11 Microsoft’s Pete Bernard Explains AI IoT And Edge 5G Guys Tech Talks

Show video

hey real quick a quick word of thanks to today's sponsor vertex Innovations before we get started for over 17 years vertex has been building the nation's Wireless and Broadband networks providing project management network engineering and construction oversight are just some of the ways vertex helps their clients so if you're looking for more of a partner to help you with your wireless network designs construction implementation or operations reach out to vertex you can find them at vertex.us.com that's v-e-r-t-e-x-us.com welcome to the 5G guys podcast the premier resource for industry insiders and newcomers alike to explore anything and everything Wireless and telecommunication we discuss explain and explore all things Wireless technology so let's Dive Right In welcome your host Dan mcbaugh and Wayne Smith welcome back for another episode of the 5G guys I'm Wayne Smith and I'm joined by my co-host Dan mcball hey everyone thanks thanks and welcome back thanks for listening another episode of 5G guys before I introduce our guest and dive in today's topic Wayne and I are excited to announce a new offering for our listeners we found that many of our listeners have follow-up questions or suggestions for future show topics so what we've decided we're going to do is ask you go to 5guys.com and connect with us and we're offering anyone with show ideas or questions for a free consultation with Wayne and I so if you got a question a problem you're trying to figure out show ideas we'll do some a free hour with you we'll record it and possibly have you be on one of our future episodes so go to 5guys.com submit your questions and maybe you'll be on a future episode so now I'll get to today's episode I'm excited to introduce Our Guest today Pete Bernard Pete is an innovator at Microsoft where he drives Microsoft strategy for the intersection of 5G iot and AI now don't worry if those acronyms are already having you spinning just sit down have a drink we'll explain what they mean and you'll help us break all that down and make sense of it but this sort of innovation isn't new stuff since getting his degree in computer science from Boston University go Terriers I think yeah Terry is all right Pete's been driving product disruption and ecosystem Partnerships into the intelligent Edge Telecom and services arena for years in a Silicon Valley and at Microsoft and he also happens to be his own host of his own Microsoft podcast the iot in the card so uh Pete thanks for joining us excited to talk about 5G Ai iot and Edge Computing yeah great to be here great to be here Dan and Wayne appreciate it I've you know I know Dan we actually met I think you reminded me a few years ago back in the old days pre-covered days when we had face-to-face meetings but I think I spoke at a sort of like the ISC Western Wireless event or something and it was like a lot of the Telco infrastructure folks and I met a lot like cool folks including yourself and your colleagues and so that was pretty cool so I it's great to be back in touch and uh let's do a bunch of your episodes really educational so hopefully I can I can add to that in some small way you bet I have no doubt that you can so I guess before we dive into the weeds I've been known for having some some kind of corny analogies to explain technology but I thought maybe we could try a couple out on edge Computing and AI for our listeners maybe you can kind of give me some feedback if that works if it's a good way to help it break it down for the Layman so when I think about Edge Computing the analogy that comes to mind for me is if I think about the old cliche of the less left hand doesn't know what the right hand is doing right and we need our brain to sort of coordinate between our left hand and our right hand and in a cellular network or really any Telecom Network that's traditionally how networks have been right there's a core brain whether it's a switch Center or a data center that's doing all the processing and decision making and then just telling all the network elements so cell sites cell towers whatever the case might be access points think of those as our left hands and our right hands and tells them how to coordinate with each other how to do things together but when we start talking about what's coming out now with Edge Computing and 5G that's almost like our left hand having its own brain and our right hand having its own brain and they can almost communicate directly to each other with our our brain in the middle not slowing things down is that does that analogy work you think is that a good way to kind of explain Edge competing I think that's I think that's a good way of thinking about it I mean basically you know in the old days of meta Computing or iot like we used to have pretty pretty dumb sort of sensors and endpoints out there maybe like think of like a moisture sensor or some sort of sensor out there and it would send information up to the cloud and that was kind of you know the basics and then then we got kind of two-way communication right so then the cloud and the edge could sort of talk to each other and you could send information from the cloud to the edge as well as from the edge of the cloud and now you know the edge itself is adding more compute power the semiconductors on the edge that everything in your from your doorbell to your car uh actually I read this about 500 worth of semiconductors in today's cars as opposed to like 80 that used to be and so now the The Edge is becoming you know having quite a bit of compute power in and of itself right so these these endpoints you know millions and billions of endpoints are able to compute and you know do do things on their own without having to talk to the cloud