>> Hi I'm Peter Burris and welcome once again to Wikibons's weekly research meeting on theCUBE. (funky electronic music) This week we're going to discuss something that we actually believe is extremely important. And if you listen to the recent press announcements this week from Deli MC, the industry increasingly is starting to believe is important. And that is, how are we going to build systems that are dependent upon what happens at the edge? The past 10 years have been dominated about the cloud. How are we going to build things in the cloud? How are we going to get data to the cloud? How are we going to integrate things in the cloud? While all those questions remain very relevant, increasingly, the technology's becoming available, the systems and the design elements are becoming available, and the expertise is now more easily bought together so that we can start attacking some extremely complex problems at the edge. A great example of that is the popular notion of what's happening with automated driving. That is a clear example of huge design requirements at the edge. Now to understand these issues, we have to be able to generalize certain attributes of the differences in the resources, whether they be hardware or software, but increasingly, especially from a digital business transformation standpoint, the differences in the characteristics of the data. And that's what we're going to talk about this week. How do different types of data, data that's generated at the edge, data that's generated elsewhere, going to inform decisions about the classes of infrastructure that we're going to have to build and support as we move forward with this transformation that's taking place in the industry. So to kick it off, Neil Raden I want to turn to you. What are some of those key data differences and what taxonomically do we regard as what we call primary, secondary, and tertiary data? Neil. >> Well, primary data come in from sensors. It's a little bit different than anything we've ever seen in terms of doing analytics. Now I know that operational systems do pick up primary data, credit card transactions, something like that. But, scanner data, not scanner data, I mean sensor data is really designed for analysis. It's not designed for record keeping. And because it's designed for analysis, we have to have a different way of treating it than we do other things. If you think about a data lake, everything that falls into that data lake has come from somewhere else, it's been used for something else. But this data is fresh, and that requires that we really have to treat it carefully. Now, the retention and stewardship of that requires a lot of thought. And I don't think industry has really thought of that through a great deal. But look, sensor data is not new, it's been around for a long time. But what's different now is the volume and the lack of latency in it. But any organization that wants to get involved in it really needs to be thinking about what's the business purpose of it. If you're just going into, IOT as we call it generically, to save a few bucks you might as well not bother. It really is something that will change your organization. Now, what do we do with this data is a real problem because for the most part, these senses are going to be remote, and there's going to be a lot of, that means they're going to generate a lot of data. So what do we do with it? Do we reduce it at the sight? That's been one suggestion. There's an issue that any model for reduction could conceivably lose data that may be important somewhere down the line. Can the data be reconstituted through metadata or some sort of reverse algorithms? You know, perhaps. Those are the things we really need to think about. My humble opinion is the software and the devices need to be a single unit. And for the most part, they need to be designed by vendors, not by individual ITs. >> So David Floyer, let's pick up on that. Software and devices as single unit, designed more by vendors who have specific demand expertise, turn into solutions and present it to business. What do you think? >> Absolutely, I completely concur with that. The initial attempts to using the sensors and connecting to the sensors were very simple things like for example, the nest, the thermostats. And that's worked very well. But if you look at it over time, the processing for that has gone into the home, into your Apple TV device or your Alexa or whatever it is. So, that's coming down and now it's getting even closer to the edge. In the future, our proposition is that it will get even closer and then those will put together solutions, all types of solutions that are appropriate to the edge that will be taking not just one sensor but multiple sensors, collecting that data together, just like in the autonomous car for example where you take the lidars and the radars and the cameras etcetera. We'll be taking that data, we'll be analyzing it, and we'll be making decisions based on that data at the edge. And vendors are going to play a crucial role in providing these solutions to IT and to the OT and to many other parts. And a large value will be in their expertise that they will develop in this area. >> So as a rule of thumb, when I was growing up and learned to drive, I was told always keep five car lengths between you and whatever's in front of you at whatever speed you're traveling. What you just described David is that there will be sensors and there will be processing that takes place in that automated car that isn't using that type of rule of thumb but know something about tire temperature, and therefore the coefficient of friction on the tires, know something about the brakes, knows what the stopping power needs to be at the speed and therefore what buffer needs to be between it and whatever else is around it. >> Absolutely. >> This is no longer a rule of thumb, this is physics and deep understanding of what it's going to require to stop that car. >> And on top of that, what you'll also want to know, outside from your car is, what type of car is in front of you? Is that an autonomous car, or is that somebody being driven bye Peter? In which case, you have 10 lengths behind you. >> But that's not going to be primary data. Is that what we mean by secondary data? >> No, that's still primary because you're going to set up a connection between you and that other car. That car is going to tell you I'm primary to you, that's primary data. >> Here's what I mean, correct use primary data but, from a standpoint of that the car in that case is submitting a signal, right? So even though to your car it's primary data, but one of the things from a design standpoint that's interesting, is that car is now transmitting a digital signal about it's state that's relevant to you so that you can combine that >> Correct. inside effectively, a gateway inside your car. >> Yes. >> So there's external information that is in fact digital coming in, combining with the sensors about what's happening in your car. Have I got that right? >> Absolutely. That to