About this episode

We all want to get more value from our data, right? So should we wrangle it, prep it, classify it, mine it, or model it? Then once we’ve got our data squared away, should we do prescriptive or predictive analytics? And does that require real-time or just-in-time data? If we had a nickel for every buzzword we hear, we could afford far better podcasting equipment. 

In this episode, Tim and Juan are joined by Kirk Borne, Chief Science Officer at Dataprime AI to demystify some of our most egregious buzzwords and look at what everyone’s going to be talking about next.

Special Guests:

Kirk Borne

Kirk Borne

Chief Science Officer, DataPrime AI

This episode features
  • What is industry 4.0?
  • How can an edge be intelligent?
  • What is the buzzword that makes you want to pull your hair out the most?
Key takeaways
  • Always ask “why?”
  • We entering the age of small data
  • Stay a life-long learner

    Episode Transcript

    Tim Gasper:
    Welcome. It’s Wednesday once again and it’s time for Catalog and Cocktails. It’s your honest, no BS, non salesy conversation about enterprise data and you’ve got Tim Gasper over here. I am a data nerd and product guy and joined by co-host, Juan. Hey, Juan.

    Juan Sequeda:
    Hey, Tim. I’m Juan Sequeda. I’m the principal scientist here at data.world and it’s Wednesday. Middle of the week. End of the day and it’s time to take a break and chat about data and all that stuff that we do hear at Catalog and Cocktails and today we have a very special guest because all our guests are awesome. They’re so cool and very special, but today, we have Kirk Borne and Kirk is … I mean, who doesn’t know who Kirk Borne is, right? He is like one of the top influencers in the world on data science and he’s also the chief science officer at Data Prime. Kirk, how are you doing?

    Kirk Borne:
    Good. Thank you. Glad to be here. This is great.

    Juan Sequeda:
    Awesome. So we always like to kind of kick it off with our Tell and Toast, so what are you drinking and what are you toasting for?

    Kirk Borne:
    Well, I’m drinking water. My wife said avoid anything else. I’m drinking water. I guess I’m toasting the start of the school year because it’s all about learning for me. Lifelong learning is a passion of mine and I have grandkids now and when I see them getting excited about going to school, I said this is wonderful. So toasting all the teachers, educators, administrators, students. Parents. Wait a minute? Have we covered everybody in the world yet?

    Juan Sequeda:
    How about you, Tim?

    Tim Gasper:
    No, that’s awesome. I’m going to end up toasting to the same thing because we’re actually getting our kids going in school now. I got one kid in in-person school and I got one in virtual school and I got one who isn’t in school yet, so oh my goodness. Things are just still crazy. But the drink that I’ve created today, it’s actually based on a shrub cocktail called La Hoya but instead of tequila I made it with pisco and actually added a little bit of blue caracao in here to make it look like a turquoise color. So I call it a Chrysocolla, which is a Peruvian gemstone. This is a Tim special. I don’t think this exists, so hopefully this tastes good.

    Kirk Borne:
    Awesome. I haven’t had pisco since my trip to Chile.

    Juan Sequeda:
    Always a good discussion is where does pisco come from? Either from Chile or Peru. That’s another …

    Kirk Borne:
    Well, I don’t know, but I had some in Chile.

    Juan Sequeda:
    Well, I’m … I only have two things right now where I am. I have beer and vodka, so I looked up what can I do with beer and vodka and here you go. It’s a twist on the shandy beer, so I have a stella. I added some of this tamarind Smirnoff vodka. I had some honey and lemon and it actually is really, really good. I’m surprised. You would not think vodka and beer would go together, at least this vodka that I have, so-

    Tim Gasper:
    It kind of does sound disgusting, I’ll say.

    Juan Sequeda:
    I would … Yes. And I’m actually toasting for today is my wife’s birthday so-

    Kirk Borne:
    Hey, happy birthday.

    Juan Sequeda:
    Happy birthday to my wife,[inaudible 00:03:08]. Cheers.

    Tim Gasper:
    Happy birthday. Cheers, Kirk. Welcome. Glad to have you here.

    Kirk Borne:
    Awesome.

    Juan Sequeda:
    So our warm up question today is what is the buzz word that makes you want to pull your hair out?

    Kirk Borne:
    Well, I don’t have much left.

    Juan Sequeda:
    You’ve gone through a lot of buzz words, right?

    Kirk Borne:
    I’ve thought about it and I guess one of the words that really gets me going is, maybe two words together, exponential growth, because people use that phrase and they have no clue really what they’re talking about. Exponential growth doesn’t mean that it’s going up. It means it’s going up at a certain rate. Anyway.

    Kirk Borne:
    So to tie that back to a data comment, I would often tell people the difference between linear growth, exponential growth, and combinatorial growth, all right? So linear growth is two times X, exponential is like two to the X power, and combinatorial is like X factorial which is like X to the X power. So that’s way faster than two to the X, X to the X. So when people say to me they have exponential growth of data, I would say, “Really? So little?”

    Tim Gasper:
    And they’re like, “Why are you making me be literal here? I’m trying to evoke a feeling.”

    Juan Sequeda:
    That’s the honest, no BS right there. Tim, how about you?

    Tim Gasper:
    One of the buzz phrases that I’ve been very annoyed with lately, makes me want to pull some of my hairs out, is the phrase data intelligence which to me … I know people are trying to apply it to represent certain areas of doing more with your data and being smarter with it and driving action back, but I feel like it’s such a generic phrase and when folks are saying, “We’re trying to achieve data intelligence,” I’m like, “Anything you do with data is hypothetical, hopefully an intelligent use of that data.”

    Juan Sequeda:
    I forget the phrase, I’ll look it up again, but it’s like the opposite. Wait? So you would do something unintelligent with data? No. It’s obvious that whatever you want to go do with data’s intelligent so you’re telling me the obvious. Thank you, I guess.

    Tim Gasper:
    Yeah. Yesterday I was doing data stupidity, but then I had a euphoric moment where I was like, “Oh, I got to data intelligence.” That’s mine.

    Kirk Borne:
    Well, maybe that’s what they’re saying.

    Tim Gasper:
    Maybe.

