Bob Friday: 0:01
The speed at which things are changing. That’s a good point. Yeah, right, I mean I think over my 40 years keeping up with the transition that’s right with software engineer to wireless engineer, system engineer Forty years ago, if you bet a dollar that in my career as chief AI officer, you guys would be all billionaires right, oh, totally yes yeah it’s like you know who would ever have thought chief.
Michael van Rooyen: 0:19
AI officer. Today I’m meeting with the co-founder of Mist and now the chief AO officer and CTO of Enterprise for Juniper Networks, bob Friday. For me, it’s quite an honor to meet Bob. He’s got quite a legacy history. Not only does he have and hold 15 patents in various technology, particularly wireless, he’s also recently won the 2024 Lifetime Achievement Award from the Wi-Fi Awards. Welcome, bob. Yeah, great to be here. Thanks for having me. Yeah, thanks for visiting in Australia. I appreciate the time and also just want to say thank you again for hosting us last year in the US and spending some time to really understand the MIS story a bit further. And certainly you’ve got an amazing, amazing history and we just want to talk a bit about that today and really think about the conversation around Juniper, mis and the history, et cetera. For me, bob, when I have a look at your journey in the radio frequency RF world, it’s really fascinating. What initially sparked your interest in that field and how did it shape the rest of your career?
Bob Friday: 1:12
Yeah. So when I started this adventure back in the 80s, it really all started around the FCC and unlicensed spectrum in 1985. And so when I came out of school I actually started as a software engineer and I started at this startup, metricom Ricochet, and we were really on the mission to build wireless mesh networks for utility companies. And so I got in the middle of the startup. I was supposed to be the software mesh engineer doing routing stuff. It turned out there was no one there. You know questions where are we going to get this radio from? And so a bunch of my amateur ham friends thought they were going to buy these radios from Japan or something. It turned out that I started building the radios faster than the software guys, wow. So back in those days there was something called PCAD. I taught myself how to lay out boards. Back in those days there was no internet, so you had to go down to the library to buy books to learn something new, and that’s where I started my wireless careers. It’s almost a competition with the software guys.
Bob Friday: 2:02
Right, right, right I taught myself to basically lay out a board a radio, build it, test it in six weeks.
Michael van Rooyen: 2:08
Wow.
Bob Friday: 2:09
So that was kind of the beginnings of the wireless adventure at Metricon and Ricochet. One thing I tell young kids at college and everything is if you want to learn something, really startups are the best place if you can handle it.
Michael van Rooyen: 2:22
Yes.
Bob Friday: 2:25
And if you look at new colleges, almost all these colleges have an entrepreneurial program, and so I always encourage kids like if you have the opportunity, find out if you can actually handle a startup. If you can handle it, it’s probably going to be the best experience you’re going to have in terms of learning something new. Yes, If you put your mind to it, you can tackle almost anything from scratch. Right, Because I wasn’t a wireless engineer back then, but I was able to teach myself fast enough. Stay ahead of the software guys.
Michael van Rooyen: 2:48
And that introduction to RF was it? You know, you were interested in RF radio or you joined the startup and then they kind of pushed you into that area. You talk about ham radios. Were that more your friends playing with hams? Like, where did the RF interest become.
Bob Friday: 3:00
You know, for me, wireless was like it was black and white. There’s laws of physics. There was no debate about right or wrong when you did path loss analysis. It was straightforward equations, right. You’re just learning laws of physics and putting them to use.
Michael van Rooyen: 3:13
Off the back of that, that deep dive. And I really like your point about startups, right, because you’re pushing limits in new areas. That’s great advice for young people who are getting into the industry or wanting to develop. You went through yourself and you’re from that area of America already. How did you end up in Silicon Valley?
Bob Friday: 3:26
I was born in Northern California. I went to school at Georgia Tech in Atlanta, on the East Coast. So when I got done with Georgia Tech, I decided it’s time to get back to the West Coast.
Michael van Rooyen: 3:35
Yes, and then from that you founded Airspace. Tell me a little bit about Airspace. What was the motivator behind it? And then off the back of that wireless LAN controller is where Airspace really started. And then bringing that product to market. And then obviously, cisco becoming very drawn to the product.
Bob Friday: 3:49
So if you look back, you know those days Metricon, ricochet I was a little young back then.
Michael van Rooyen: 3:53
I was young, naive.
Bob Friday: 3:55
If you look at Metricon, ricochet, we were basically both a service provider and the vendor, right? You know, we had raised like one and a half billion dollars and the vision was we were going to build the technology and we were going to build a nationwide infrastructure like a service rider. Looking back on it, that’s probably a little naive. Both those things are big tasks building technology and the service rider thing. So when the internet bubble burst, that’s when we saw the opportunity around aerospace, because we saw that, hey, we had enterprise customers out there and coming from Metricon and Ricochet, to me they look like service writers. We had these very large customers and they were basically trying to maintain hundreds, thousands of radios, thousands of users.