or be reliant on other parts of the network and so um that's kind of the essence of what when people say Edge Edge is also relative term so it sort of depends on where you're standing where the edge is right uh and sometimes the edge is that moisture sensor out in the field in an agricultural kind of deployment or sometimes the edge is the edge of a uh like an att's like cellular network the edge of their data center in the Metro or Edge could be you know a server you know in in you know on a manufacturing floor so so wherever sort of the workloads are the farthest out is typically where the edge is but even in that you have sort of a tiered system going back up to the cloud so it can get a little complicated but at the end of the day what it really means is that the the all these compute points are getting more and more horsepower and they can do more and they can do more independently so that means you can you have a lot more options and and possibilities to hooking this stuff up right yeah and and as a matter of fact a good example of that that I heard from somebody that's talking to the other day that's a smaller wireless internet service provider that's deploying Wireless uh 5G networks for internet access at your house and they've got some cool stuff going on where they're pushing content to the edge where their edges basically the device that's in your house that's serving serving streaming video to your TV into your phone for example well that Netflix episode actually is not going to be residing on the cloud now it's going to be residing right on that device in your home and so now you're not using up the network at all when you're streaming it's just downloading what it thinks you want using things like AI which we'll get to next to figure out what it thinks you like right yeah yeah there's a lot of like uh the reason why people use Edge or Edge technology could be for that you know because they want to be really smart about how they use the network and when they use the network also it could be for privacy reasons maybe they want to keep the data local and on-prem like imagine like AI vision cameras you want to keep all that sort of secure on-prem and not back up into a cloud or a public Cloud uh or just low latency right so you want to be able to do things in you know single millisecond latency on a manufacturing floor for like worker safety things like that you don't want to be traveling all the way back up you know through the network all the way back up to some cloud and back down again to be able to do those sorts of things there's a lot of good reasons to use it and and I think what we're seeing is customers are getting really creative on how to use the tech for those that don't know what latency means think of latency not as speed but latency is how quickly you're asking for something to be done how quickly the network responds to what you're asking for right so my analogy of like a a super car being a 5G handset you know you hit the gas and it freaking goes right away right it responds immediately that's low latency so yes exactly and I've heard I had the guest on my show actually from Qualcomm who was saying like the single flap of a of a Bumblebee's Wings is about five milliseconds oh okay wow that gives you an idea it's pretty fast yeah Pete one of the things you you know you mentioned is workloads yeah and so what do you you know give us the definition of what a workload is because I think in a case study you know what is that what do you mean by that yeah so you know workloads and that's another one of those kind of industry terms but it's basically it's a it's a it's a set of code that needs to be run and that so that's a that could be a set of code that's you know doing some level of compute or analysis or you know or some some sort of compute so that workload can run that work right now I'm running on my Surface Book my beautiful Surface Book and I'm running a bunch of workloads on it to you know capture my voice and to display on the screen and to do all sorts of things uh but in some of these iot or Edge deployments those workloads may be running on the cloud on the public cloud like on Azure or those workloads could be running on the camera itself in the in the convenience store or it could be running on a Gateway box in the lunchroom of the convenience store so those workloads what's really exciting about that world is that you're now kind of Distributing workloads is Right Between The Edge and the cloud and and you know on the plus side it's like a lot of great flexibility for customers and especially commercial customers to design some really great Solutions I guess on the downside it makes things you know kind of infinitely more complicated so that's the negative so there was a couple other acronyms we threw out along the way here um you know we talked about iot and you touched on what iotism internet of things and the best way to describe that to people is it's not individuals it's not you and I using the network it's our ring doorbell it's that sensor in a factory it's and by last I heard and maybe you've got more recent numbers I've heard there's there's something like 10 times more iot devices in the world right now than there are human beings that may not be the right number but it's huge right I probably have 50 more I have probably 50 iot things right here in my office and it's just me so there's a 50 to 1 ratio right there but I wouldn't be surprised yeah I mean everything you know and so many things are internet connected today and and in the consumer world right you got doorbells and you've got your your Bluetooth scale and your Fitbit and your phones and all kinds of things and in your car you know so many things that are happening in your car that are that are connected to the cloud too so