me is a sort of sengrey one, then you've got the tertiary data which is the big picture about the traffic conditions >> Routes. and the weather and the routes and that sort of thing which is at that much higher cloud level, yes. So David Vellante, we always have to make sure as we have these conversations. We've talked a bit about this data, we've talked a little bit about the classes of work that's going to be performed at the different levels. How do we ensure that we sustain the business problem in this conversation? >> So, I mean I think Wikibon's done some really good work on describing what this sort of data model looks like from edge devices where you have primary data, the gateways where you're doing aggregated data in the cloud where maybe the serious modeling occurs. And my assertion would be is that the technology to support that elongating and increasingly distributed data model has been maturing for a decade and the real customer challenge is not just technical, it's really understanding a number of factors and I'll name some. Where in the distributed data value chain are you going to differentiate? And how does the data that you're capturing in that data pipeline contribute to monetization? What are the data sources, who has access to that data, how do you trust that data, and interpret it, and act on it with confidence? There are significant IP ownership in data protection issues. Who owns the data? Is it the device manufacturer, is it the factory, etcetera. What's the business model that's going to allow you to succeed? What skill sets are required to win? And really importantly, what's the shape of the ecosystem that needs to form to go to market and succeed? These are the things that I think customers are really struggling with that I talk to. >> Now, the one thing I'd add to that and I want to come back to it is the idea that, and who is ultimately bonding the solution because this is going to end up in a court of law. But let's come to this IP issue, George. Let's talk about how local data is going to be, is going to enter into the flow of analytics, and that question of who owns data, because that's important and then have the question about some of the ramifications and liabilities associated with this. >> Okay well, just on the IP protection and the idea that a vendor has to take sort of whole product responsibility for the solution. That vendor is probably going to be dealing with multiple competitors when they're sort of enabling say, self-driving car or other, you know edge, or smaller devices. The key thing is that, a vendor will say, you know, the customer keeps their data and the customer gets the insights from that data. But that data is informing in the middle a black box, an analytic black box. It's flowing through it, that's where the insights come out, on the other side. But the data changes that black box as it flows through it. So, that is something where, you know, when the vendor provides a whole solution to Mercedes, that solution will be better when they come around to BMW. And the customers should make sure that what BMW gets the benefit of, goes back to Mercedes. That's on the IP thing. I want to add one more thing on the tertiary side which is, when you're close to the edge, it's much more data intensive. When we've talked about the reduction in data and the real-time analytics, at the tertiary level it's going to be more where time is a bigger factor and you're essentially running a simulation, it's more compute intensive. And so you're doing optimizations of the model and those flow back as context to inform both the gateway and the edge. >> David Floyer I want to turn it to you. So we've talked a little bit about the characteristics of the data, great list of Dave Vellante about some of the business considerations, we will get very quickly in a second to some of the liability issues cause that's going to be important. But take us through how, which George just said about the tertiary elements. Now we've got all the data laid out, how is that going to map to the classes of devices? And we'll then talk a bit about some of the impacts on the industry. What's it going to look like? >> So if we take the primary edge first, and you take that as a unit, you'll have a number of senses within that. >> So just released, this is data about the real world that's coming into the system to be processed? >> Yes. So it'll have, for example, cameras. If we take a simple example of making sure that bad people don't get into your site. You'll have a camera there which will be facial recognition. They'll have a badge of some sort, so you'll read that badge, you may want to take their weight, you may want to have a infrared sensor on them so that you can tell their exact distance. So, a whole set of sensors that the vendor will put together for the job of insuring you don't get bad guys in there. And what you're insuring is that bad guys don't get in there, that's obviously one, very important, and also, that you don't go and- >> Stop good guys from going in. stop good guys from going in there. So those are the two characteristics >> The false-positive problem. the false-positives. Those are the two things you're trying to design that- >> At the primary edge. at the primary edge. And there's a mass amount of data going into that, which is only going to be reduced to very, very little data coming up to the next level which is this guy came here, this was his characteristics, he didn't look well today, maybe you should see a nurse, or whatever other information you can gather from that will go up to that secondary level, and then that'll also be a record of to HR maybe, about who has arrived there or what time they arrived, to the manufacturing systems about who is there and who has those skills to do a particular job. There are multiple uses of that data which can then be used for differentiation for whatever else from that secondary layer into local systems and then equally they can be pushed up to the higher level which is, how much power should be generating today, what are the higher levels. >> We now have 4,000 people in the building, air condition therefore is going to look like this, or, it could be combined with other types of data like over time we're going to need new capacity, or payroll, or whatever else it might be. >> And each level will have its own type of AI. So you've got AI at the edge, which is to produce a specific result, and then there's AI to optimize at the secondary level and then the AI optimize bigger things at the tertiary level. >> So we're going to talk more about some of the AI next week, but for right now we're talking about classes of devices that are high performance, high bandwidth, cheap, constrained, proximate to the event. >> Yep. >> Gateways that are capable of taking that information and start to synthesize it for the business, for other business types of things, and then tertiary systems, true private cloud for example, although we may have very sizable things at the gateway as well, >> There will