>>Okay, welcome back everyone. To the cubes coverage of ADB has re-invent 2021. I'm John fare with Dave Nicholson. My cohost we're here exploring all the future innovations. We've got a great guest we'll lead negam who's the EVP executive vice president chief research innovation officer cap, Gemini engineering will lead. Thanks for coming on the cube. Thank you. So I love the title, chief research, innovation engineering officer. >>I didn't make it up. They did. >>You got to love the cloud evolution right now because just more and more infrastructure as codes happening. You got this whole data abstraction layer developing where people are starting to see. Okay. I can have horizontally scalable governed data in a data lake. That's smart, someone intelligent and use machine learning. It seems to be the big trend here from AWS. More serverless, more goodness. So engineering kind of on the front lines here kind of making it happen. >>Yeah. So, uh, the question that our clients are asking us is how do these data center technologies moving over into cars, planes, trains, construction, equipment, industrial, right? And you know, maybe two decades ago it was called IOT. Uh, but we're not talking about just sensors, vertical lift aircraft, uh, software-defined cars, um, manufacturing facilities as a whole, you know, how are these data center technologies going to impact these companies? And it's not a architectural shift for say the Evie, the electric vehicle, many OEM, it's a financial transformation, right? Because if they can make their vehicle containerized, uh, if they can monitor the cars, behaviors, they can offer new types of experiences for their clients. So the questions we were asking ourselves is how do you get the cloud into the car? >>Yeah. And software driving, all that. So you've got software defined everything. Now you've got data-driven pun intended with the cars cloud everywhere. How does that look? What are the concerns, obviously, latency moving data around. They got outposts. Am I moving the cloud to the edge? How are you guys thinking? How are customers thinking through the architectural, I guess foundational playbook? Is there one? Yeah. >>I, you know, coming into this, I did ask my, my son, the question is hardware or software more important. And then he, you know, he's not, and he said, you know, we're coding our way out of hardware. It was very interesting insight software rules. That that is for sure. But when we're talking about physical products and these talking about trillions of dollars of investments going into green energy, uh, into autonomous driving into green aviation. So we're not, it's not just the matter of verse here. We're dealing real physical products. I think though the point for us as engineers or as an engineering businesses, how do you co-design hardware and software together? What are the questions you to ask about that machine learning model being moved over from AWS? For example, into the car, is the Silicon going to be able to support the inferencing rates that are required right. In real time and whatnot. So some of the things like that, >>Well, that's been a, it's been an age old battle between the idea that, uh, the flour that's nurtured in a walled garden is always going to be more beautiful than the one that grows out in the meadow. In other words, announcement, uh, at, in Adam's keynote, talking about advances in AWS Silicon. So what's your view on how important that is? You just sort of alluded to it as being important, the co-development of hardware and software together. >>Yeah. We're seeing product makers again, think, you know, anybody from a life sciences company building a digital therapeutics product, maybe a blood glucose monitor or, um, an automotive or even an aerospace, uh, going direct to Silicon asking questions around the performance of the Silicon and designing their experience around that. Right. So, uh, if they need a low latency, low power efficiency, green networks, they're taking those questions in-house or asking those questions in house. So, you know, AWS having a, sort of a portfolio of custom or bespoke Silicon now as part of the architectural discussion. Right? And so I look around here, I see a lot of developers who are going to have to get a little bit more versed in some of these questions around, you know, should I use an arm based chip? You know, do I use this Silicon partner? You know, what happens when I move it into the vehicle? And then I have over the air updates, how do I protect that code in an enclave in the car just to continue to use the so there's are a lot of architectural questions that I don't think software engineers typically ask when they're just dealing in the cloud. Uh, although at the end of the day over time, a lot of these will be abstracted from the developer to some degree, you know, that is just the nature of the game. >>It reminds me of the operating system theory of system software meeting hardware. And because you have software developers just want to code now, you're saying, well, now I'm responsible hardware. Well, not if it's programmer, was there a hard top two it's over, these are big questions and important ones I think is we're in a major inflection point, but it comes back down to, you mentioned aerospace space is the same problem. You can't send that break, fix engineer in space. Right. You've got software now. So you've got trust that security supply chain who's right. And who's doing the hardware now you've got the software supply chain. So a lot of interesting kind of, yeah. >>So you, you, you know, you check them off, back in into it, the supply chain problems with Silicon, and there are now alternatives to try and get around the bottlenecks using high-performance computers versus hundreds of ECS and a vehicle allows you kind of get away from the supply chain shortage. Uh there's you know, folks moving from one architecture to another, to avoid kind of getting locked in and then of course creating your own Silicon, or at least having more ownership over the Silicon. I think suffer defined systems, uh, are the way to go regardless of the industry. Uh, so you're going to make some decisions on performance, characteristics of the hardware, but ultimately you want a software defined system, so you can update it regularly. >>I was talking with doc some of the top hair executives. I talked to, um, the marketplace guys here, Deepak, uh, over at the, here at Amazon and containers comes up. You start to see a trend in containers where you see certified containers because containers are everywhere. You can put malware and containers. So, you know, think about like just hacking software. It's a surface area now. So you bring the software security model in there. So to see this kind of like certified containers, I can imagine certified infrastructure now because I mean, what's a processor, it's just a hardened top to a PC. Now you've got the cloud. If I have hardware, how do I know it's workable? How do I trust it? You know, how could it not be hacked? I don't want my car to be hacked and driven off the road. >>So, so, um, when you're dealing with a payment system or you're dealing with tick-tock different than when you're dealing with a car with life consequences. So we are very active in the software defined transformation of automotive. And it's easy to say, I'm just going to load it up with all this data center technology, but there's safety criticality issues that you have to take