it's you know that's one of the benefits of 5G by the way is the high density capabilities of 5G is the ability to handle you know billions of connections and much higher density for a given Tower and deployment than uh traditional like 4G which is pretty exciting for us because we can then you know a farm can outfit you know lots of connected sensors uh and robots and and other detection systems you know in a huge you know multi-acre deployment using 5G or you could do that in an airport or you know shipping containers all kinds of things and they can all communicate on the network without any you know you know issues around latency or or you know other confusion in the network all right so let's do one more let's talk about AI so artificial intelligence right I'll let you take this one how would you break down AI for Laura Layman and and you know is it is it the Terminator cyborg that uh you know the world's gonna be taken over by robots or right yeah not yet so that's good news but um now ai and ai's been around forever you know not forever but you know decades and and in a computer science world I mean they've been experimenting with AI for a long time what's happened probably over the last decade is that the amount of compute power and going back to semiconductors and things like that have gotten so uh significant that you're able to run these AI models you're able to train models AI models which is just code you know to recognize things and to and to to be able to do things and so there's there isn't because of the compute and the storage and the network capabilities you're able to run artificial intelligence in much more profound ways than used to be I mean one simple example of AI would be you know when you go to Netflix and you watch a whole bunch of mobster shows or whatever they recommend a monster show hey you might like this mobster show that actually is a form of AI very simple you know machine learning type of AI more advanced AI would be well I have a camera pointed to my production line and it's visually inspecting you know equipment that's being manufactured and it's looking for defects and so I've trained it to know what a a good part looks like and I've trained it to know what a bad part looks like and it can do all kinds of on the Fly you know anomaly detection through Visual AI so so AI is really just that that ability to to do to kind of decision making uh based on training and based on input and now now we have the compute horsepower and other infrastructure horsepower to do it at a much more significant level not quite Terminator yet right the the the analogy I've given people is you know when I went to engineering school and people asked what did you learn in engineering school that you use in your career and I said well I learned how to learn and that's really what we're doing right we're teaching these compute platforms how to learn and then they go off and learn on their own without the old way of us them them just doing what we tell them to do right right yeah there's training and then there's inferencing and that's kind of another little breakdown in terms of of workloads going back to that term you know how do you train the models to recognize you know license plate recognition that's a big one in parking lots and things like that or how many you know customers or where in the store at certain times of the day and looking at end caps and things like that so you're training using lots of data you know hundreds of thousands of images and then there's inferencing which is now that you have the model you want to run the model on some piece of equipment to then you know recognize license plate so it's taking the model and it's recognizing it so that's a little bit less horsepower in that in that kind of workload but yeah no the AI you know if you start combining some of these Technologies together things like 5G which I'm sure you're familiar with and Ai and Edge you kind of get this really interesting Venn diagram overlap right and that that piece in the middle is pretty transformative stuff and that's what we're seeing with a lot of our customers working with Microsoft on like how do we accelerate our digital transformation with this kind of now interesting new combination of really fundamental technology yeah wow you know a part of that what you need to make AI where work is a lot of data correct I mean the more data the better the AI is that exactly you get that yeah the model is trained you know the more data you can feed into it the better the model is trained to be the more accurate it's going to be okay is that is that pipe in order to get your data from one place to another fast enough or is that is that a true assumption or no yeah I would say you know there's getting back to sort of training and inferencing you know a lot of model training is happening on the cloud today so you're using some pretty big compute horsepower to your training like you would not do a lot of training on your on your local PC it would take a long time you know so really what you want to leverage your your your Azure or your AWS or gcp or something like that to do your training and you're feeding in a ton of data you know a ton of data we have a partner that does a lot of lumber stuff and so they can recognize you know discrepancies in the lumber pattern or whatever and so they've got you know millions of images of lumber that they put into the cloud and crunch through and create model to do that so typically you're not using 5G for the Ingress of the data for training you're more of you know using 5G to communicate you know with those cameras you know over over the network to to do the inferencing maybe of the of the you know actual real world using those AI models if that makes sense yeah so 5G is basically allowing us to