know from fiction writing understood. The technology understood the service space for me. It became very logical to actually focus on that problem. So i started doing this freelance at a bunch of clients and then i met my co founder and they had a an agency and behavior design so they were a psychologist thinking really about how people interact with products designed design for certain behavior and made a ton of sense. Because they were already doing stuff with conversational interfaces as well and we really. We realized that if you have a conversation between an artificial brain on a human brain you know you should make psychology and technology equally important in all right and so we. We started playing around with that and we had an agency was called robo copy at the time so so we just got a bunch of clients got more experience about designing conversations Started a little academy which is really focused on chat bots and Got in touch with google. Wally brill that that reached out to us and he really took us under his wings. And explain there's a lot of the fundamentals of voice design and connected with other designers So that got us more experienced gutters to work with you know the best designers in the world and and learn from there and as we went on that journey we really saw that the education was a big part of it So we created the academy and now mission or honest. Really that you see that enterprises are looking for conversation. Designers does lots of people that are actually designing conversations. But there is no alignment job title skill sets. It's very fragmented service. Enterprises can't find designers designers can't promote themselves very well tech companies on the other hand. Don't really know who's going to be using the products and how to design freedom universities at university at the same time note that i have to train people for jobs in the future Don't really know what jobs are so. We really tried to seek alignment now in an industry of conversation design. Therefore for conversation design institute was a name more suitable for that as well. So that's really how. I went from failed novelist conversation driver

Conversation Design Institute Zayn University Of Amsterdam Ace Venture Van Damme Siham Kayla Wally Brill Acquired Jobs Google
AI driven Privacy tool developed to protect COVID-19 tracing data

Cyber Security Weekly Podcast

05:48 min | 3 weeks ago

AI driven Privacy tool developed to protect COVID-19 tracing data

"Welcome to carry tv and ad tech insect weekly chris coverage on the editor with more security media. And this is al. Friday morning episode. Normally stream live on tuesday afternoons and on fridays and today's episode where with dr did hornets strategic advocacy manager would stand into strategy and up the sushmita rush at senior research scientists with data. Sixty one gonna be looking at data. Sixty one's recent ion driven promising tool. Personally my shin factor. I think it's go piff and with with don't this mehta rush a senior research scientist with data sixty one. Thank you very much for joining us. Law soc me that look thank you. So much We covered off. And we're gonna be talking about It's great to have data sixty one on obviously as well but this caught mile. I released a new data privacy tool for a anonymous covid. Nineteen tracing data and keeping that secure. And it's cold personal information factor or this. You're the senior research saunas on the project. Might be the adult. Talk us through It's a big topic. Accident had to stop. Whether we start with the i ought to the covid. Non tain tracing data Update class. I can just a little bit about fifth avenue. Exactly the information and Just to be fit. That need talk about why we are doing this. And one of the use cases of course the covid nineteen of data about this is a much more universal kind of a tool which which actually helps to share data a to protect the privacy of individuals whose data is there in the assets and its stock. V personal information factor is essential information content in the data affect and. Just imagine that if i were to the custodian was to release the state asset. Then it looks reveal definitely information individuals so freshman is that went to release and when not to release and this personal information factor is a measure of that information content in that deep affect the identified data and what the tool does is that not only. Does it publishes the data in an in an in some kind of transformed fashion. But it's also evaluates. The risk of free identification. Is very very important. Like when i when i want to share my data. The first thing that i ask is that what is happening to my data. What are the risks associated rely. We get out in this whole list of data that has been you know released. No you can. I ask you. Is it reverse engineering. The fact it's released. And we reverse that. Can i identify that will happen. Writer the tool essentially doctor that you know it. I evaluate what happens. What are the risks. And if it feel that you know the risk is low then it's released the data at the high than it suggests very thoughts of transformations Aggregations techniques so as to make the data more suitable to be released that in it's not reeducation is not possible so the two of you know a lot lot more. To protect the privacy of individuals this is donald sixty runs on another saying albumin. Doctor men whose new south wales chief scientists it shifts on and We've he's also hit it up. We'll previously headed up. The new south wales at data analytics as well is that this is all great working together on this because we've heard from duct tape and previously about the work that they're doing with a lot of this data across the south wales in sydney in particular. Very interesting work. But yeah it's that that The day anonymous anonymously information and they identified. Information is a challenge. Because if you join. The dots suddenly can start to identify. They so yeah. This is actually a on oprah man. The project actually started with the initiative of yet overman dr yang obama and what we are essentially trying to do at the data was also involved from the very beginning. But what we are trying to do. Is that enhance that tool so we want to enhance such way that weekend. Mitigates against various attacks so we are trying to identify what other attacks. What are the attack vectors. That are possible that might breach the privacy of individuals and beth. Israel comes into picture. We are essentially studying what a- what the attacks are and can be do it in a more sophisticated sway to learn from the attack and suggest suggests techniques to protect the privacy of individuals of be it aggregation be. It's probably secure Secure approval private algorithms for like Differential privacy or there would be other solutions that can help to the data at make it fit to be