    Juan Sequeda:
    Well, mine is … There’s a combination. One is the old one that we’ve been … Now old, but big data. I mean, the whole, “We’re doing big data. We’re doing big data.” I’m like, “Wait a minute. The amount of your data you’re dealing with fits in my pocket. Literally, it fits on my phone. Heck no, it’s not big data.” Then when we start talking about efficiency and scalability, it’s like, “Wait, wait. What does efficiency actually mean in your context and stuff?” So people like to throw those words around and what does this actually mean? So hey, always say words matter.

    Juan Sequeda:
    Talking about words matter, let’s jump into this discussion. So Kurt, honest, no BS. We’re going to start demystifying some buzz words here. What the heck is industry 4.0 or industrial revolution 4.0? I’ve heard and seen this more from the Europeans, right, especially the Germans use a lot industry 4.0. This is something I don’t see a lot in the U.S. but we’re kind of starting to see it more and what the heck is industry 4.0?

    Kirk Borne:
    Well, I think the Germans are on to something there. I like to talk about it in the context of the first three industrial revolutions, so interesting 4.0 represents the fourth. The first came about with the steam power, steam engine, over 200 years ago and I specifically make reference to how the Industrial Revolutions change the way work is done. At the dawn of the first Industrial Revolution, 90% of the U.S. work force was doing farming and then a hundred years later the Electricity Revolution started and electricity became the main source of power and industry and by the time this was underway, as a result of the preceding Industrial Revolution, in that hundred years, the percentage of the U.S. work force that was farming dropped from 90% to 50%.

    Kirk Borne:
    Well, in the next hundred years, or eighty years or, probably about eighty years or so, we got to the Computer Revolution which is the third Industrial Revolution, which really started about fifty years ago. So the computer revolution, by the time that started the number of people, work force percentage doing farming dropped from the previous revolution from 50% now down to 5%. So the key thing during all these Revolutions is not just sort of the power source, if you will, but also the way we do work, okay?

    Kirk Borne:
    Starting with almost everyone doing farming to half the population doing the farming to just a tiny percentage of the population work force doing farming to the current age where literally 0.5%, like a half a percent, of the work force in this country are farmers and we’re feeding ten times as many people as we were two hundred years ago, at least these people. So what is it that’s really different? We’re still in the computer age, right? That hasn’t changed in fifty years.

    Kirk Borne:
    So what’s really changed is this era of hyper-connectivity. Everything is connected, right? Through our smartphones, through our devices. Even our cars are connected. Our homes are connected. The roads are connected. So it’s really an era of hyper-connectivity. So what powers thing is the knowledge that is transported across the computer networks so it’s not just using computer networks to connect to a computer, which is sort of how the computer revolution began. It’s now knowledge itself is being transmitted across the networks and it’s powering work, it’s powering homes, it’s powering life. Think about work from home in the last year. I mean, all this hyper-connectivity, I don’t think we could have worked from home fifty years ago, or even let’s say thirty five, forty years ago when the PC revolution started. If we just gave everyone a PC and said, “Go work from home,” it just never would have worked out, but I think because we have this hyper-connectivity across everything, our platforms, our banks, our homes, our cars, our smartphones, our apps, Zoom calls, platforms like this. We’re just all connected and so the way work goes now is completely different than it ever used to be. So that’s how I see industrial revolution. It’s sensors and data everywhere, powering and informing and changing the way we do things.

    Juan Sequeda:
    So to kind of pick … Be a little pedantic on here, wouldn’t this just be like 3.2 or 3.3 because the thing that you mentioned before, like steam, electricity, computer, these things fundamentally changed society, right? Going from steam to electricity and then going to computers but what you’re talking about more specific type of computers, right?

    Kirk Borne:
    Well, I would say like AI for example is probably the thing that’s fundamentally different than just having the computers we had fifty years ago. It’s sort of like a parallel to farming because people didn’t stop doing farming when the steam engine came along. They just did their farming with steam powered plows and steam powered combines and things like that, right? So they were still doing things similar, it’s just that the way they did it was so much more effective, it freed up people to do other things and that’s what I think AI is doing.

    Kirk Borne:
    I mean, I’ve been looking a lot into sort of voice assistance and how voice assistance in customer call centers and a lot of different ways how it’s really changing the way that customers interact with businesses and how clients and business providers interact with one another. So it’s not really necessarily changing how we’re doing business, it’s just the whole transformative way of efficiencies and effectiveness of different things are made so much better that we can now start trying many new different kinds of things and so that’s really the revolution. It’s really this transformation and disruption of the old ways and introducing all kinds of new ways and I think the AI part of that story, because AI is … To be really an intelligence, and maybe this is sort of my closing comment on this, is that intelligence to me is really about not just having … Finding one pattern in one stream of data, right?

    Kirk Borne:
    So let’s say you have a time series and then you see a sudden rise or decline or a glitch or something in the time series. That’s not really intelligence. Intelligence is the full cognitive awareness of, “Why did that happen? What’s going on around that?” So we talk about … In psychology, you talk about cognitively impaired people, right? So a cognitively impaired person could walk into a room and make some really off the cuff comment that’s just doesn’t work in that room, right? They can’t read the room. I walk into a board meeting and I make some silly comical comment during a very serious time in my business. That’s really not being very cognitively aware of what’s going on.

    Kirk Borne:
    So cognitive intelligence is about seeing and getting data and information from everything else in the environment so you make the best decision and take the best action. So I think what AI does for us is if it’s really going to be an AGI, a general intelligence, it has to have this sort of cognitive ability of getting context and this is where the hyper-connectivity of Industrial Revolution 4.0 gives us that connectivity, the context in which things are happening in our world, in our work, in our lives, in our homes, even in our cars.

    Tim Gasper:
    That’s interesting. When you talk about hyper-connectivity, obviously there’s a connectedness and networking aspect to what you’re talking about here, but then when you talk about AI, there’s more of a sort of the way we’re applying the models, that fact that we’re able to leverage the data in a smarter way. When I mentioned about my buzz phrase, data intelligence, it sounds like you’re also kind of connecting it to how that phrase can actually be a useful phrase which is to think of that as a goal, as a higher sort of plane of what we’re doing with our data, with our computers, that actually is now a step up. Is that the right way to think about it and then secondly, so why are people getting so excited? What is actionable about this that’s different than maybe the computer revolution and what that enabled?