be true private clouds. that are capable of integrating data in a more broad way. What's the impact in the industry? Are we going to see IT firms roll in and control this sweeping, (man chuckles) as Neil said, trillions of new devices. Is this all going to be intel? Is it all going to be, you know, looking like clients and PCs? >> My strong advice is, that the devices themselves will be done by extreme specialists in those areas that they will need a set of very deep technology understanding of the devices themselves, the senses themselves, the AI software relevant to that. Those are the people that are going to make money in that area. And you're much better off partnering with those people and letting them solve the problems, and you solve, as Dave said earlier, the ones that can differentiate you within your processes, within your business. So yes, leave that to other people is my strong advice. And from an IT's point of view, just don't do it yourself. >> Well the gateway's, sound like you're suggesting, the gateway is where that boundary's going to be. >> Yes. That's where the boundary is. >> And the IT technologies may increasingly go down to the edge, but it's not clear that the IT vendor expertise goes down to the edge >> Correct. at the same degree. >> Correct. >> So, Neil let's come back to you. When we think about this arrangement of data, you know, how the use cases are going to play out, and where the vendors are, we still have to address this fundamental challenge that Dave Vellante bought up. Who's going to end up being responsible for this? Now you've worked in insurance, what does that mean from an overall business standpoint? What kinds of failure weights are we going to accommodate? How is this going to play out? What do you think? >> Well, I'd like to point out that I worked in insurance 30 years ago. (men chuckling) >> Male Voice: I didn't want to date ya Neil. (men chuckling) >> Yeah the old reliable life insurance company. Anyway, one of the things David was just discussing sounded a lot to me like complex event processing. And I'm wondering where the logical location event needs to be, because it needs some prior data to do CEP, you have to have something to compare it against. But if you're pushing it all back to the tertiary level, there's going to be a lot of latency. And the whole idea was CEP was, you know, right now. So, that I'm a little curious about. But I'm sorry, what was your question? >> Well no, let's address that. So CEP David, I agree. But I don't want to turn this into a general discussion and CEP. It's got its own set of issues. >> It's clear there have got to be complex models created. And those are going to be created in a large environment, almost certainly in a tertiary type environment. And those are going to be created by the vendors of those particular problem solvers at the primary edge. To a large extent, they're going to provide solutions in that area. And they're going to have to update those. And so, they are going to have to have lots and lots of test data for themselves and maybe some companies will provide test data if it's convenient for those, for a fee or whatever it is, to those vendors. But the primary model itself is going to be in the tertiary level, and that's going to be pushed down to the primary level itself. >> I'm going to make an assertion here that the, the way I think about this Neil is that the data coming off at the primary level is going to be the sensor data, the sensor said it was good. Then that is recorded as an event, we let somebody in the building. And that's going to be a key feature of what happens at the secondary level. I think a lot of complex processing is likely to end up at that secondary level. >> Absolutely. >> Then the data gets pushed up to the tertiary level and it becomes part of an overall social understanding of the business, it's behavior data. So increasingly, what did we do as a consequence of letting this person in the building? Oh we tried to stop him. That's going to be more of the behavioral data that ends up at the tertiary level, will still do complex event processing there. It's going to be interesting to see whether or not we end up with CEP directly in the sensor tower. Might under certain circumstances, that's a cost question though. So let me now turn it in the last few minutes here Neil back to you. At the end of the day, we've seen for years the question of how much security is enough security? And businesses said, "Oh I want to be 100% secure." And sometimes see-so said "We got that. You gave me the money, we've now made you 100% secure." But we know it's not true. Same thing is going to exist here. How much fidelity is enough fidelity down at the edge? How do we ensure that business decisions can be translated into design decisions that lead to an appropriate and optimized overall approach to the way the system operates? From a business standpoint back, what types of conversations are going to take place in the boardroom that the rest of the organization's going to have to translate into design decisions? >> You know, boy, bad actors are going to be bad actors. I don't think you can do anything to eliminate it. The best you can do is use the best processes and the best techniques to keep it from happening and hope for the best. I'm sorry, that's all I can really say about it. >> There's quite a lot of work going on at the moment from Arm, in particular. They've got a security device image ability. So, there's a lot of work going on in that very space. It's obviously interesting from an IT perspective is how do you link the different security systems, both from an Arm point of view and then from a X86 as you go further up the chain. How are they going to be controlled and how's that going to be managed? That's going to be a big IT issue. >> Yeah, I think the transmission is the weak point. >> Male Voice: What do you mean by that Neil? >> Well the data has to flow across networks, that would be the easiest place for someone to intercept it and, you know, and do something nefarious. >> Right yeah, so that's purely in a security thing. I was trying to use that as an analogy. So, at the end of the day, the business is going to have to decide how much data do we have to capture off the edge to ensure that we have the kinds of models we want, so that we can realize the specificity of actions and behaviors that we want in our business? That's partly a technology question, partly a cost question. Different sensors are able to operate at different speeds for example. But ultimately, we have to be able to bring those, that list of decisions or business issues that Dave Vellante raised, down to some of the design questions. But it's not going to be throw a $300 micro processor everything. There's going to be very, very concrete decisions that have to take place. So, George do you agree with that? >> Yes, two issues though. One, there's the existing devices that can't get re-instrumented, that they already have their software, hardware stack. >> There's a