access that AI after it's been trained and take advantage of it yeah and so you know one of the interesting things about it is if you have ai vision and you know you're doing um over 5G you can do some interesting things where you can recognize instead of streaming all the data back to the network back up to the cloud which could be quite expensive you could actually do a lot of inferencing on the edge and actually use very little data and use you know what we call lpwa or low power wireless access and use very little data and only send sort of What's called the metadata back to the cloud for action and to say you know we we have the camera has now recognized that someone has fallen on the floor and then you can actually start streaming the data because of that action back to the cloud flat as opposed to streaming all the data so so you know when you start combining things like the 5G and Edge and AI you can do some really smart things and use the network in a much smarter way I think than if you had to just sort of just pump data to the cloud and look at it later right it's all it can all happen very real time yeah I know a good a good application for AI specifically in a cellular network that actually the idea was born 15 20 years ago it just is taken until now to have the compute powder and the edge technology that to enable it is Network operators now can have cell sites that can self-optimize and they can say Hey you know I know where traffic's located and I know hey this cell site's overloaded and it can't handle all the traffic but the cell site next door can handle more so I'll I'll use AI to shed traffic off of this cell site onto that cell site and the Network's doing it on its own without someone telling it to do it in the past we'd have to send a technician out there climb the tower physically mechanically down tilt the antenna so that the foot footprint got smaller so the site next to it would take it and now that's all happening through AI yeah and so you know we have a lot of customers including telcos we've made some announcements with ATT and BT recently where they're using Azure ML and they're using a lot of azure Services big cloud services to optimize their networks and move those workloads uh workload word again onto Azure and onto the public Cloud to get a lot more kind of scalability and so yeah we're seeing AI applied back in the network itself like you said for like routing and traffic optimization and then you know my area is more of like outside the data center where we're using Ai and Edge for some of these interesting kind of scenarios as well but yeah it seems like on a daily basis and I guess that's one of the benefits of working at Microsoft we work with so many Fortune 500 customers that come to us and they have some really kind of gnarly problems that they're trying to solve and so part of what my team does is help them think about how they could use 5G and nji to solve some of those problems so it's kind of a never-ending list of interesting uh challenges I would say do you have a great case study that you know you can kind of share with the listeners about and practical application you know not not customer wise but more or less you know just sure yeah one of the interesting things that's emerging originally hadn't thought of this I mean when they think about like 5G and Aji you think about like deploy anywhere right so you can peel and stick a camera and and you know it can be in a moving vehicle or it could be out in the middle of a field and whatever so that's pretty straightforward so we had that and then we also heard a lot about deployment costs and smart cities or critical infrastructure deployments if you can have a device out there that doesn't write any cabling for example if you can be self-powered and 5G connected then you're reducing your deployment cost by like 90 right because you don't have to you know Jackhammer the sidewalk and lay new cable and stuff like that so that was interesting and then the third thing that was kind of a surprise was we heard from a lot of retailers and other folks with storefronts and things that they did not want to have a lot of heavy Edge Equipment in their facility I don't know if you've ever encountered in real life like a server server blade like you would see in a data center you know when they run they kind of like sound like a hive of bees and so it's not not the kind of thing you would want to stick in your lunchroom um and run your store off of so that we had a lot of customers saying hey can I use 5G can I can I have these servers co-located somewhere else you know a few miles away and I'll use 5G to go from like cameras in my store to these this heavy Edge Equipment like you know a few blocks away or a few miles away and maybe I can share that compute across multiple stores in the neighborhood and so that was a really neat way of leveraging like a low latency 5G connection and also getting the benefit of not having a bunch of heavy age equipment in your store which by the way is kind of like a you know someone could steal it someone could kick the plug out it also sounds like a high the bees so it's not good so so that was kind of a neat a neat scenario that we hadn't really thought of but our customers came to us and we're like that's you know that would be fantastic we could do that so we're now working with with a lot of our Telco partners and and Hardware Partners to to do just that which is pretty cool yeah well and and I think the other thing that like that scenario described what it does is it creates this what's called democratization of Technology right it makes this advanced technology more accessible to the larger world where you know prior to these these convergences of Technologies it took a lot of money and capital and resources to