    Kirk Borne:
    Well, I think you hit the nail on the head with connected intelligence, right? So one thing for sure, you just think about breaking down the silos in business. Business has always been, not always but for many years, decades, have been doing business intelligence. But that business intelligence usually lived in silos so the marketing department, the product department, the sales department, the customer care department, those people didn’t share data with one another. So maybe people are calling up about some defect in the product and no one’s getting the message in some other part of the business.

    Kirk Borne:
    So the connected intelligence is what’s really making this difference and of course, there’s so much data it’s hard to do that without something to orchestrate the flow of that information and knowledge and that’s where the hyper-connectivity and AI come together.

    Kirk Borne:
    So let me give you an example of how something can be really different from what historically it is and that’s the digital twin. So digital twins are really big in Germany. You talk about Germany being pushing this and-

    Juan Sequeda:
    I’m glad we’re bringing digital twin because that’s another buzz word that comes in here. What is a digital twin?

    Kirk Borne:
    A digital twin is a computer model of a real, physical thing. So NASA was one of the first groups in the world to introduce digital twins where they made a computer of the thing which they deployed in deep space. So they could run it through the paces of the thing going on in deep space and whenever there was a problem, they could sort of it replay it. They have the data. They’ve collected the data from the real, physical system and they can just basically re-record it, not re-record it but replay the recording so to speak and figure out sort of what was going on when it glitched, okay? What kind of things were happening and find a solution.

    Kirk Borne:
    So industries are doing this now in manufacturing plants, on engines of all sorts, jet engines, all kinds of engines. I’ve seen … I actually walked through a virtual reality demo of a windmill, a solar wind powered … Not wind powered. Wind energy generated windmill where they actually had the ability in this VR environment to actually climb into the gearbox of this 200 foot tall tower and poke around in the electronics to see what was the problem. But you’re not really in the gearbox. You’re actually in the simulated environment of a virtual reality world playing with the simulation of that windmill. So if I tweak this, tweak that, how does it behave? So you can actually fix problems, identify precursor problems like preventive and predictive maintenance and that kind of thing in the computer before something really bad happens.

    Kirk Borne:
    So building these sort of connections between real world and simulated world is really one of the real amazing differences in this hyper-connectivity we’re now talking about. In the past you could probably rebuild that simulated instance of the physical thing, but you never could really connect the two data sources.

    Juan Sequeda:
    So a couple things I want to dive in here. One is as I mentioned, Industry 4.0 is something that we hear more coming from Europe, especially from Germany in particular. I postulate then that the reason why this happens is because you have … The Germans are great on manufacturing, right? On cars and stuff and when you talk about the digital twin, it’s about manufacturing, about engines and stuff like that. So is that kind of the connection right there? That’s one.

    Juan Sequeda:
    Second, here in the U.S. I’m starting to hear this more and actually I think many, many years ago it started with GE, right? GE was one of the first companies I started hearing more they talk about digital twins for manufacturing. Now I get the definition of digital twin. It comes from the actual engine, right? I like how you just said it’s a computer model of the real thing that you go want to simulate. Basically you want to go replay what happened before. The thing that really … I call BS at this moment because it’s so unclear to me and people are just throwing this word is when they say digital twin for any other non-manufacturing enterprise, so like, “Here’s my … I’m an e-commerce company and whatever and we have a digital twin ourselves,” it’s like, “Digital twin of what? Your entire company or what?”

    Tim Gasper:
    Yeah. It is fair to apply it that broadly? For example, I’ve heard people talk about digital twins in the context of marketing and trying to do something around a person doing things and this is their digital twin. Can we use it in that way or is it really more of a sensors and more of a manufacturing type of situation that it’s more appropriate for?

    Kirk Borne:
    Well, I think you get to these extremes where people will use that word, again, as a buzz word, okay, let’s say it that way, versus a more real practical thing. I’ve seen digital twins of entire manufacturing plants. Articles that talk about the digital twin of an entire manufacturing plant. What I think sort of softens the blow in that conversation is it’s all the sensors and the data we collect from sensors that lead to predictive models that is predicting when something will happen given the stream of data. It’s also about prescriptive models. If we tweak this, what will the outcome be? If we tweak that, what will the outcome be? So I can see that being a very reasonable thing in marketing, for example, or a very reasonable thing in a manufacturing assembly line, for example. If we change this or change that, what would the outcome look like?

    Kirk Borne:
    So you have to have a pretty high fidelity model of the thing you’re simulating, right? You don’t want to have some simple model. You need to have a model that really captures all the moving parts, so to speak, whether it’s a customer, an e-commerce store or a windmill or a jet engine-

    Tim Gasper:
    And you need a lot of data, right? A good signal.

    Kirk Borne:
    Exactly. So now you need a lot of data. You need the predictive models, okay, the causal models which I call the prescriptive models. What will cause what to happen if we tweak something. But also root cause analysis models where you can replay the data when the thing did fail and say, “What was it in the history of this that caused the failure so we don’t allow that to happen again?”

    Kirk Borne:
    So you got to have high fidelity, you got to have data, and again, there’s got to be this connectivity because it’s not just simulation of data. It’s taking the real data from the real system and it’s what called assimilating into the model, into the simulation. So data assimilation, I learned from NASA days, that’s a really powerful tool where you’re bringing the real live data into the simulation to update.

    Juan Sequeda:
    So I think … A big thread here is about connectivity and we’re talking about when it comes to connectivity of data, it seems like we have a lot, in the manufacturing space at least, there’s a lot of sensors and I think that’s one of the stuff that we get a lot of data and that data needs to be connected and stuff but then there’s also this connectivity of knowledge which in reality is going to give us that extra context, give us all that data that we’re producing to understand what it means such that we can go and do more fancy AI stuff because if we’re just focusing on the data itself, without that context, without that knowledge, it’s almost kind of still garbage in, garbage out in a way. We get enough but we’re getting the obvious in a way.

    Juan Sequeda:
    That’s kind of the takeaways I’m getting right now. Connectivity of data with sensors and connectivity of knowledge is what we really need for AI.