legacy in place? >> Yes. But there's another thing which is, some of the most advanced research that's been going on that produced much of today's distributed computing and big data infrastructure, like the Berkeley Analytics lab, and say their contributions spark in related technologies. They're saying we have to throw everything out and start over for secure real-time systems. That you have to build from hardware all the way up. In other words, you're starting from the sand to re-think something that's secure and real-time that you can't layer it on. >> So very quickly David, that's a great point George. Building on what George has said very quickly, the primary responsibility for bonding the behavior or the attributes of these devices are going to be with the vendor. >> Of creating the solution? >> Correct. >> That's going to be the primary responsibility. But obviously from an IT point of view, you need to make sure that that device is doing the job that's important for your business, not too much, not too little, is doing that job, and that you are able to collect the necessary data from it that is going to be of value to you. So that's a question of qualification of the devices themselves. >> Alright so, David Vellante, Neil Raden, David Floyer, George Gilbert, action item round. I want one action item from you guys from this conversation. Keep it quick, keep it short, keep it to the point. David Floyer, what's your action item? >> So my action item is don't go into areas that you don't need to. You do not need to become experts, IT in general does not need to become experts at the edge itself. Rely on partners, rely on vendors to do that unless of course you're one of those vendors. In which case, you'll need very, very deep knowledge. >> Or you choose that that's where you're value stream your differentiations is going to be which means you just became one of those values. >> Yes, exactly. >> George Gilbert. >> I would build on that and I would say that if you look at the skills required to build these full stack solutions, there's data science, there's application development, there's the analytics. Very few of those solutions are going to have skills all in one company. So the go-to market model for building these is going to be something that, at least at this point in time, we're going to have to look to like combinations like IBM working with sort of supply chain masters. >> Good. Neil Raden, action item. >> The question is not necessarily one of technology because that's going to evolve. But I think as an organization, you need to look at it from this end which is, would employing this create a new business opportunity for us? Something we're not already doing. Or number two, change our operations in some significant way. Or number three, you know, the old red queen thing. We have to do it to keep up with the competition. >> Male Voice: David Vellante, action item. >> Okay well look, at the risk of sounding trite, you got to start the planning process from the customer on in, and so often people don't. You got to understand where you're going to add value for customers and constructing and external and internal ecosystem that can really juice that value creation. >> Alright, fantastic guys. So let me quickly summarize. This week on the Wikibon Friday research meeting in the cube, we discussed a new way of thinking about data characteristics that will inform system design and a business value that's created. We observe that data is not all the same when we think about these very complex, highly distributed, and decentralized systems that we're going to build. That there's a difference between primary data, secondary data, and tertiary data. Primary data is data that is generated from real world events or measurements and then turned into signals that can be acted upon very proximate to that real world set of conditions. A lot of sensors will be there, a lot of processing will be moved down there, and a lot of actuators and actions will take place without referencing other locations within the cloud. However, we will see circumstances where the events that are taken, or the decisions that are taken on those vents, will be captured in some sort of secondary tier that will then record something about the characteristics of the actions and events that were taken, and then summarized and then pushed up to a tertiary tier where that data can then be further integrated in other attributes and elements of the business. The technology to do this is broadly available but not universally successfully applied. We expect to see a lot of new combinations of edge-related device to work with primary data. That is going to be a combination of currently successful firms in the OT or operational technology world, most likely in partnership with a lot of other vendors that have demonstrated significant expertise and understanding the problems, especially the business problems, associated with the fidelity of what happens at the edge. The IT industry is going to approach very aggressively and very close to this at that secondary level, through gateways and other types of technologies. And even though we'll see IT technology continue to move down to the primary level, it's not clear exactly how vendors will be able to follow that. More likely, we'll see the adoption of IT approaches to doing things at the primary level by vendors that have the main expertise in how that level works. We will however see significantly interesting true private cloud and public cloud data end up from the tertiary level end up with a whole new sets of systems that are going to be very important from an administration and management standpoint because they have to work within the context of the fidelity of this overall system together. The final point we want to make is that these are not technology problems by themselves. While significant technology problems are on the horizon about how we think about handling this distribution of data, managing it appropriately, our ability, ultimately, to present the appropriate authority at different levels within that distributive fabric to ensure the proper working condition in a way that nonetheless we can recreate if we need to. But these are, at bottom, fundamentally business problems. They're business problems related to who owns the intellectual property that's being created, they're business problem related to what level in that stack do I want to show my differentiation to my customers and they're business problems from a liability and legal standpoint as well. The action item is, all firms will in one form or another be impacted by the emergence of the edge as a dominate design as consideration for their infrastructure but also for their business. Three ways, or a taxonomy that looks at three classes of data, primary, secondary, and tertiary, will help businesses sort out who's responsible, what partnerships I need to put in place, what technologies and I going to employ, and very importantly, what overall business exposure I'm going to accommodate as I think ultimately about the nature of the processing and business promises that I'm making to my marketplace. Once again, this has been the Wikibon Friday research meeting here on theCUBE. I want to thank all the analysts who were here today, but especially thank you for paying attention and working with us. And by all means, let's hear those comments back about how we're doing and what you think about this important question of different classes of data driven by different needs of the edge. (funky electronic music)