into considerations, but containers are well suited for that. Just requires some thought. I mean, my excitement, enthusiasm about this product engineering is if you just take any of these products and, and apply them into a product engineering context, there's so much invention and creativity can happen. Uh, but on the safety side, we're working through security enclaves using containers and hardware based roots of trust. So there's ways around, you know, malware and bad actors at the edge. Um, >>What's your, what's your take on explainable AI? Why got you might as well ask because this comes up a lot, explainable AI is hot in college right now, AI, that can be explained. It's kind of got some policy, uh, to it. What's your thoughts on this AI trend? Cause obviously it's everywhere. Um, I mean, what is explainable AI? Is that even real or how do you explain AI? Is that democratized? >>Yeah. Computer vision is a great example. I think to bring it to life I'm all of the audience probably knows this, but you could, you know, you can tell your kid that this is a cat once and they'll know every single cat out there is a cat, but if you, you, you need a thousands of images, uh, for a computer vision model to learn that this is a cat. And even, you know, you can probably give it an example, um, out of say a remote region of the world and it going to get confused. So to me, explainability is about adding some sort of certainty to the decision-making process. Um, and when there's a, some confusion, be able to understand why that happened. I think in, in automotive or any, even in quality assurance, being able to know that this product was definitively defective or this pedestrian definitively did cross the crosswalk or not. You know, it's very important because it could, you know, there are, there are consequences. So just being able to understand why the algorithm or the model said what it said, why did it make that judgment is super important, super important. >>So I've got to ask you now that we're here, re-invent from your engineering perspectives, you look at the landscape of AWS, the announcements. What, what, how do you think about it to other engineers out there trying to, uh, grok all the technology who really want to put innovation in place, whether it's creating new markets, new categories or innovating their existing business, how do you grab the class out and make it work for you? I mean, from an engineering standpoint, how do you look at AWS and say, how do I make this work better for me? >>Uh, so I mean, over the years, I, um, I think it's true. AWS has started to really look like a utility, you know, the days where it was called utility as a service. And, um, you know, I, I, I did attend a workshop on, I think it was called LightSail or something like that, but they are simplifying the way that you can consume this infrastructure to a degree that is somewhat phenomenal. Uh, and they're building any, yeah, they continue to expand the ecosystem. Um, so I mean, for me, it's, it's a utility. Uh, it's it's, it's, it's, it's, it's consumable. Uh, if you got an idea pick and roll your own. >>Okay. So back back to the, uh, the concept of AI and explainability, uh, one of my cars won't allow me to unlock certain functions because of the way that I drive. No one needs to explain to me why, because I know what I'm doing wrong, but I'm still frustrated by it. So that that's sort of leads to kind of the larger philosophical question to you about what you're seeing, where are we in this kind of leapfrog, constant pace of the technology exists, but people aren't culturally ready to accept it because it feels like right now to me that there isn't anything we can't do with cloud technology from a technical perspective, it can all be done. Swami's keynote today, talking about integrating all sorts of sources of data and actually leveraging them in the cloud. Um, technically possible yet 85% of it spend is still on prem. So, so what's your thought there? What are the, what are the inhibitors, what are the real inhibitors from a technology perspective versus the cultural ones? Uh, setting aside my lack of, uh, adherence to, uh, to driving lawful >>I industry by industry. I think in, um, you know, if you're trying to do a diagnostic on an MRI in an automated way, and there's going to be false positives, false negatives, and yes, we know that yeah, we know that there's going to be a physician participating in the final judgment call. Um, I think just getting a really good comfort level on the trustworthiness of these decision points, um, is really important. And so I don't blame folks for being reticent about, you know, trusting or, or asking some questions about, does this really work and are these autonomous systems as they become more and more precise, are they doing the right thing? Uh, I think there's research that has to be done on agency. You know, am I in patrol? What happened? Did I lose control? I think there's questions around handoffs, you know, and participation in decision-making. So I think just overall, just the broad area of trust and, uh, the relationship between the participants, the humans and the machines still. I think there's some work to do, to be honest with you. I think there's some work to do maybe in a manufacturing facility where everything's automated, you know, maybe it's a solved problem, but in an open road, when the vehicles driving, you know, in the middle afternoon, you know, you probably should ask some more questions. >>Well, I want to ask you what we got a couple of minutes left, real time data near real time, real time, always a big, hot topic. Seeing one more databases out there in the keynote today from Swami real-time are we there yet? How are we dealing with real-time data, software consuming the data? It comes to cars and things that are moving real time versus near real time. It could be life or death. I mean, this is big time. Where are we? >>So, um, I was trying to conduct a web conference. I won't tell the vendor because it has nothing to do with the vendor. Um, and I couldn't get a connection. I couldn't get a connection at reinvent. I just couldn't get it. I'm sorry guys. I can't get it. So I, you know, so we talk about real time talking about real-time operating systems and real time data collection at the edge. Yeah. We're there, we can collect the data and we can deploy a model in, you know, in the aircraft on the train to do predictive analytics. If we got to stream that data back home to the cloud, you know, we better figure out how to make sure we have a reliable and stable connection. 5g is a, you know, is, is, will be deployed, right? And it has ultra low latency, uh, and can achieve those types of, uh, requirements. But, uh, you know, it has to be in the right setting, right? That's to be the right setting and a facility, uh, very well controlled where you understand the density of the cell sites, small cells sound cells, and you really can deploy a, uh, a mobile robot, uh, wirelessly. Yes know, we can do that, but you know, kind of in, in, in other scenarios, we have a lot of ask that question about >>With the connections and making that false, huh? Well, he, thanks for coming on. Great insight, great conversation. Very deep, awesome work. Thanks for coming on and sharing your insights from cap Gemini. We're here in the cube, the worldwide leader in tech coverage live on the floor here at re-invent I'm John fare with Dave Nicholson. We write back.