    Kirk Borne:
    Yeah, and I think I’m going to get a little philosophical here. I mean, there’s been this diagram that people have used for years to talk about going from data to information to knowledge to wisdom. We’ve all seen that pyramid. I actually use that a lot in my talks. I say, “Well, what is data?” Well, data are the ones and the zeros. Those are the bytes you collect from your sensors so by themselves, they’re just ones and zeros. Information is something you extract from the ones and zeros. So let’s say I have a satellite image. This is my example from NASA. People look at a satellite image, “Oh, that’s a cool picture of a volcano or a flood or an iceberg or something,” but if you just gave someone this file and you didn’t tell them what it was and they tried to break it apart, they would just see ones and zeros. That’s all you’d see, all right? We know it’s an image, so we actually display it as an image but what you extract from that image is, for example, the location of the event, the type of event, how big it is, time of day. You extract information from the ones and zeros.

    Kirk Borne:
    Now you connect that to something else. So here’s a flood. Well, gee, there was this wildfire in that part of the country last year, burned all the brush, which made all the topsoil run off and therefore you end up with this flood plain because all the fires basically burned out the brush that would have withstood the heavy rain. So now you got this runoff and you got this flood. So NASA’s analyzing floods across the world in the context of climate change when you look at increased number of wildfires and increased drought conditions and all kinds of stuff and so now you’re connecting all these pieces of information from the data bits to create knowledge and from that knowledge, they say, “When this happens, this will happen,” and so that connectivity now gives you the wisdom to know how to act, know what to do next because you see the sort of causal connection between things.

    Kirk Borne:
    Instead of, like I said before, I take an image of this location on Earth, I take an image a year later and I see something’s different. Oh, okay. So there’s something different. Okay. I saw something different. Someone wants to know, “Why? How? What?” That’s how you become a cognitively intelligent person and that’s how our intelligence, our data, AI, becomes cognitive and real AI becomes intelligent about seeing those factors. So it’s causal connection, but also able to build predictions from it. Predictive models and then, again, prescriptive models, that is, “How can we do something so it doesn’t happen this way again?”

    Juan Sequeda:
    Yeah, I love how you’re saying the notion of being cognitive is ask the why. I think that’s something really important that we as humans ask the why and if we are expecting machines, AI, to go do that, that’s a true kind of measure right there is not the intelligence but how cognitive things are because we ask, “Why is this happening?” I love that same example that you’re doing about the data, information, knowledge. My example is if I see a number, right, okay here’s a number 42. What does that mean? Hmm? I don’t know. Well-

    Kirk Borne:
    That’s a good choice by the way.

    Juan Sequeda:
    Yeah. Perfect idea there, right? Okay, then next to 42 I have this thing called USD. Well, that’s a currency, so now 42 USD can mean something about a price or whatever, but then you want to go get more context around it. Well, I have this thing, an attribute called the net sales. Oh, something was sold and this and then you get more context about it. Oh, there’s an order and there’s a customer and so forth, right, and then you start asking these questions. It’s like, “Wait. Who’s buying this and why are they buying it? How come they’re not buying more? How come they’re buying less?” And so forth. This is a great way of thinking about it, having these examples all the way from data all the way to wisdom. I love this.

    Kirk Borne:
    And you just described the whole connectivity thing. You’re connecting all these different dots, if you will, all the data points to get the understanding of that number, that 42.

    Juan Sequeda:
    Yeah. Tim, as we always do [crosstalk 00:24:35]

    Tim Gasper:
    Oh, go ahead, yeah.

    Juan Sequeda:
    No, knowledge first, right? This is the stuff that we’re realizing right now that we’ve always been in this data first world that we just need to start thinking about it’s not just data first, it’s knowledge first. It’s connectivity.

    Tim Gasper:
    People and connections and context. Yeah. I love where this conversation’s going here. As we talked about sensors and around Industrial Revolution 4.0, this idea of sensors and getting all these information, it kind of leads to our second buzz phrase here, which is intelligent edge. One of the things that I’m very curious, Kirk, is how do you think about what is the intelligence edge and what are the implications of that? Does that just mean like … Right now we’re gathering lots of data and it’s going in our data warehouse and we’re running our big models on it and that kind of thing and we’re going to push more of that to edge? We’re going to process more at the edge? What is intelligence edge?

    Kirk Borne:
    Well, it has a lot to do with autonomous systems, first of all. Clearly you can collect data, store it in a data warehouse and spend the next six months analyzing it. You don’t want to do that in a self-driving car, right? So a child walks in front of the car while you’re driving and a camera is watching the front of your car and it sends the data back to the Cloud and you analyze the data and fifteen minutes later you say to the car, “Oh, you need to apply your brakes.” Well, it’s kind of late fifteen minutes later, right, because the kid is in front of you car fifteen minutes ago.

    Kirk Borne:
    So the intelligence edge is the ability for a system to respond and take an action, take a decision or whatever, at the point of data collection. Now that requires something called edge computing that is putting the smarts as close to the point of data collection so we now have sensors that have ML built in. If you ever heard of TinyML, so TinyML is an example and there’s other things like this. They’re sort of neural chips, sort of neural processing chips that will actually do process in neural networks even on the chip as you collect the data.

    Kirk Borne:
    So computer vision is a straightforward example, right? Your car needs to interpret what’s in that image so the car can take the right action in that moment, but there’s also intelligent edge to things which are sort of less critical or less intense as that. For example, just postal code scanners, right? So a package runs through your mail and it scans like zillions of packages per minute and it reads those postal codes which are written by human hands. People write numbers in all kinds of different ways. So there’s an example.

    Kirk Borne:
    There’s even an e-commerce store that the moment a person is clicking on something, maybe there’s an action that could be taken autonomously driven by the web analytics of that particular customer interaction and even in Fitbits or wearable health devices. You want to get some kind of feedback. For example, if you’re getting some kind of alarm that your heart rate is out of balance when you’re exercising, you want to get that alarm now not after you’ve passed out on the street.

    Kirk Borne:
    So intelligent edge is really about getting that intelligence and it includes that cognitive thing we were just talking about, the contextual information around that at the point of data collection. So the edge can be anything. It can be at the point self driving car, it can be in a health environment, it can be in a manufacturing environment. Anywhere where you sort of need information, knowledge and understanding at the point of data collection.