>> [Peter] Hi, this is Peter Burris. Welcome to Wikibon Research's Friday research meeting on theCUBE. (tech music) Today, we're going to talk about something that's especially important, given the events of this week. As many of you know, Apple announced a new iOS 11, and a whole bunch of new devices. Now we're not going to talk about the devices so much, but rather some of the function that's being introduced in iOS 11. Specifically, things like facial recognition. An enormous amount of processing is going to go into providing that type of service on devices like this, and that processing capability, those systems capabilities, are going to be provided by some new technologies that are related to artificial intelligence, big data, and something called deep learning. And the challenge that the industry's going to face is, where will this processing take place? Where will the data be captured? Where will the data be stored? How will the data be moved? What types of devices will actually handle this processing? Is this going to all end up in the cloud, or is it going to happen on increasingly intelligent smart devices? What about some of the different platform? And, ultimately, one of the biggest questions of all is, and how are we going to bring some degree of consistency and control to all of these potentially distributed architectures, platforms, and even industries, as we try to weave all of this into something that serves all of us and not just a few problems. Now to kick this off, Jim Kobielus, why don't you start by making a quick observation in what we mean by deep learning. >> [Jim] Yeah, thank you, Peter. Deep learning. The term has been around for a number of years. Essentially, it's machine learning, but with more layers of neuron, and able to do higher level abstractions from the data. Abstractions, such as face recognition, natural language processing, and speech recognition, and so forth, so when we talk about deep learning now, as a client, to what extent can more of these function? Face recognition as in iOS 11, or iPhone 10. What then will this technology, with capability, be baked into all edge endpoints now? >> [Peter] Jim, I'm having a little bit of a problem hearing you, so maybe we can make sure that we can hear that a little bit better. But, very quickly, and very importantly, it suggests that, the term deep learning suggests something a little different than I think we're actually going to see. Deep learning suggests that there's going to be a centralization, a function for some process. It's going to be the ultimate source of value. And I don't think we mean that. When we talk about deep learning, let's draw a distinction between deep learning, as a process, and deep learning as a set of systems and designs and investment that's going to be made to deliver on this type of business function. Does deep learning fully capture what's going to happen here? >> [James] Is this for me, Peter? Can you hear me, Peter? >> [Peter] I can hear you better now, a little bit saturated. >> [James] Okay, I got my earbuds in. Yeah, essentially the term deep learning is a broad paradigm that describes both the development pipeline function that quite often will, more often than not, will be handled in the cloud among distributed teams, and those function of deep learning that can be brought to the edge, to the end devices, the mobile devices, the smart sensors. When we talk about deep learning at the edge, as enabled through chip sets, we're talking about functions such as local sensing, local inference, from the data that's being acquired there, local actuation as it were taking action, like an autonomous vehicle steering right or left, based on whether there is an obstacle in their path. So really, in the broadest sense, you need that full infrastructure to do all the building and the tuning and the training of deep learning models, and, of course, you need the enabling chip sets and tools to build those devices, those functions, deep learning functions, that need to be pushed for local, often autonomous execution at endpoints. >> [Peter] So, David Floyer, that strongly suggest that, in fact, deep learning is suggestive of a new system architecture model that is not going to be large and centralized, but rather is going to be dependent upon where data actually exists and how proximate it is to the set of events that we're both monitoring, and ultimately trying to guide, as we think about new automation, new types of behavior. Take us through our thinking on some of these questions of where the data's going to reside, where the function's going to reside, and ultimately, how the architecture's going to evolve. >> [James] I think you're on mute, David. >> [David] Yes, I would put forward the premise that the majority of the processing of this data and the majority of the spend on equipment for this data will exist at the edge. Neal brought forward a very good differentiation between second-hand data, which is where bit data is today, and primary data, which is what we're going to be analyzing and taking as decisions on at the edge. As senses increase the amount of data and smart senses come, so we're going to see more and more of a processing shift from the traditional centralized, to the edge. And taking Apple as another example, they're doing locally, all of this processing of data. Siri, itself, is becoming more and more local, as opposed to centralized, and we're seeing the shift of computing down to the edge. And if we look at the amount of computing we're talking about, we're talking, with the Apple 10, it's six hundred billion operations a second. That's a lot of computing power. We see the same thing in other industries. There's the self-driving car. If you take the Nvidia Drive-2 it has a huge amount of computing power within that to process all of the different sources of data in a device which is costing less than $1,000, $600, $700. So much lower pricing of processing, et cetera. Now the challenge of data, the traditional model, is that all of the data goes to the center, is the cost of all this data, moving it from the edge to the center is just astronomical. It would never happen. So only a sum set of that data will be able to be moved. And people who develop systems, AI systems, for example, at the edge, will have to have simulation factories very local to them to do it, so car manufacturers, for example, having a small city, if you like, where they have very, very fast communication devices. And the amount of data that