 Live from Barcelona, Spain. It's theCUBE, covering Cisco Live Europe. Brought to you by Cisco and its ecosystem partners. Welcome back to theCUBE's live coverage of Cisco Live 2019 here in Barcelona, Spain. I'm Stu Miniman, my co-host Dave Vellante and John Furrier is here with us, wall-to-wall coverage going through all the areas of what Cisco is covering. Their transformation become more of a software company. Help us dig into a very exciting area. We have Nikola Rorsitz, who's the lead strategic AI program at Cisco. Nikola, thanks so much for joining us. Nice to meet you. All right, so AI is something that, you know, pervading everything that we talk about. We definitely have the buzz and the hype in the industry. You sit at the nexus of all the different areas inside of Cisco. So, you know, give us a little bit about, you know, your role inside the company. You've been there about two years and kind of the scope of what you do there. AIML has a long history inside of Cisco. We have not been very vocal about it, but it's been used throughout the company. And we once put together a map of all these things. This was one of my first activities and I was, wow, it's amazing. In all of these products, we have some element of AIML. And this showed also that we have a very pragmatic approach to AIML. It's not this killer robots or you name it. It's more like, okay, how do we use this to solve specific problems? Yeah, so, you know, I think back, you know, definitely analytics, you know, when you talk about networking and flows, you know, there's always been lots of data and I've had tools to be able to access there. When I talk to most people in the industry though, there definitely is something new and different, you know. You know, AI is not new. We've been talking about artificial intelligence for about 150 years, you know, machine learning, there's been movements on there. So, maybe you can give us, you know, what's the same as what we've had before and what's new and different about, you know, the error that we're in today and the products that just goes bringing to market. It's a spectrum. On one side you have analytics, on the other side you have AI. And basically the fundamental difference is that in analytics, usually it's accepted that it's a person that creates the rules, that draws the rules. On your hand, in AI, it's a computer that draws the rules. So, in between those, there is a gray zone in which you evolve and there's more and more done by the machine. But it really depends from the specific area in which you want to apply it. And it is true, we are moving more and more towards artificial intelligence, more done by the machine. And the reason is clear, we have more and more data and up to a point in which AI will be the only option to do business because everything needs to be automated. I want to ask you, as an AI expert, when I talk to security experts, I always ask them who was your favorite superhero because when they were little kids, they dreamt about saving the world. So, as an AI expert, did you think about artificial intelligence or whatever you called it back then as a child? How did you get into and interested in artificial intelligence? I've been always fascinated by how the brain works. I did my PhD in neuroscience and physics because of that, because when I was a kid, suddenly I thought, well, how do we think? How is this possible that we create stories in our mind and that we dream at night and wake up? And then slowly, little by little, I kept on asking a question, how do you make this into a technical solution? How do you engineer something like this? And then started looking into computers, well, it's not like our brain works, so there's a difference. And now we're sort of like coming slowly together, despite having started from very different paths from the new phone-knowing machine and so on, now we're moving more and more to our brain-inspired technologies. And so you're seeing those two worlds come together. I mean, I was under the impression that, despite that vision, that today's AI anyway is really a lot about maybe automating processes, robotic process automation as an example, but you're talking about a world that much more mimics the human brain as we understand how the human brain works. Is that correct? It is correct in the sense that, you want to mimic certain fundamental capabilities. So intelligence is about perceiving information, storing knowledge, thinking, and adapting. And you need all these four components to create a truly intelligent system. And you don't need to replicate individual neurons to make this happen, but at least understand the fundamental principle behind it, what's the computation like? And as you go along, because we're in a business, we need to find concrete solutions to business challenges, and therefore we apply whatever we need from these principles to make something out of it. What are the things that humans can do today that computers have trouble doing, and how is that changing? One of the clearest things is that computers are not able to think. They are still executing machines, they don't have a representation of what it means to do whatever they're doing, to solve their problems. And one of the next steps, which the researchers are very interested in now, is trying to understand the context in which a machine operates. Now, if you ask a machine to do a certain task, and it can fail miserably, because it's not able to connect the dots between different elements of the context. And part of the reason is that context, context on information is so broad and large, you have so much data, so which one do you pick? And this is still an unsolved problem. Yeah, Nikola, help us understand how we should think about Cisco when it comes to AI. People hear about Facebook and Google and IBM with their Watson pieces there. Obviously, things like scale of networks and managing infrastructure and moving to some of these multi-cloud environment theme. A natural fit for Cisco, but how should I, as a user, be thinking of when do I come to Cisco? How does AI and ML fit into what Cisco does compared to some of those other software and enterprise IT providers? So doing AI at Cisco is super exciting, because it's still an open field. AI ML for networking is something that has not been solved yet, and there are other areas where other companies operate in, they're much more advanced. Well, for us, there's lots of room still to innovate. And for us, it's a business opportunity, it's a tremendous business opportunity. Some market research talks about $1.2 trillion that's got to be captured by companies that adopt AI compared to those who don't. But for Cisco is really a necessity, because data is going to flow