    Juan Sequeda:
    So this is reality today. I mean, all those example that you were stating right now, those are happening right now. So this intelligent edge … It’s not a buzz word anymore, then. It’s a reality. Or-

    Kirk Borne:
    Well, what makes it different now is the internet of things, because the internet of things is predicted around 25 trillion sensors on the world so we’re really creating an intelligent edged world, right, and I’ve seen the numbers that different groups have sort of projected the market value of the intelligent things market is anywhere from $1 trillion to several trillions of dollars in the next ten years. So businesses are realizing that they just can’t bring all that data back to a warehouse and try to extract the knowledge and intelligence out of it over a period of weeks or months. They need it now, action now.

    Kirk Borne:
    So I like to talk about internet of things a lot, about this, and I like to use a phrase for the internet of things. I call it the internet of context. All these sensors in your manufacturing plant, in your car, in your smart home, in your smart city are giving you contextual information about that environment, okay? What’s going on in that space, so when you see something unusual in a particular sensor, you know have a context of what’s going on at that same time in that same area. So all these sensors by themselves give you information about whatever it’s sensing, but also gives you contextual information about the environment it’s situated in, whether it’s in a manufacturing plant, your home, your car, your city, whatever. So internet of context, to me, is powering the intelligent edge because it’s not just the one … Again, like you said before, it’s not just the ones and zeros, the 42, it’s what else does that 42 correspond to or what is related to? What is it connected to? What is it associated with?

    Tim Gasper:
    Being able to embed more meaning in all of these. You know, we’ve got a lot of folks on this podcast that are data people, whether that’s analysts or architects or folks that are organizing governance initiatives and things like that and as they think about the intelligent edge, there’s obviously an aspect of this that is more how this is … The sensors and the intelligence are being implemented in the products themselves into the assembly line. How does this impact data professionals, right? Does the stack change? Does the skill set change? What’s the impact and relevance there?

    Kirk Borne:
    I’m sure it does. I’m not a data-

    Tim Gasper:
    We’ll find out together.

    Kirk Borne:
    I’m not a data engineer. I’ve worked with a lot of data engineers but I never pretended to be one, so I’m sure the stacks going to change for one thing. Like I said before, a lot of the smarts is being … It’s now in sort of the microcode in these newer chips, right, and so at some level you’re not even needing … You can imagine almost not even needing to know machine learning because it’s already in the chip so you need to buy the right chip. For example, there’s particular chips that actually have cameras in the chip, right, so it actually collects the data, the image, and does object detection in real time.

    Kirk Borne:
    I heard someone talk once at a conference, I still don’t understand exactly what they meant, but they basically did … It was a behavioral chip, right? They detected behavioral patterns in real time and I don’t know what that meant. Maybe it meant trend lines. Maybe it meant anomaly detection. I don’t know what they meant, but behavior-

    Tim Gasper:
    The word behavioral is another tricky one.

    Kirk Borne:
    Behavior on a chip. I don’t know. That sounded kind of interesting to me, but I can imagine that for example in a stock market, high frequency trading instance where you want to see what the behavior is instantaneously as best you can when there’s a very high volume of stuff going across the wire, okay? So what is the behavior at this moment of this ensemble of traders, or something like that. So I don’t know if that’s what they meant, but it’s an interesting concept to me of how you can … Whether behavior means like ensembles or segmentation or outlier detection or trends or correlations, I’m not sure but it sounded really interesting.

    Tim Gasper:
    Yeah. Well, and the idea of more of the … Even just that comment you made about the chips and more of the machine learning being executed on a chips or behavioral analysis or whatever things they have deployed there, that’s different because obviously the state of data science today is much more focused on the idea that, “Well, I’m going to collect all the data and then I’m going to run a bunch of stuff on it and maybe I’ll build my knowledge graph and stuff on it,” but it’s much more of a collect and then build whereas there’s this whole aspect of the edge which is like, “No, there’s things that are forward deployed,” some of which is hard coded. How does that interplay and how do you iterate on that? I don’t know. It’s interesting.

    Juan Sequeda:
    So there’s a cycle here, right? So at the end if you’re pushing something back on to the edge, right, that’s going to do the machine learning, the AI, that has to … That model has to come from somewhere and that’s going to come from … There’s an initial model that was created so you’re going to bring in … You’re going to still do your traditional data, gathering data, integrate and bring it all together, doing some machine learning stuff. Then you have to go figure out that, “I need to take that, put it back into the edge, onto the sensor,” or whatever because it’s going to feed that and then there’s new stuff that’s going to come out, right? I’m going to be making some decisions at that moment. I’m going to learn new things.

    Juan Sequeda:
    I need to then send that data back to the central place and keep updating the model, then send the model over there. There’s this whole cycle that needs to occur over here and I’m … Are we actually prepared … People we talk to, I don’t even think … Do you know how to go take your model and update it and go send it to all the different millions or billions of sensors that are out there and making sure that they are … Because just doing an update, right? I got to go update my sensors for the model that we have because we realized that the previous one was screwed up, it had an error, there was bias and stuff. There’s going to be so much governance and things we need to go track about this. This is … Yeah. I’m not realizing this. This next year is going to be so much more things to keep track of.

    Kirk Borne:
    Oh, you’re such a doomsayer.

    Juan Sequeda:
    Hey, it’s more problems, more things to go work on.

    Tim Gasper:
    Opportunity.

    Kirk Borne:
    Opportunity is here. [crosstalk 00:34:48] Well, let me tell you a story. I worked at NASA for twenty years in some capacity. I was never a NASA civil servant employee, but I worked on projects with NASA, Hubble Space telescope for one. I left that role eighteen years ago, but even then, eighteen plus years ago we were talking about this very topic and so we didn’t have too many GPUs in those days, but they had these FPGAs, floating point graphic accelerators, and the FPGA was essentially what you were just describing. It’s a programmable chip. So NASA’s idea was we were going to deploy these chips on our deep space probes and these probes would be programmed to do certain things, like collect soil samples on Mars, take images of the moons of Saturn, measure electron density in interplanetary medium, things like this, right? But you want to be able to recode that if you discover something and you say, “Well, we were sampling this every thirty seconds but we better start sampling this every five seconds because there’s something going on here,” or conversely, “Gee, we don’t need to sample every thirty seconds. Maybe every thirty days because it never changes.”