can be stored, as well, from this new primary source of data is going to be very, very small, so most of that data either is processed immediately, or it disappears. And after it's processed, in our opinion, most of that will disappear, 99% of that class will disappear completely. So the traditional model of big data is being turned upside down by these new and prolific sources of data, and the value will be generated at the edge. That's where the value is in recognizing a bad person coming into a building, or recognizing your friends, or recognizing that something is going wrong with a smart sensor locally. The vibrations are too high, or whatever the particular example is. That value will be generated at the edge by new classes of people and new classes of actors is this space. >> [Peter] So, Neil Raden, one of the interesting things that we're talking about here, is that we're talking about here is that we're talking about some pretty consequential changes in the nature of the applications, and the nature of the architectures and infrastructures that we're going to build to support these applications. But those kinds of changes don't take place without serious consideration of the business impacts. Is this something that companies are going to do, kind of willy-nilly? How deeply are companies going to have to think about how deeply are users going to have to think about deploying these within their business? Because it seems like it's going to have a pretty consequential impact on how businesses behave. >> [Neil] Well, they're going to need some guidance, because there just aren't enough people out there with the skill to implement this sort of thing for all the companies that may want to do it. But more importantly than that, I think that our canonical models, right now, for deep learning and intelligence at the edge are pretty thin. We talk about autonomous cars or facial recognition, something like that, there's probably a lot more things we need to think about. And from that we can derive some conclusions about how to do all this. But when it comes to the persistence of data, there's a difference between a B to C application, where we're watching people click, and deciding next best offer, and anything that happened a few months ago was irrelevant, so maybe we can throw that data away. But when you're talking about monitoring the performance of an aircraft in flight or a nuclear power plant, or something like that, you really need to keep that data. Not just for analytical purposes, but probably for regulatory purposes. In addition to that, if you get sued, you want to have some record of actually what happened. So I think we're going to have to look at this whole business, and all of its different components, before we can categorically say, yes we saved this data, here's the best application. Everything should be done in the cloud. I don't think we really know that yet. >> [Peter] But the issue that's going to determine that decision is going to be a combination of costs today, although we know that those costs are going to change over time, and knowledge of where people are and the degree to which people really understand some of these questions. And, ultimately, what folks are trying to achieve as they invest to get to some sort of objective. So there's probably going to be a difference in the next few years between, in which we do a lot of learning about deep learning systems, and some steady state that we get to. And my guess is that the ecosystem is going to change pretty dramatically between now and then. So it may be the telcos think that they're going to enjoy a bonanza on communications costs over the next few years, as people think about moving all this data. If they try to do that, that's going to have an impact on how Amazon and Google, and some of the big cloud suppliers invest to try to facilitate the movement of the data. But there's a lot of uncertainty here. Jim, why don't you take us through some of the ecosystem questions. What role will developers play? Where's the software going to end up? And to what degree is this going to end up in hardware and is going to lead to or catalyze kind of a Renaissance in the notion of specialized hardware? >> [James] Yeah, those are great questions. I think most of the functionality, meaning the local sensing and inference, and actuation, is inevitably going to end up in hardware, in highly specialized and optimized hardware for particular use cases. In other words, smart everything. Smart appliances, smart clothing, smart lamps, smart... You know, what's going to happen is that more and more of what we now call deep learning will just be built-in by designers and engineers of all sorts, regardless of whether they have a science or a computer background. And so it's, I think going to be part of the material fabric of reality, the bringing intelligence that, with that said then, if you look at the chip set architectures, and if we can use the term chip set here, that will enable this vast palette of embedding of this intelligence in physical reality. The jury is really out about whether it will be GPUs, like in video, of course, it's power out behind GPUs, versus CPUs, versus FPGAs, A6, there's various neuromorphic chip sets from IBM and others. It'll be, it's clearly going to be a fairly very innovative period of great ferment in innovation in the underlying hardware substrate, the chip sets, to enable all these different use cases in embedding of all this. In terms of developer, take the software developers. Definitely, they're still very much at the core of this phenomenon, when I say they, data scientists, as the core developers of this new era who are the ones who are building these convolutional neural networks and recurrent neural networks, and long, short-term, and so forth. All these DL algorithms very much are the province of data scientists, for the new generation of data scientists who specialize in those areas and that who work hand-in-hand with traditional programmers and so forth, to put all of this intelligence into a shape that can then be embedded and might, containerized, whatever, and brought into some degree of harmonization with the physical hardware layer into which hardware could be used for terms like, clothing, smart clothing. What gave us that, now we have a new era where the collaborations are going to be diverse among nontraditional job, or skills categories, who are focused on bringing AI into everything that touches our lives. It's wide open now. >> [Peter] so David Floyer, let me throw it over to you, because Jim's raised some interesting points about where the various propositions, the value propositions, and how the ecosystem is going to emerge. This sounds like a, once again, going back to the role that consumer markets are going to play from a volume, cost, and driving innovation standpoint. Are we seeing kind of a repeat of that, are the economics going to, of volume going to also play a role here? Muted? >> [David] Yes, I believe so, very strongly. If you look at