more and more through our networks. How do we handle that? And what people don't realize, in general, compared to what's out there, that ML for networking is a different beast. For one, the data is different, and often it's encrypted. So how do you do AI on encrypted data? And every network is unique. And these are two fundamental differences that force us to be creative and to pioneer new ways of doing AI. And this is super exciting. Does open source play into the activities that your different product groups are working on? So in general, AI has been driven by a very lively AI community in the open source world. And then the question comes, when we talk to our partner and customers, how do you bring these solutions to production? Because certain packages of open source cannot be applied directly. And this is one of the main pain points of the IT teams and data scientists and data engineers. I want to ask you about the black box phenomenon. As a human, I can look at a dog and I can see it's a dog immediately, but I can't really explain how I know it's a dog. I can, but I could be describing another animal. Computers can figure out, but we don't really know exactly is there sort of a black box inside. Is that a problem? Do we need to make AI more transparent, or is it increasingly going to be a black box that we just trust? What are your thoughts on that? It depends on the situation. You came here by plane. Do you know exactly how the plane works? I don't. I sort of know the principles, but I trust the industry, the regulations, everything that they have checked everything. And to me, it's a sort of a black box. However, if there are certain things that I have to go under like surgery, so I want to know exactly what's going on. And the same thing here in AI. So there's the black box phenomenon. You don't know exactly how does this work? And on one side, I understand it and it makes sense. You want to be sure that you know what's going on. On the other hand, sometimes you want a result and you don't really care about exactly how it works because ultimately the risk is not that high. And so you have to really think about what kind of risk management, how deep you want to look into it. And the problem of transparency has been researched a lot because of course there are certain phenomena that touch the social sphere. And there we have to be careful. When it touches private data, how are its private data hand lands on? That is very important, of course. Yeah, Stu and I often, when we do these conversations, John as well, we often ask ourselves, okay, how far can we take AI and how far should we take AI? So maybe a couple of examples, if I may. Do you expect that within, let's say the next 10, 15 years, that machines will make better diagnoses than doctors? Oh, they already do. They already do. The research has been shown that in certain cases, specific cases, they have better accuracy. However, to bring that again into production at the level that we go to the hospital and there's a machine that helps us diagnose, well, we're still at least some years away because there's all the process of certification. And it must be added that on one side it's really about augmented intelligence rather than artificial intelligence. The machines will help us diagnose, but then the responsibility should stay with the human. Another question we'd like to ask is around automobiles. Do you think it will become the exception rather than the rule that individuals will own and drive their own automobiles? It's going to be the exception in the future that there's going to be a ownership and a driving, active driving. It's interesting because it's going to become like when it started, like a pleasure to drive. You drive because you want to drive. You're going to drive those hills up and down and really enjoy it. Otherwise, if you go on your commute, you have work to do. Yeah, I still have a stick shift. You're going to enjoy it. Now, I got to ask you, so the likes of Elon Musk and Hawking's have said, you know, projected that AI is a bad thing. It's going to, machines will take over the world. I don't sense that you're of that mindset, but what are your thoughts on that, those dire predictions? Are they ultimately going to come true or do you feel like they're overblown? Who knows, but it's hard to forecast, but what are your thoughts on that? It's important to acknowledge these forecasts of a dire future because AI is capable of lots of things at scale and this is the key differentiator. So whatever you can do, you can do it at scale automatically, things on their own. So it's more than predicting a dire future. It's like I'm wanting to say developers, managers, be careful of your choices because they're going to have an effect at scale and this is not just an AI-rated effect. It's really like a technology-rated effect because also if you look at the AI today, there are lots of pieces that come together. Lots of pieces that come also from the big data era and now they're being transformed and you add a little bit of AI in the mix but to make it work, there's a lot around it. So AI is the culprit because of the science fiction history and everything but ultimately the ability to do things at scale automatically, that is really where we have to be careful. So Nicola, what should we be looking for when we watch Cisco going forward for the next couple of years in this space? What are some key milestones that you think will come to reality? Well, we're going to release products that have more and more AI into it and the whole industry will evolve and have a better understanding of what's possible and whatnot and AI and Cisco revolves around three axes. One, infrastructure, tools that fit and unique data. Infrastructure is how do we deal with increase of data? Create these future-proof networks. This is like our core business. The tools that fit is that we provide end-to-end solutions to our customers and partners so that they can implement their AI ML strategies and this is a really interesting topic because AI ML is moving into the enterprise and other organizations but it's still in the early stages because of all these operational challenges which we at Cisco are very good at solving. And the third point is unique data, unique in terms of volume, breadth and type of data. This is where on one side we have systems that work at scale but also we have kind of data that can be used by our customers to better understand their own business. All right, well, Nikola, I really appreciate you giving us a nice overview of all the areas for AI. Dave Vellante and I are still humans here doing the interviews here until the robot take over all of our jobs. Until then, thanks as always for watching theCUBE.