    Kirk Borne:
    So we were having this conversation about how do you update the microcode on the chip in a remote sensor based upon data you’ve collected and then you upload new code back to the sensor so that as it’s collecting data, it’s doing the right thing. So it’s not quite as much detail as, Juan, you were just describing but it was the same general idea is that you need to update the computing that’s happening in the world environment. So NASA never called it edge computing but that’s basically what they were doing where they’d send these probes in deep space and they’d basically … You couldn’t wait for a command from Earth because the light travel time for the probe to report back to Earth, “Hey, I got a problem,” and then Houston to send back a reply would be hours. It needs to decide in real time.

    Kirk Borne:
    So a lot of the programs that NASA sent up there, probes, they had what they called Safe Mode. If something just go bonkers, it puts itself … It turns off all the sensors and puts everything in a safe operating condition so it stops doing all things that might be upsetting to the machine and just say, “Let’s turn everything down. Put it in low power mode. Wait for a command from Houston to tell us what to do next.”

    Tim Gasper:
    It’s like a timeout, right?

    Kirk Borne:
    A timeout. Exactly. And so that’s how they dealt with it, because they said, “We can’t put”… The things that are shipped up there in space are small, right? We’re only getting to the age now where we’re talking about sending really huge things up in space, but the initial probes, they were the size of a water bottle, okay, so they’re not very … Okay, a little bit bigger than that, but not much. So they had to have sort of limited resources on the ship and then have these ways of dealing with issues like this. So then they start talking about, “What can we do about sending reconfigurable chips up there? Things that we can reprogram in that remote environment?”

    Kirk Borne:
    So anyway, I’m just saying this because it’s not a new conversation. It’s just now we have capabilities we never had before to do a lot of cool things. We never talked about putting machine learning on chips like that. We weren’t talking about AI in the chip. We were just talking about instead of sampling every thirty seconds, sample the data every five seconds of this particular data. So we weren’t doing any rocket science. I guess we were, but-

    Tim Gasper:
    We weren’t doing data science. It was only rocket science, right?

    Kirk Borne:
    Yeah. It was only rocket science, that’s right.

    Juan Sequeda:
    A lot of the conversation we’re having now, it seems to me that it’s more about the sensors, right, the actual things, the internet of things but not all companies are dealing with sensors? I just got a bunch of data … I mean you were mentioning before, Kirk, like all … I have my e-commerce and we have e-commerce company and they want to go do something realtime very quickly and you said something like, “Let the web analytics.” That’s the edge right there. How does this whole notion of edge, intelligent edge and stuff work with companies who are not dealing with sensors or is this more around sensors and that type of hardware?

    Kirk Borne:
    Well, you’re right. If you’re not dealing with sensors it makes less sense, but more of us are dealing with more sensors than we care to admit, right, because there’s this thing called a smartphone and every employee has a smartphone and they tell where people are, where they’re moving, what they’re doing. I mean, there’s all kinds of information being collected there, right? In that case, some of that might be surveillance and we got some legal issues here but not all of it. So when people are shopping online, where are they shopping? Hey, if you’re shopping in this mall and I know that you’re shopping at that mall and my store has a sale, maybe I can just sort of drop an ad in one of those apps you’re using, right?

    Kirk Borne:
    So that location intelligence can be very powerful for even small stores, right? Then again, it’s edge intelligence because you got to know that the person is doing this thing at this place at this time or shopping for this particular product. So you need to have that extra contextual information besides, “Oh, maybe this person would like to buy a book on TinyML today.” Well, they’re actually at a restaurant. Maybe they would like to know what are the sales for these types of foods instead of TinyML books.

    Juan Sequeda:
    So this is a real good point because technically everybody is connected somehow to sensors because of at least their phone, right? So we always have this notion you’re going to be related to this intelligent edge somehow because if you’re dealing with phones. Now one of the things that worries me now is you look at all the apps I have on my phone and I’m like, “Wait. This is an app that deals with pictures and I have to go download this 300 MB app to go do some stuff.” It’s like really? Are people not even being trained, then, to go deal with … Think about being cognizant with the amount of space and time that they have on a phone?

    Kirk Borne:
    Well, I think we are in the era of small data, right, so this is one of the emerging trends in the Gartner hype cycle this year, small data, because we’re realizing that we need laser focused, targeted, personalized, hyper-personalized data. So hyper-personalization is one of the consequences of this Industry 4.0 we were talking about, this hyper-connectivity. Once we see all these connections, it’s like a big knowledge graph, right? We focus down on the very edge and nodes in that knowledge graph that matter in this moment in this time in this context and so what you really brought up to me is what I call the space-time continuum of data because no matter what your business is, something is happening at a place at a time, but it’s not just those coordinates, right?

    Kirk Borne:
    So I’m a physicist, so I learn about hyper-dimensional coordinate systems. We got time axis and the space axis. Albert Einstein correctly pointed out that this a space-time continuum. Those axes, space and time, are interchangeable in the right way. Well, we also have another kind of continuum of data … The characteristics, the dimensions that describe a customer or the dimensions that describe a product or the dimensions that describe the market or describe your company. Columns … Think of them as just columns in a database. You’ve got XYZ coordinates of something or a thing or a product, you’ve got the time or day or whatever, and we got these other dimensions. What’s the product? What’s the customer? What are they buying?

    Kirk Borne:
    So it’s a hyper-dimensional space. It’s a space-time continuum that every business, every company lives in. Now whether they want to explore it that way is up to them, obviously, and whether it makes sense for them to explore it that, again is up to them. I’m not saying it makes sense for every company to do this. But look at that hyper-dimensional space and saying, “When this thing happened, what else was happening in that space?” So I talk to people about what I call the stellar analytic score card and three of those dimensions … S-T-E-L-L-A-R. Is Edge, Location, and Related Entity. E-L-R. I won’t bore you with all the other ones. Team Analytics. Agile Analytics. I got some of those letters in there.

    Kirk Borne:
    But time … So edge computing is basically the edge at the moment of time. What’s happening at this moment in time. Location analytics is what’s happening at this location? For example, you might be at a sporting event and it’s important to know you’re at a sporting event before you get recommended a certain thing, all right? There’s also R, the related entity analytics, which is those other dimensions in the space-time continuum. What other things are like this in my data space? Products? Services? Other customers? Other behaviors? Other preferences.