Mehta Rush AL Chris South Wales Dr Yang Obama New South Wales Donald Trump Sydney Oprah Beth Israel
Leaderless Consensus

Data Skeptic

04:51 min | 3 weeks ago

Leaderless Consensus

"Minus biology. And i'm a student in the system. Softer research group virginia tech. I'm broadly interested in building distributed systems. And i've been doing that since my master's and my phd degrees or the past five years. Broadly i focus on building reliable and high-performance distributed systems very specifically i work on this topic of consensus and agreement protocol van idress different properties of them such as scale ability and resiliency to fox previously also worked on addressing the performance aspects of certain taxes of consensus call the leaderless consensus protocol. Which i believe is the topic of today's discussion. Yeah could you draw that distinction. We've talked a little bit about paxos on some episodes but had a very leader paxos biased to that. I guess what does it look like to leaderless. Paxos the main reason you would want a leader in paxos protocol is because you won't agreement among a collection of processes it could be notes that are spread around in a wide area network or it could be in the same local area network and important a problem. The leader saws is the type of conflicts because if you allow anybody to propose values than they might not reach agreement quickly in fewer communication on. That is critical. So that is why classically a paxos protocols have been proposed the leader so that you can have domination in fewer communication steps decision can be made in. Let's say to communication steps. However what differentiates leader less paxos is that you remove the requirement for leader and introduce a different mechanism so that even without a leader. These different notes can agree on a same set of values that the agree and execute as part of their statement. Do we have to give anything up to go. Leaderless we lose eventual consistency or. Yeah what's the cost of this consistency vice. We are still able to get leaner is ability. We don't lose anything per se. But however the protocol itself gets more complex and more nuanced and subtle paxos itself is a complex protocol to understand dissect however these littlest protocol because of more addition to the original access protocol can get more complicated. You'd mentioned scale. Ability is one of the interesting things to study in these consensus protocols. Most of the papers and research. I did learning these things. We'll give me examples of like you know quorums of five or seven nodes. Which of course are great right. That's the way to learn it but in reality it's internet and cloud scale. Maybe we'd like to have orders of magnitude more nodes. What are some of the scale. Ability challenges you bump into down that path you mean. In terms of leila's protocols are in just in terms of consensus. For god's will either way maybe we can stick to your precise work or if you think it's valuable to contrast it with the more generic case that works to terms scale ability in the crush fault tolerant space before i get deep into cash phone torrens. Let me specify a little bit on the fault models themselves. So typically people work in the crash fault donald space if they're building consensus for a single organization use case data center use case fair people deeply into databases and stuff like that and the assumption in the crash. Fault tolerant space is that the machines can simply crush but they cannot behave maliciously in the sense that when they reach agreement they don't behave in a way as to deviate the consistency of the values. That is being agreed on. This is in contrast to the byzantine fault tolerant agreement problem which addresses a different set of use cases like permission blockchain allegations that require much more scale obliterated then that is required in the trash falter spaced typical even people talk about the fda talk about scale ability to fifty or hundred or more than hundred up to thousands soft notes but in crush on space. People typically talk scale ability up to five or seven notes and that is for very specific reason. And that is that consensus protocols. In general they tend to be very expensive in coordination and the performance tend to get much worse as you scale to higher number of notes and for blockchain applications. If you look at the absolute numbers the performance of the f. d. protocols for blockchain applications are much lower than safety protocols. The reason we stick to a few notes in the safety protocol is because they're not other mechanisms that use like shotting and stuff like that in order to achieve scale ability in a data center sitting.