Bob Friday: 4:34
And what does that sound like? That sounds like a service writer, right? For those who remember back then, microsoft was probably one of the biggest companies deploying Wi-Fi back then and they were just getting to that point where, you know, running their network with scripts and everything has become unmanageable. And you look at the controller architecture, it looks a lot like an SP base station type of architecture, right, and that was kind of the theory behind the original Airspace was bringing that SP architecture into the enterprise space to help them manage all these access points and all these clients that they were trying to take care of back then.
Michael van Rooyen: 5:07
So it’s really solving the problem of this sprawling wireless deployments. You know how do we configure and manage them. You know you took your experience from building the previous capabilities around an SP type service. How do we put that into an enterprise type environment to help manage, maintain, run wireless effectively from that point of view? Then you ran that organization for a while and then Cisco came along, was obviously very interested in it because it was a market they wanted to chase, of course, with wireless becoming the new mechanism. So tell me about that acquisition with Cisco, because obviously Cisco acquired Airspace. Was there some lessons that you learned from it about scaling technology, startups integrating with a big organization, et cetera?
Bob Friday: 5:40
A couple of interesting things I learned on that adventure, especially at Mitchicon Ricochet. One of interesting things I learned on that adventure, especially at. Mitchicon Ricochet. One of the things I learned going through the bubble burst was the importance of executive teams making good decisions and building businesses.
Michael van Rooyen: 5:52
Yes.
Bob Friday: 5:53
So when we started the aerospace adventure, that was my first real startup. Raise money, let’s build a business. The one thing I learned quickly was hey, you’ve got to make sure that you have a good executive team in a startup, because it’s one thing if you’re in a big company and you make a $20, $30 million mistake, they shake that off. You make a $20, $30 million mistake in a startup, you’re probably not surviving that mistake.
Bob Friday: 6:13
Definitely not when you’re working towards trying to create value for your investors. You’re always building towards building a company that you can take public right, and I think that was part of the Cisco acquisition. You’re trying to build a real business and something you can take forward to a public outcome. The other thing I learned during that adventure was you know, I’d been a startup guy. Cisco was my first big company adventure and I had to adapt culturally. You know, when I got to Cisco I had some product manager tell me that they were going to delay something for two quarters or something. Hr basically told me I overreacted because I’d never been in an environment where, you know, two quarters was like half the money you had in the bank normally at a startup. So then I finally learned that big companies that’s okay, they have plenty of, you’re not going to be running out of money anytime soon. That transition not everyone can handle it. You know, big companies have a slightly different culture and business than working in a very high-paced, chaotic startup adventure.
Bob Friday: 7:09
I would say the other thing I learned during the aerospace adventure was the importance of scale. At Aerospace we built a controller architecture. But what I did learn at Cisco we basically had to stop for a couple of months and basically re-architecture that for scale. When you decide you have a big company like Cisco, there is no speed dial on the sales team. Once you put that product into a big sales team like that, it has to be ready to scale. And that’s actually something I took into the MIST adventure was making sure that when you go from small to big you want to make sure the foundations, the architectural foundation, can actually scale to handle a big sales team.
Michael van Rooyen: 7:42
Yeah, that’s a good point, so it would seem counterintuitive. Right, when you get absorbed into a larger company, you know, take a pause, but make sure it’s designed right, because once it’s lit in the wild, it’s got to be ready for that. It’s going to be used everywhere by many different versions, so that is a really interesting point.
Bob Friday: 8:02
Off the back of that, actually, as the technology decisions significantly impacted customers’ outcomes, Actually, one of the inspirations for the MIST adventure really started with some big customer discussions at Cisco. One of my more famous stories is around talking to some big retail customers. I was trying to sell them this connected mobile experience. Hey, we’re going to put an app onto your customer’s phone when they walk into the store. I can remember the IT CIO leader basically telling me Bob, we’re not putting anything on this network, any business critical thing on this network, until you promise me the controllers are going to stop crashing. They wanted to make sure that they were going to be able to deliver code faster than once a year. You know, because they were kind of that regression test mode, build code regression test. You know it could take us six to 12 months to actually get a code release out.
Bob Friday: 8:47
And more importantly back then was the paradigm shift from managing just network elements. Right, you know, yes, we have to keep access points, switches and routers green. But it was more important to actually make sure that whatever the application was running on top of it was going to be good. Users are going to have a good mobile experience. Robots were going to stay connected. That was probably, as CTO at Cisco, the big paradigm shift when I realized, hey guys, there was a paradigm shift from Wi-Fi going from being a nice-to-have to a must-have.