technologies and how they evolve. If you look for example at Intel, and how they became so successful in the chip market. They developed the chips with Microsoft for the PC. That was very, very successful, and from that they then created the chip set for the data senses, themselves. When we look at the consumer volumes, we see a very different marketplace. For example, GPUs are completely winning in the consumer market. So Apple introduced GPUs into their ARM processes this time around. Nvidia has been very, very successful, together with ARM, in producing systems for self-driving cars. Very, very powerful systems. So we're looking at new architectures. We're looking at consumer architectures, that in Nvidia's case came from game playing, and in ARM, has come all of the distributed ecosystems, the clients, et cetera, all ARM-based. We're seeing that it's likely that consumer technologies will be utilized in these ecosystems because volume wins. Volume means reduction in price. And when you look at, for example, the cost of an ARM processor within an Apple iPhone, it's $26.90. That's pretty low compared with the thousands of dollars you're talking about for a processor going into a PC. And when you look at the processing power of these things, in terms of operation, they're actually greater power. And same with Nvidia with the GPUs. So yes, I think there is a potential for a big, big change. And a challenge to the existing vendors that they have to change and go for volume and pricing for volume in a different way than they do at the moment. >> [Peter] So that's going to have an enormous impact, ultimately, on the types of hardware designs that we see emerge over the course of the next few years. And the nature of the applications that the ecosystem is willing to undertake. I want to pivot and bring it back to the notion of deep learning as we think about the client. Because it ultimately describes a new role for analytics and how analytics are going to impact the value propositions, the behaviors, and ultimately, the experience of consumers, and everybody, has with some of these new technologies. So Neil, what's the difference between deep learning-related analytics on the client, and a traditional way of thinking about analytics? Take us through that a little bit. >> [Neil] Deep learning on the client? You mean at the edge? >> [Peter] Well deep learning on a client, deep learning on the edge, yeah. Deep learning out away from the center. When we start talking about some of this edge work, what's the difference between that work and the traditional approach for data analytics, data warehousing, et cetera? >> [Neil] Well, my naive point of view is deep learning involves crunching through tons of data in training models to come up with something you can deploy. So I don't really see deep learning happening at the edge very much. I think David said this earlier, that the deep learning is happening in the big data world when they have trillions of observations to use. Am I missing your point? >> [Peter] No, no. We talked earlier about the difference between deep learning as a process and deep learning as a metaphor for a new class of systems. So when we think about utilizing these technologies, whether it's deep learning, or AI, whatever we call it, and we imagine deploying more complex models close to the edge, what's that mean from the standpoint of the nature of the data that we're going to use, the approach, the tooling that we're going to use, the approach we're going to take organizationally, institutionally, to try to ensure that that work happens. Is there a difference between that and doing data warehousing with financial systems? >> [Neil] Well, there's a difference in terms of the technology. I think that 10 years ago, we were talking about complex event processing. The data wasn't really flowing from centers, it was scraping Web screens and that sort of thing, but it was using decision-making technology to look for patterns and pass things along. But you have to look at the whole process of decision making. If you're talking about commercial organizations, it's not really that much in commercial organizations that requires complex, real-time, yeah, making decisions about supply chain or shop floor automation, or that sort of thing. But from a management point of view, it's not really something that you do. The other part of decision making that troubles me is, I wrote about this 10 years ago, and that was we shouldn't be using any kind of computer-generated decision making that affects human lives. And I think you could even expand that to living things, or harming the environment and so forth. So I'm a little bit negative about things like autonomous cars. It's one thing to generate a decision-making thing that issues credit cards, and maybe it's acceptable to have 5% or 3% of decision just completely wrong. But it's that many wrong in autonomous driving, especially trucks, the consequences are disastrous. So we have to be really careful about this whole thing with IoT, we've got to be a lot more specific about what we mean, what kinds of architectures, and what kind of decisions we're trying on. >> [Peter] I think that's a great point, Neil. There's a lot that can be done, and then the question is that we have to make sure that it's done well. We understand some of the implications, and again, I think there's a difference between a transition period and a steady state. We're going to see a lot of change over the next few years. The technology's making it possible to do so, but there's going to be a lot of social impacts that ultimately have to be worked out. And I'll get, we'll get to some of those in a second. But George, George Gilbert, I wanted to give you an opportunity to talk a little bit about the way that we're going to get this done. Talk about how we're, where's this training going to take place, per what Neil said? Is the training going to take place at the edge? Is the training going to take place in the cloud? Institutionally, what is the CIO and the IT organization have to do to prepare for this? >> [George] So I think the sort of widespread consensus is that the inferencing and sort of predicting for the low latency actions will be at the edge, and some smaller amount of data goes up into the cloud training, but the class of training that we will do over time changes. And we've been very fixated on sort of the data centricity, like most of the data's at the edge a little bit in the center. And Neil has talked about sort of secondary, or reference data, to help build the model from the center. But the models themselves that we build in the center and then push out, will change in the sense that we look at the compute intensity. The compute intensity of the cloud will be, will evolve, so that it's more advantageous there to build models that become rich enough to be like simulation. So in other words, it's not do I, if I see myself drifting over the lane marker on the right, do I correct left? But you have a whole bunch of different, different