Virginia Tech FOX Leila Blockchain Donald Trump FDA
Clearview AI’s Facial Recognition App Called Illegal in Canada

Daily Tech News Show

00:25 sec | 3 weeks ago

Clearview AI’s Facial Recognition App Called Illegal in Canada

"New report by canada's privacy commission found clear view is facial recognition database to be illegal mass surveillance and said sent a letter of intention to the company telling it to cease sevices in the country and delete canadian faces. Clearview has not operated in canada since july due to the investigation and says it will allow canadians to opt out of the database. But we'll challenge the determination in court

Privacy Commission Canada Clearview
Amazon faces spying claims over AI cameras in vans

Techmeme Ride Home

01:26 min | 3 weeks ago

Amazon faces spying claims over AI cameras in vans

"Amazon has started deploying. Ai powered netra dine cameras which are always on an automatically uploading footage so that amazon can monitor drivers out in the real world quoting cnbc. Amazon is to play the cameras in amazon branded. Cargo vans used by a handful of companies. That are part of its delivery service partner program which are largely responsible for last mile deliveries. The cameras could be rolled out to additional. Dsp's over time and amazon has already distributed an instructional video dsp informing them of how the cameras work. Dsp's are contracted. Delivery providers usually distinguishable by amazon branded cargo vans responsible for picking up packages from amazon delivery stations and dropping them off at doorsteps. The program launched in two thousand. Eighteen has allowed the company to quickly scale up. Its last mile delivery capabilities and compete with shipping partners such as ups and fedex amazon's dsp program has faced criticism for lax safety protocols in the past investigations by nbc news propublica and buzzfeed news identified safety issues and described poor working conditions at some. Dsp's based on interviews with drivers and former amazon employees. The cameras could help improve safety but privacy advocates and several. Dsp drivers said they're concerned about potential privacy offs. The drivers who asked to remain anonymous out of fear of retaliation from amazon described the cameras as unnerving big brother and a punishment system and

Amazon DSP Cnbc Buzzfeed News Propublica Fedex NBC
interview With Geoffrey Tate Of  Flex Logics

Artificial Intelligence in Industry

04:19 min | 3 weeks ago

interview With Geoffrey Tate Of Flex Logics

"So jeffer-. I wanna start off with certain difference between doing business with the data center versus doing business at the edge. I know that the hardware you folks are working on. Is you know what i think. Most people think hardware. They think about the big rack sitting somewhere in the data center the edges different the edges. New it's its burgeoning. How do you define the edge when you talk to people. Because i think people always think about those racks but it's clearly a blooming ecosystem. Yeah well azusa terms of different people defined differently but where we basically look at businesses any system. Outside of the data center there can be things like cellphones stations in verizon stations that are kind of in between what we're looking at your robots this. He'll cars field ultrasound systems in the field. So these are systems that are separated in well removed from today's yet. Okay and and obviously pretty wide. Berth as to what that could be almost any industry this this could be applied retail. You've got cameras energy you've got i don't know some some big turbine out there. You know generating some power killing in occasional bird. You know it's pretty pretty vast swath of of what what can imply does that. Broad world of edge cluster. In any interesting ways. I think industry would be one that makes sense. Maybe you can talk a bit about that but we also use case what you see edge sort of us for a certain way tammy how do you think about this whole new space. Those of us at home. it's it's sort of. It's new it's novel but how we want to break it up. Well we're just touching the top of the iceberg so we've engaged with a lot of customers see a lot martin's segments and they have different potential sizes so one obvious one is countered. There's cameras all over the place. You mentioned walmart's wells fargo's there's cameras today wired into servers in the back offices of these places and serve servers not in the data center and right now those cameras us recording video in case somebody shoplift something that got her the tape. Now they can add inference start tracking their stores checking buying behavior along the lines along the take to get through the lines things like that. So that's an application where you need object detection and recognition similarly when you're talking about robots robots moving around into distribution for a warehouse. They need to know you know what. Say iraq to put things on. What's the what's the person to make sure you don't get them. Yeah so you're detecting. You're recognizing them taking action appropriately. Same thing happens with cars so those are all object detection recognition models like yolo the three which do an excellent job of doing that. Which are people are flying now and we see medical imaging and there's many types of limiting there's crazy. Mri machines are much less expensive or numerous ultrasound. Machines x-ray cd stamps and stuff in between so there with the people are using models for his more specialized object detection recognition typically. You're standing your knee. It's stuck in there. It's not moving but you're looking to detect some anomalies in the x ray on the ultrasound. South is the baby. Ok is the got busted acl. So it's helping. The radiologist do occur. Job of diagnosing powell. That's what those kinds of models need to be doing. Things like Scientists gamma joining or life sciences. And they are in many cases what they're doing is looking to clean up images using network approaches through extraneous information clarify the. Which if you've ever seen like an ultra sounds like i just recently hasn't surgery. The doctors were trying to find a vein of the my shoulder. I can see the ultrasound. They were looking at a teaching hospital. And i couldn't tell what they said when they when they started a doctor they couldn't tell was on either but eventually they figured it out so these ultrasounds are hard to make out in. Computers can make better judgments which results in better outcomes says. A wide range of applications are seeing for inference models. I think we're just scratching. The surface gets more powerful and cheaper is going to go into more more more systems.