    Kirk Borne:
    So looking at hyper-dimensional space, do the segmentation and the clustering analysis and those hyper-dimensions in space, time and related entity. Just your basic features and your feature engineering so you still need the data scientist to figure out what are the important features and in that space, now you’re not doing just edge computing, you’re doing intelligent edge because it’s all about where in time, where in space and where in that related entity space is something happening and what do I need to do? What action do I need to take next because of this?

    Tim Gasper:
    Mm-hmm (affirmative). You know, this idea of small idea and this idea of the space-time continuum of data both feel like not only are they very interesting in terms of spelling out where the intelligent edge is trying to drive us towards but also they feel intimately connected with these idea of the knowledge graph and the fact that knowledge and wisdom, if we go back to that analogy we talked about earlier, if it’s a pyramid, ideally the wisdom and the knowledge is actually smaller data. It’s the moments of intent. It’s the critical pieces of information that are the conclusions that are the things that you were trying to boil down to as opposed to the big ocean of everything. It’s seems like that’s … We’re trying to do a better job of getting to that top of the pyramid there.

    Kirk Borne:
    Yeah. So the knowledge graph itself can be large, right, but the part that you need, the key edges and nodes in that graph that say, “This causes that to happen and therefore when I see this happening I need to do this to cause the better outcome.” So knowledge is knowing how to navigate through that knowledge graph just like through your data graph to say, “What are the things that I need right now in this moment to do the right thing?”

    Tim Gasper:
    Mm-hmm (affirmative). That could be a lot in the knowledge graph.

    Kirk Borne:
    Can I just say something about e-commerce stores?

    Tim Gasper:
    Yeah, sure.

    Kirk Borne:
    Back in the day when I was working on these really big data astronomy projects and we invited this guy from eBay. I won’t name him, because he’s still pretty active in the world, but anyway, one of the probably smartest people in analytics I think I’ve ever met in my life. I mean, he just told the most amazing years. This was like fifteen plus years ago.

    Kirk Borne:
    So he was the director of analytic at eBay and he said, “You know, we do AB testing on eBay,” and of course at that point as an astronomer fifteen, twenty years ago I wasn’t thinking about what AB testing was so he explained how they change the font sizes and the colors. They change the size of the image of the product that’s being sold, the location on the page, where do they put the price, where do they put the bids, latest bid. They move things around all over the place to see what really works and he said, “So we do AB testing. We try the change, so that’s the A group and then we have the control group, the B group, where we don’t make the change and see what the response is to the change.”

    Kirk Borne:
    So I’m thinking now, “Okay, I get it now. So what are you telling me here?” He said, “Well, guess how many AB tests we do everyday on the eBay e-commerce site.” I said, “I have no idea. A hundred thousand.” I was just being real extreme, gross. It’s not going to be a hundred thousand so I just said that number. He said, “Thirty million.” I said, “Wait a minute. Thirty million per day?” He said, “Yes. Thirty million tests per day.” We’re talking edge intelligence here. They’re finding out in real time, thirty million times a day, what works and what doesn’t work on their e-commerce. Now granted, they’re eBay and they have an enormous e-commerce site but nevertheless it just completely startled me and this was very highly irrelevant because I was working on this astronomy project which is just going to start … It’s actually going to start next year, the survey. It’s taken twenty years to build the telescope in Chile, by the way. [inaudible 00:47:11]

    Kirk Borne:
    It’s going to collect images of the sky, the whole southern sky, every twenty seconds for ten years during the night of course, not daytime, every twenty seconds and it’s going to image a hundred million objects every twenty seconds and because the night sky is variable. Stars vary, black holes vary, galaxies vary, all kind of things vary. They move, they change brightness. There’s going to be like twenty million events a night. That’s nothing compared to the thirty million things this guy was testing. At twenty million times a night, the astronomers … I no longer work on the project but I did for many years. Twenty million times a night, you got to make a decision. Do I take an action or not take an action when I see this unusual thing happening in the night sky? An action might be to notify someone or do whatever or do whatever.

    Kirk Borne:
    But anyway, the idea is that that’s edge intelligence because you can’t send it back. What they going to do is they’re going to ship all the data back up, by fiber optic cable, under the Atlantic Ocean, back to the United States, a process. It’s going to take a year to process a lot of data. So the National Science Foundation, which funded this project along with the Department of Energy for different reasons, a billion dollars of your tax dollars at work, one of the requirements on the project is within sixty seconds of an image you have to alert the world what’s going on in the sky. Ten million times a night. You have to publish an alert, every single one of those events, so you have to define, “What do I mean by an event? What do I mean by something changing? What do I mean by something being interesting?” Okay. Now we’re talking domain expertise to the max here, but it all starts with the data bits. The data bit that is an outlier, there is a trend change, there’s a color change, there’s a movement change. Oh my gosh.

    Kirk Borne:
    So we can define in words that any human can understand. It got brighter. It moved. But then say, “Okay, why is that important to an astronomer?” Now you have the domain expertise that defines sort of the knowledge that’s in that particular piece of information that you’re sending to the world ten million times-

    Tim Gasper:
    I mean, you’ve got domain knowledge here. You’ve got big data here. You’ve got a lot of moving pieces that are part of this.

    Kirk Borne:
    Yeah, so what you can do is you can subscribe to their email stream and get ten million emails a night for ten years, okay? No, I’m kidding. I’m kidding. You don’t do that. So what they’re creating is basically a data catalog and I think you’ll love that phrase since we’re on the Catalogs and Cocktails. They’re creating a data catalog where you can subscribe. I want specific types of events to be alerted. I’m a super nova person, let’s say, so a super nova exploding star has a very specific characteristic so when you see this characteristic, I want to be notified and that’ll happen maybe ten or a hundred times a night so I don’t mind getting ten or a hundred emails a night.

    Kirk Borne:
    If you see this kind of fast moving object at this brightness, it might be the killer asteroid that’s going to wipe out civilization. I want to know about that one.

    Tim Gasper:
    Fast, please.

    Kirk Borne:
    I’m going to check my 401k, make sure my children are safe. Oh, wait a minute, all civilization’s going to be wiped out so what difference does it make? I want to go get that steak dinner I’ve been denying myself for the past five years.

    Juan Sequeda:
    Oh, Kirk, this has been an awesome discussion. We could just sit, talk to you for hours. I got so much stuff, but let’s go into our lighting round here. So quick questions, yes and no answers and I think we’re going to just ask you to predict the future, Kirk, so I’ll go first.