Michael van Rooyen: 9:14
Yes.
Bob Friday: 9:15
Because I remember in the early days back when we started, wi-fi was kind of a nice-to-have in the conference rooms and it really started to become a must-have when Intel put Citrino and Wi-Fi into laptops right. Once Intel did that, that was a transition for the Wi-Fi industry and that started that whole nice-to-have-to-must-have transition Happy Wi-Fi.
Michael van Rooyen: 9:33
happy Wi-Fi, as I keep saying, is that’s what people want right.
Bob Friday: 9:37
Yeah, there’s a generation of kids who’ve never seen an Ethernet cable. I’m sure that’s right. What is?
Michael van Rooyen: 9:43
this and then you know you really drove the Cisco wireless strategy from mobility, went through lots of life cycles, lots of new product development, lots of new features, functions, and continued to drive that market. Then came along Meraki and I know that you were involved or had some input into the Meraki acquisition. Can you tell me a little bit about how pivotal that was in wireless and the fundamentals of that, as well as what led to Meraki being acquired by Cisco, knowing they’d already had a wireless portfolio?
Bob Friday: 10:11
I met the Meraki team when they first were getting started. Back then Cisco was doing very good at large enterprises, medium enterprises, but we were not doing good at small business. And when you look at the Meraki team, what they really brought to the market was that first cloud managed networking, wi-fi networking experience. And that’s when it became obvious that hey, managing things from the cloud was going to be a much easier way for dealing with, especially in a small environment where you didn’t have IT teams right. And that was kind of the original theme and initiative for Meraki was around that cloud managed simplification where you didn’t need all these large enterprise features, and Meraki did well.
Bob Friday: 10:46
I mean, meraki was another one of Cisco’s success stories of acquisitions and that was probably the other reasoning for Mist was hey guys, cloud is going to be a big part of managing networks going forward, moving software to the cloud. The cloud is just a much better way to develop and maintain software. You talk a little bit about Mist. You’re going to see extending that over to the cloud. That is a big thing. That transition to cloud is a better way of maintaining and developing software A hundred percent and off the back of it again.
Michael van Rooyen: 11:12
I think about you know you being acquired at Airspace, being very involved with Cisco helping acquire Meraki, or at least you know having input into that when you’re evaluating a build or buy decision, because you would have had to consider these sometimes. What are some of the factors as a tech leader you should consider a way in on the interesting thing.
Bob Friday: 11:29
A couple of comments on build versus buy. So I was at Cisco. We actually had a small meeting business team there. I think it’s hard when you’re trying to do build versus buy and one question is well, why couldn’t Cisco just build Meraki? It’s like it was there, I would tell you. Culturally it was very hard for hey guys. Can we just basically build what they’re doing? Can we just copy? The question of build-buy is more of a question of do you really have the right team to build it or do you need to buy it? But if it’s strategic to the business, you need to get it some way.
Michael van Rooyen: 12:00
Otherwise you as well. Back at that. Then, bob, tell us a bit about the inception of MIST Systems as it started out to be, and what made it stand out to you and obviously it was an idea that you’d really drove and then maybe a little bit about your partnership with Sujay Hajela and how that contributed to the success of you both bringing this together.
Bob Friday: 12:15
There’s a couple of different themes going on here. First theme is the inspiration for MIST came from that big retail discussion which kind of when it became obvious, hey guys, there’s a paradigm shift going on for managing the network elements to actually trying to manage this end-to-end cloud experience. That was one inspiration for Mist. I think the other inspiration for Mist for back then was I don’t know if you remember the Watson IBM Jeopardy, because I had actually during my master’s I had done and built a neural network for seeing if I could build a demodulator for a radio. But then I was like, hey, if they can build something that can beat the champion Jeopardy’s, we should be able to build something that can answer questions about networking right.
Michael van Rooyen: 12:52
It’s like this stuff.
Bob Friday: 12:53
Okay, guys, this technology is mature enough now. So I would say that was the inspiration for hey, there’s an opportunity here in the marketplace. The big bet that Sujay and I made when we left Cisco was it was an architectural change. If you look at airspace, that was really an emerging market opportunity and that’s a market timing thing. And you look at Mist, this is a mature market Cisco, arubo and you’re not going to beat these big, mature companies with a new feature. This is a bet on architectural and that was the bet on cloud and AI that architecturally, this is going to require a blank sheet of paper.
Bob Friday: 13:27
From my experience at Metricon, ricochet and watching an executive team make some bad decisions and tying up $300 million in inventory, that’s when I learned you guys, you have to have a very good executive team to make decisions. And when you go into these adventures like Mist or Airspace or whatever, they’re what I call the ultimate team sport and it starts with a strong CEO, because ultimately, the CEO is going to make a couple of critical decisions that make or break that startup. So that was kind of, you know, making sure that you have the right CEO, and Sujay and I started the adventure and then you have to make sure you bring on product sales marketing. You want to make sure you have a strong team as you go along in the adventure.