knobs that get tuned, in that it happens over time. So that the idea of the model is almost like a digital twin, but not of, let's say, just an asset or physical device, but almost like a domain, in that that model, it's very compute intensive, it generates a lot of data sets, but then the model itself can be distilled down and pushed out to the edge. Or, essentially, guiding or informing decisions, or even making decisions with a lot more knobs than you would have with a more simplistic model. >> [Peter] So, Ralph, I know that we've spent some time looking at some of the market questions of this. Based on this conversation, can you kind of give a summary of how much data volume we think is happening, data movement's happening? What's the big, broad impact on some of the segments and opportunities over the course of the next couple of years? >> [Ralph] Yeah, I think the, think back maybe 10 years, the amount of unstructured data that was out there was not all that great. Obviously, in the last 10 years of war, there's a lot more of it. So the growth of data is dramatically increasing. Most of it is going to be in the mobile area. So there's just a lot of it out there. And this, I think fishing for where you derive value from that data is really critical for moving optimization of processes forward. But I think I agree with Neil that there's a lot of work to be done yet about how that actually unfolds. >> [Peter] And there's also a lot of work to be done in areas like, what will the role of, who's going to help define how a lot of these platforms are going to be integrated together. What's the role of standards? What role will government play? There's an enormous number of questions here. But one thing we all agree on ultimately, is that this is an emerging source of, or this technology is an emerging source of dramatic new types of business value taking on problems that we've never thought about taking on before. And it's going to have an enormous impact on how IT organizations work with business, how they work with each other, how businesses work together. This is the centerpiece of the new digital business transformation. Alright, so let me summarize this week's findings. The first observation we make is that this week, Apple introduced facial recognition directly in iOS 11, and it wowed much of the industry, and didn't get a lot of people excited for a variety of reasons, but it does point to the idea that increasingly we're going to see new classes of deep learning, AI, machine learning, and other big data-type technologies, being embedded more deeply in systems as a way of improving the quality of the customer experience, improving operational effectiveness and efficiency, and ultimately, even dramatically improving the ratio between product and service revenue in virtually everything that we can think about. Now, that has led folks to presume that there's, again, going to be this massive migration of workload back into the cloud, both from a data standpoint, as well as from a workload standpoint. But when we stop and think about what it really means to provide this value, it's pretty clear that for a number of reasons, including real-time processing to provide these services, the cost of moving data from one point to another, and that the characteristics of the intellectual property controls, et cetera, restricts the pressure to try to move all this data from the edge, client, and device back into the cloud. And that the new architectures, increasingly, are going to feature a utilization of dramatic new levels of processing on devices. We observe, for example, that the new iPhone is capable of performing 600 billion instructions per second. That's an unbelievable amount of processing power. And we're going to find ways to use that up, to provide services closer to end users without forcing a connection. This is going to have enormous implications, overall, in the industry. Questions, for example, like how are we going to institutionally set up the development flow? We think we're going to see more model building at the center, with a constrained amount of the data, and more execution of these models at the edge. But we note that there's going to be a transition period here. There's going to be a process by which we're learning what data's important, what services are important, et cetera. We also think it's going to have an enormous impact, for example, on even describing the value proposition. If everything is sold as a product, that means the cost of moving the data, the cost of liability, et cetera, on these devices is going to be extreme. It's going to have an enormous impact on the architectures and infrastructures we use. If we think in terms of services, that might have a different, or lead to a different set of ecosystem structures being put in place, because it will change the transaction costs. The service provider, perhaps, is going to be more willing to move the data, because they'll price it into their service. Ultimately, it's going to have a dramatic impact on the organization of the technology industry. The past 25, 30, 40 years have been defined, for the first time, by the role that volume plays within the ecosystem. Where Microsoft and Intel were the primary beneficiaries, or were primary beneficiaries of that change. As we move to this notion of deep learning and related technologies at the edge, providing new classes of behavior, it opens up the opportunity to envision a transitioning of where the value is up and down the stack. And we expect that we're going to see more of that value be put directly into hardware that's capable of running these models with enormous speed and certainty in execution. So a lot of new hardware gets deployed, and then the software ecosystem is going to have to rely on that hardware to provide the data and build the systems that are very data rich to utilize and execute on a lot of these, mainly ARM processors that are likely to end up in a lot of different devices, in a lot of different locations, in its highly distributed world. The action item for CIOs is this. This is an area that's going to ensure that a role for IT within the business, as we think about what it means for a business to exploit some of these new technologies, in a purposeful and planful and architected way. But it also is going to mean that more of the value moves away from the traditional way of thinking about business systems with highly stylized data to a more clear focus on how consumers are going to be supported, devices are going to be supported, and how we're going to improve and enhance the security and the utilization of more distributed, high quality processing at the edge, utilizing a new array of hardware and software within the ecosystem. Alright, so I'm going to close out this week's Wikibon Friday Research Meeting on theCUBE, and invite you back next week where we'll be talking about new things that are happening in the industry that impact your lives and the industry. Thank you very much for attending. (tech music)