    Tim Gasper:
    Hopefully your space-time continuum machine is working well.

    Kirk Borne:
    Let me check.

    Juan Sequeda:
    So enterprise knowledge graphs will become the top buzz words in the next couple of years. Yes or no?

    Kirk Borne:
    Say the question again? My time machine wasn’t listening.

    Juan Sequeda:
    Okay. Now you’re listening, ready. Enterprise knowledge graphs will become a top buzz word in the next couple years.

    Kirk Borne:
    Yes.

    Juan Sequeda:
    Tim.

    Tim Gasper:
    Perfect. All right. Number two. So we didn’t have a chance to talk too much about ML and model ops and that sort of thing, but blank ops sure is kind of the thing these days, right? Blank ops, devops, now anything ops. The blank ops trend will continue to become more prevalent and more over-used over the coming couple years here.

    Kirk Borne:
    If you emphasize the word over-used, yes. Prevalent, therefore yes.

    Juan Sequeda:
    All right. I’ll follow with my trend still on knowledge graphs. So knowledge graph at the edge will become a thing in the next five years.

    Kirk Borne:
    Good question. I’d like to be optimistic, but five years might be too soon for all of that power at the edge.

    Juan Sequeda:
    So that’s a no.

    Kirk Borne:
    Prove me wrong. I want to be proved wrong.

    Juan Sequeda:
    All right.

    Tim Gasper:
    That’s a challenge, Juan.

    Juan Sequeda:
    All right.

    Kirk Borne:
    I think it’s wonderful if that happens, but prove me wrong.

    Tim Gasper:
    I appreciate your honest answer there, because you know, that one I was also thinking about. I was like, “Hmm, ten years [crosstalk 00:52:26].”

    Kirk Borne:
    I’m not saying I don’t want that. I want that. Just like I want that steak dinner, but it ain’t going to happen.

    Tim Gasper:
    Yeah. Once we do that, we can really get this intelligent edge going.

    Tim Gasper:
    Okay. Final lighting round question here for you, Kirk. Will all companies eventually have their digital twin?

    Kirk Borne:
    No. It doesn’t make … There’s a lot of companies it doesn’t make sense.

    Juan Sequeda:
    I love that. So whoever’s listening, if you’re thinking you need to have a digital twin initiative or whatever, really, really think about it if you need a digital twin. That’s the honest, no BS, right? I love this stuff.

    Juan Sequeda:
    All right. Takeaway time. TTT. Tim, take it away with our takeaways.

    Tim Gasper:
    Oh, my goodness.

    Juan Sequeda:
    You go first.

    Kirk Borne:
    Tim’s Takeaway Time.

    Tim Gasper:
    Tim’s Takeaway Time. T3. There’s so much ground we covered here. I think one of the things that really came across to me was your use of the phrase small data. That moment kind of struck me and I’m excited that that’s becoming a little bit more of a thing because I think when we talk about knowledge, we talk about wisdom, we talk about what’s the point of all this big data, it was to actually unlock the next step here, which is really the small data. That is … It’s like it all came back. It’s all full circle. We just wanted to have better small data and we had to pass through the big data to get to it.

    Tim Gasper:
    When you think about all the things that we’re doing and pushing to the edge around intelligence and around the AI things, your example of the NASA probes and things like that, it seems like it all goes around this idea of trying to be more dynamic, be smarter and actually get back to small data at the edge. That’s one of my big takeaways here.

    Juan Sequeda:
    And mine, quickly, are I think Industry 4.0. It’s all about hyper-connectivity. Your knowledge is being transported. It’s all being connected everywhere and this honest, no BS definition of the digital twin. It’s a computer model of the real thing, so if there was a problem you can go press play and replay it and see where the glitch happened so you can go simulate.

    Juan Sequeda:
    The intelligent edge is a system to respond at the point of data collection and basically let’s put the smarts where the data is and hey, everybody potentially is part of it, right? We all have our phones and our phones are our sensors and I love the whole large amounts of small data. I think that’s the way we need to go think about it.

    Juan Sequeda:
    So we got a couple seconds here left. Kirk, throw it back to you. What’s your advice and second, who should we invite next?

    Kirk Borne:
    Well, my advice ties right back to our toast at the beginning and that is about education, start of the school year. My advice is to stay a lifelong learner. No matter where you are or what you’re doing, always be in learning mode. My background’s physics and astronomy. I learned physics and I did colliding galaxies and black hole accretion disks, if any of that makes any sense to people, years ago and now I’m advising companies in medicine and cybersecurity and behavior analytics and all kinds of stuff and it’s because I’m kind of special person but I just never stopped learning. I just always was learning new things. It’s opened the doors for many opportunities in my life and I encourage people to always do that. Just be on the lookout for new things to learn because the door might open not this week or next week, but many years later. I started learning about machine learning in 1997, 24 years ago, and I started my own data leadership group LLC four months ago. Sometimes it takes a little while but now I work at a startup, Data Prime. I should mention that. I work at a startup which is doing AI in all kinds of cool places and that’s what I’ve been doing the last four months. Anyway, so-

    Juan Sequeda:
    Who should we invite next?

    Kirk Borne:
    I would suggest if you haven’t already had Kate Strachnyi on, she is just a fireball of building up the data community. She runs the DATAcated Conference three times, four times a year. I don’t know. Most conferences, people run them one year and they’re exhausted. I used to do that at NASA. We ran the Data Center conference every two years. Kate runs the DATAcated Conference like four times a year. I think she’s just amazing knowledge, community building person. I’d love to have you have her on and watch that episode.

    Juan Sequeda:
    Love it. Kirk, this was awesome. Thank you so much for this discussion talking about Industry Revolution 4.0, intelligent edge and there’s so many more buzz words we can do this. We should do this again-

    Kirk Borne:
    I’ll say. All right.

    Juan Sequeda:
    … find more buzz words.

    Juan Sequeda:
    Kirk, cheers. Thank you very much.

    Kirk Borne:
    Cheers.

    Tim Gasper:
    It’s been fantastic. Really glad to have you here.

    Kirk Borne:
    Toast to you. Thanks for the opportunity. Really loved it.

    Tim Gasper:
    Awesome.

    Juan Sequeda:
    Awesome.

    Tim Gasper:
    Cheers.

    Enter Content Here.