take advantage of Technology now with all of this like you said I don't have to have large compute hardware for me to do that I can have a service where I'm paying a marginal amount maybe monthly to have access to that compute power without having to buy it I think a good example that's happening right now is Mobile 5G gaming and this will be something this is an area where 5G is really going to start benefiting the consumer is you have almost I think I heard you've got more Gamers now on mobile devices than you have on you know Xboxes and PlayStations and PCs because because they can now leverage 5G Edge Computing to do gaming from their phone or from their Chromebook or from whatever you know and so it's now people that couldn't afford to have a gaming platform now can use 5G and they don't have to have internet service at their house anymore either by the way it's a Mobile 5G service um yeah yeah I think we're seeing a lot of folks you know and I think on the business world what that means is like small business you know other startups and stuff can you know like you said in the old days you'd have to drop in half a dozen different boxes into the back room and have someone know how to configure them and run them and all that good stuff and buy all this equipment and now yeah I mean you know you can sort of get all that stuff on a monthly basis and not install pretty much anything uh except the Edge Edge pieces and uh you you know that then like you mentioned also gaming and um you know we have lots of folks with our Xbox service using that via the cloud as well and you know that means you just kind of play anywhere right you can certainly you'd want you to buy an Xbox and we should all buy Xboxes and sit at home and play Xbox but you know can't always sit at home play Xbox sometimes you're on the bus or you know you're somewhere where you want to use so yeah being able to extend that out and have that horsepower in the cloud and leverage a high performance network is a fantastic thing yeah I read an article before using 5G for gaming mobile Gamers one out of three times ended their gaming session because latency made the game impossible to play right you're a first shooter gaming you know kind of thing and and you got shot because you couldn't pull the trigger on the game fast enough so yeah so now with 5G now all of a sudden that goes away that's where that latency Improvement comes in it responds faster right yeah and also you know a lot of these telcos are reaching a lot of capacity frankly there's just so many people with phones and other equipment connected to these networks that it makes some of these kind of high performance experiences not really work too well unless you're on 5G so it's getting back to that kind of high density high density benefit is that you know you can get a lot of these things connected at the same time over a single Tower a lot more of them over a single Tower than you used to be able to do got it so I think there's one other term that I realized we've been throwing around in our in our talk today that I want to make sure we clarify for folks and it may actually be a good clarification for Wayne and I as well what's the cloud the cloud yes I remember actually a funny story so I have three kids and uh they've all grown up here we've been up here in Billy Washington for like 16 years and I remember sitting around the dinner table and my daughter who is I don't know she was in grade school or something she asked about the cloud and she was like this you know she didn't understand how the computers could be in the cloud like that didn't make any sense to her like up in the sky you know and so so first rule number one is it's actually not in a cloud it's actually on the ground so and they're in these huge data centers and so you know you can think of the cloud a big collection of compute in these huge data centers in these data centers if you go on search on Microsoft we I think on YouTube we have some cool video tours of some of these data centers and we have something like 50 or 60 of them around the world and they're just like you know a dozen football fields of concrete and and cabling and just racks and racks of servers and you know each one of those serves a different part of the part of the world and and that's that's what is collectively known as the cloud and so you know we have Azure from Microsoft there's AWS there's you know Google has a cloud Oracle IBM um Azure is is a huge one I think the biggest and uh it's just a collection of compute resources that you you as a as a developer would pay for on a as use basis right so you're based on how many VMS you want and what kind of services you want you pay as you use it instead of paying for Hardware that you would install locally and so that's kind of that's what when people say the cloud that's effectively what they mean perfect yeah so before the before the the quote-unquote cloud companies had huge data closets with their own servers which you called earlier on-prem on-premise Computing on-prem yes right and so Cloud allowed them to no longer have to fill a closet with a bunch of servers yeah and you know and we still there's a lot of hybrid stuff going on which makes a lot of sense too so you have some compute on on-prem but not a full data center actually when I was in school back in Boston I had a summer job at Prime computer who's long since gone out of business and truth be told my cousin worked there so he got me a job for the summer and you know one of the benefits and I love Boston but sometimes in the summer it gets really hot the benefit of working at Prime computer in the summer was that the data center was super air conditioned and so I would spend all my time there in this beautiful air-conditioned room and like 64 degrees or something