Verizon Tammy Wells Fargo Walmart Yolo Martin Iraq Powell
Interview With Ahmer Inam  And Mark Persaud At Pactera Edge

AI Today Podcast: Artificial Intelligence Insights, Experts, and Opinion

06:24 min | 3 weeks ago

Interview With Ahmer Inam And Mark Persaud At Pactera Edge

"We do have some great guests with us today. So we're really excited to have amer in phnom who's the chief. Ai officer and mark persaud. Who's the head of emerging experiences at pact-era edge. So welcome guys and thank you so much for joining us today but thank you got clean and wrong. Thank you for having in forward to this exciting conversation. We are to. 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 pact-era edge Kathleen i'll go ahead and start amazon up. I'm the chief a offers. Subtle factor edge. We are global solution and services firms that balance intelligent digital platforms using human center design as a cool concept in philosophy the hallway bill systems maya background. I've been in the space of a medal for essentially my entire career Having played with fairly early machine learning. And you'll eulex model for almost twenty years at this point To most recently. A factor edge. My journey has taken me to companies like fargo sonic automotive. Vw see nike can be solution and at the now. I pass it off to mark to introduce himself awesome. I've I'm the head of emerging experiences at a bacteria edge So i have the job of being able to look across different technologies whether that's a our and our in vr virtual reality immersive Or things like voice and conversational. I and understanding how a i can play a better role in the technologies or with the technology whether it's within or atop a different digital ecosystems for clients though. I personally have a lot of fun with that role in general just because it gives me the ability to see how we can create value for users of creative ways with technology is where we might not see very consumer or user friendly and i might add like one of the reason why this is such a differentiation. Headed what you're talking about. Even sent to city in is market. Ni- expedience must. Genie are working together. Cohesively issue the cool part of the conversation that will be having at the at the upcoming event with community and is about building and designing intelligent digital platforms that are built with humans entity the human in the mind and building them to drive adoption so that we can take a lot of these concepts that are explored in a typical machine learning ai. In women in an enterprise and then take them to an enterprise capability and the part of their journey. At least an odd philosophy is that it has to lead with human simplicity. Really great insight. We actually had a podcast not too long ago with chad moro. Who is the cto chief. Data officer at fulton bank and he actually made a great point about the human center city of systems especially of systems that depend on data because he was saying you know at the end of the day the data represents people represents what people are doing it represents their money represents their finances and those finances represent their retirement college savings. They're they're living right and you and you can never it's people's names treat data abstractly day. Sometimes it's really very critical and You know one of the great things. Of course you can. You mentioned that you'll be sharing a lot of these insights at our upcoming machine learning life cycle events so for our listeners. You may have heard this on previous episodes but of course if this is your first time We run these online free conferences. That are focused on some of the hottest topics. Ai machine learning and our objective is to help audience and help people take that next step and move their projects and forward We ran a huge data for ai. Conference back. In september twenty twenty twentieth thousands of attendees. It was amazing. Hundreds of of presenters actually well. Over one hundred plus presenters was was gigantic and we heard as people wanted to get that same sense of insight into what's happening with machine learning so we have the machine learning life cycle of that which talks about the full life cycle of machine learning from building the mall to managing an ops and govern insecurity and that is the live part of the event is january twenty sixth through twenty eight th twenty twenty one if you elect to register go to m. l. life cycle conference dot com. We'll have that in our show as well and Yeah we have some fantastic presenters in five topics and three tracks and our guest here. Terra their edge. They're they're actually doing. A session. called accelerating accelerate concept to value human centric design driven a lot of words there. There's a lot of terms of people may be familiar with some of them. They may not be. So maybe if you can. Can you give our listeners. A quick of what the session is about. And maybe some of the main questions and pain points that you're going to be addressing. Yeah thank you. Ron and actually just to right and it may come across as a laundry list of technical jargon and it's it's an i wanna make sure we can talk about it. In some of the audiences are going to be ingenious and audit back on both mock. You come from. Jean backgrounding ingenious with talking about the art of of humanity which the human centric design. What are we going to talk about. Is this first. Thing is gonna lay out the burning platform. We have seen the statistics enough data from gartner to idc that talks about the failures off a adoption. the data continues to show about eighty to ninety percent of machine learning data signs. Big data these initiatives famed to drive value. Because they're not getting adopted and if they're not getting a doctorate in driving value

Mark Persaud Fargo Sonic Automotive Amer Chad Moro Fulton Bank Human Center City Of Systems E Phnom Kathleen VW Amazon Nike Jean Backgrounding RON Gartner IDC
Tesla recalls 135,000 vehicles over touchscreen safety issue

WSJ What's News

00:29 sec | Last month

Tesla recalls 135,000 vehicles over touchscreen safety issue

"And tesla is recalling about one hundred. Thirty five thousand model s luxury sedans and model x. Sport utility vehicles over a touch screen failure. The national highway safety administration requested the action. Last month they say tesla's touchscreen can fail when memory chip runs out of storage capacity which could impact functions like defrosting turn signal functionality and driver assistance tesla said it would replace hardware for free as part of the recall but would do so in phases due to parts

Tesla National Highway Safety Admini
How to Be Human in the Age of AI with Ayanna Howard - #460

The TWIML AI Podcast

04:47 min | 1 d ago

How to Be Human in the Age of AI with Ayanna Howard - #460

"Your host. Sam charrington right. Everyone i am here with dr ianna. Howard is the dean of the college of engineering at the ohio state university as well as founder and cto of zyu- robotics. Ianna welcome to the tool a podcasts. Or i should say welcome back to welcome back. Just exciting lot has happened since we last talk. I think bit like three years. Almost to the day was february twenty eighteen. A lot has happened like alarm. world has happened. that's right and considering that twenty twenty was like ten years in and of itself it was like twelve years ago. Exactly i agree. I agree so when we last spoke and in fact at the time of this conversation you are chair of the school of interactive computing at georgia tech. But we're recording this on thursday and on monday us. Start your role at the ohio state. University congratulations first and foremost on that. Thank you. I'm actually really excited about this. Next transition in my life in my career and the things that i can do especially around engineering and you know access to the wonderful world of engineer awesome awesome. You know. i think i'd like to have you start with an introduction. You did an introduction and a little bit of background last time. We did this. I'd love for folks who haven't had an opportunity to get to know you to of your story and then we'll jump into the main topic of conversation for today and much of your research since the last time we spoke which is your recently published book sex race and robots how to be human in the age of ai up before we dive into that also a little bit about your background and story. Okay so I'm basically a hybrid engineer. Computer scientists but at the end of the day. I'm a robot. asus doesn't matter my title or my job. I'm a robotic. i design. Build the hardware as well as the algorithms to make my machines intelligent and interact with people and so started off as a robotic juicer. Nasa continued on as professor researcher at georgia tech and the key is people interacting with robots to really improve the quality of life whether it's health care whether it's education whether it's sending robots to glaciers to figure out the global warming climate change. Cleven do a lot of stock and again for those who are classically. Ai robotics is a lotta times. They call it embodied by ray which is a with the body beans robotics. I actually think of it as robotics. And then those virtual robotics which is why is virtual robots i e physical robots without the body. And so i think. I n robotics tightly couple. Bug lower botox is the lens through fix. Nice nice i was mentioning before we got rolling that this example that you gave in our last conversation with really centered on your research into the relationships between humans and robots is probably one of my most frequently quoted moments from the podcast. I tell people all the time about this story that you told about your research that shows that for whatever reasons humans look to robots as like authority figures in the example you gave was robots that ostensibly were set up to lead people out of burning buildings. You and your research program nadu really crazy. Things like bang up against walls for periods of time and humans would in many cases. Just sit there waiting for them to do the right thing. Because of that relationship that we have with robots in maybe expand on that a example briefly for folks. That didn't catch that interview. Which i recommend the go back and check out. Yeah it's actually to the day it's it was the most exciting a study that a group did mainly because it broke all of our hypothesis. So basically what happened was we. Were trying to examine what would happen. When are humans interacted with a robot that was faulty. That was the original premise and we wanted to create an environment. Where was you know. High risk time critical so that people were basically more reactive. They didn't have time to think. And so emergency evacuation fit that because we can set off the alarms. We can fill the building with smoke. And you're going to go into this rack with him.

Sam Charrington Dr Ianna ZYU Georgia Tech College Of Engineering Ohio State University Cleven Howard Ohio Asus Nasa