like that but yeah you had to you know use a ton of power and resources to cool these machines that ran at a fraction of the compute right in these Big Data Centers and they're still around and you know we're helping a lot of customers basically decommission that stuff and get to a much more energy efficient more carbon neutral footprint we use a lot of hydro and a lot of data centers for Azure a lot more efficient use of power and a lot more elastic use of the resources so you know as your business grows or as your need grow you just get more compute you know click click on the screen as opposed to actually plugging stuff in no I think a question that came up in my search you know by having you as a guest is that everyone's asking you know why is Microsoft getting into 5G as a long-term plan yeah do you want to elaborate on that you know I mean sure I mean we have you know if you think about it we have a lot of different businesses obviously the windows business Xbox business we talked about Azure uh a lot of the Microsoft 365 business I mean all this stuff has to kind of connect over a network right the age of the disconnected worker you know with their laptop that doesn't connect anything is kind of over so uh you know so we have to be connected and so you know high performance networks is a real strategic thing for us if you think about how did the world change back in 99 and 2000 when Wi-Fi became mainstream right all of a sudden you could take your laptop anywhere you could take it to your meeting room and look check your email during the meeting you can go to Starbucks and work there so it was kind of a huge change for uh workers around the world and that affected Us in terms of how we had you know we had office and windows and things like that and I think we think 5G is kind of similar it's one of those once in in a decade once in a couple of decade transformative technologies that will impact how people work how people play you know we acquired a meta switch and affirmed past year or two fantastic companies uh combine them with Azure networking we have a group called Azure for operators and I mentioned the announcements we had with ATT and BT recently you know really leaning in and helping them move their workloads onto Azure digitally transform their networks and their businesses so it's it's affecting not only our big customers in terms of their own digital transformation but just everyday folks that are trying to get stuff done uh they can do it over 5G and some of the Microsoft products that we have then everything will just work faster better you know and and and with the edge AI combination now we're you know creating some new capabilities it's just weren't possible before without 5G so that's really exciting and now we get you know energy companies and transportation companies healthcare companies are all like well I can do a private 5G Network here I can have a you know we have someone who's working on a network 5G Network on a train car that can roll out and you know be anywhere and you know on a construction site and all of a sudden there's like 5G Network there so it's really a big enabler for a lot of businesses and lots of customers that we have awesome awesome go ahead Dan what did you have well all I was gonna to say is uh maybe Pete you can get us a link to that YouTube video we'll put it in the show notes for our listeners on the website then they can kind of see what the Azure cloud looks like yeah there you go yeah it's uh pretty impressive stuff and there's like you know thousands of people that plan that stuff and run that stuff and uh it's really impressive yeah definitely um thanks for joining us super fascinating you've you're touching on things that'll to your point where it'll transform the future and thanks for sharing your wisdom yeah no problem and I was I would also suggest if anyone's interested in Edge AI we actually have a product called Azure percept that we introduced back in March and if you go to https forward slash get Azure percept uh you can get a dev kit there it's pretty cool within like around about 30 minutes you can plug it in and start doing like AI vision and you can do some audio and you can do all kinds of things it comes with a lot of AI models built in so if you want to kind of experiment with with Edge AI yourself that's a great way to kind of get into it and you can show it to your boss get promoted be a hero you know all kinds of cool things um but it's it's pretty cool so Azure percept is definitely something that people can get Hands-On after the show maybe and experiment a little bit awesome we'll drop that in the show notes as well for everyone so go check that out so yeah thanks thanks again Pete and is there any other any other ways people can connect with you personally or with other things that are going on at Microsoft yeah well I'm on Twitter digital dad is my handle so you can follow me there and uh pretty regular contributor to that whole situation and then I'm on LinkedIn as well you can reach me there so Pete Bernard and yeah look forward to uh more exciting things coming down the pike likewise thanks for joining us and thanks again to everyone for listening uh remember to subscribe or follow of the show go to 5guys.com to engage more with us we definitely look forward to everyone taking advantage of our offer for a free consultation so give us your show ideas or questions you want to have solved when I'll give you a free consultation and maybe make it a future episode so until next time thanks for joining thanks thanks everybody thanks for listening listening to the 5G guys for more reasons website at 5guys.com if you enjoyed this

episode be sure to hit that follow button and share this episode with your friends and family [Music]

2023-03-08

Show video