Michael van Rooyen: 14:04
Absolutely, and just thinking about the inception, was there a particular moment where it just hit you? Was it one of those middle of the night we should do this, or was it kind of building towards?
Bob Friday: 14:11
building this. I mean, I think it was building towards it. After the aerospace adventure I was at Cisco watching the industry kind of mature. I would say in general, if you talk to VCs, whether it’s CRM, networking, usually there’s an architectural change every 10 years. That disrupts industries. And that’s where I kind of saw that after doing the Meraki acquisition, it’s like okay, it’s clear that cloud is going to be a better way of doing this. Now, if you look back in the history of Meraki, you will see that they actually started before AWS and Azure and Google, so they actually had to almost build their own cloud right, they were racking, stacking things to actually get the servers up and running, and so that was the other transition right.
Bob Friday: 14:53
You know, we were seeing these public clouds where compute and storage was starting to get cheaper and easier. There was also an open source trend right, the open source ecosystem was getting bigger and that’s where you saw a lot of startups leveraging all the open source code. Right, because you didn’t have to build everything from scratch. So all those things kind of came together in the 24 timeframe when MIST started. If you actually look back in hindsight, go back to Google Trends, you’ll find also that was kind of the transition when machine learning and AI Corey started to take off. Those same trends were building into AI right Compute storage, open source, tensorflow. All of a sudden, these open source packages came out that made it easy to actually develop an AI solution.
Michael van Rooyen: 15:30
And I know today that there’s obviously since ChatGPT landed, you know, in 2022, it’s on everyone’s front of mind of it. But I think the important thing for the listeners here and I know you’ve touched on it, but is that you know, mist was really created to solve cloud management, re-architected from scratch for scalability I think that’s the reference you’re making. But also it was built with AI in mind from the start, and that was, I think, 2014, if I’m correct, or before that.
Bob Friday: 15:52
Yeah, no, the MIST adventure started in 2014.
Michael van Rooyen: 15:54
2014.
Bob Friday: 15:55
And I would say when we started the adventure, it was really focused on cloud and probably the differentiation from Meraki was really around real-time. So there was a cloud-managed component and making it faster innovation development, you know, because you’re able to do weekly production pushes versus yearly production pushes. But, more importantly, it was about real-time hey, we’re going to build an architecture that allows us to ingest data real-time and do something with it. You know, and that was basically the foundation for moving on to what I call fancy math. Usually, when I tell people about AI, it’s either you know we got simple math and we got fancy math.
Bob Friday: 16:27
Usually, when I tell people about AI, it’s either we got simple math and we got fancy math, and ChatGPT falls into the fancy math category. This is like very these are very large models 175 billion parameter things, weights models and we’re training very large models to do very interesting things. So I stay away from the term AI. This is fancy math and, yes, it does. When you do chat GPT, is that AI? I don’t know if I would call it AI, but it is doing some very interesting things now that look like a human cognitive reasoning. Right, you have a conversation with this thing. That’s very scary. Now.
Michael van Rooyen: 17:02
It is, it is and on that, disruptive technologies, like for your organization, you know it’s really around AI native and particularly AI networking. How do you foster the innovation, continue the innovation whilst maintaining stability?
Bob Friday: 17:15
This is what I call the typical Horizon 1 roadmap, horizon 2 roadmap. Horizon 1 is really around making sure that you can deliver the features that your customers want with stability next quarter right, and you have to do that very well. Make sure that happens. Horizon 2 is really trying to understand, typically, what customers aren’t going to tell you what they need. I mean, customers are very good at telling you what is broken right now and what they need fixed next quarter. They’re usually not totally insightful to tell you exactly what’s going to happen two years from now and that’s what I call the balancing act between Horizon 1 innovation, horizon two innovation.
Michael van Rooyen: 17:54
Where are we headed in two or three years? That’s a good point. As the chief AI officer for Juniper, what role do you see AI playing further in the networking industry? Where do you see it going further than that? The customer pain points that you’re thinking in horizon two that’s to come, or where do you think it’s going.
Bob Friday: 18:04
I mean, this is where I say you know, when you look at AI, I always describe it as the next step in the evolution of automation. The interesting thing with this step, and why it’s so disruptive, is the automation we’ve done in the past has been around very deterministic scripts, right.
Bob Friday: 18:18
You know, I build a robot that can build a car, I automate that. That robot does the same thing day in, day out. I write a script for my network that does something, but it does the same thing day in and day out. We’re starting to build now with these solutions are much more closer to cognitive reasoning, right? You know? You look what’s happening with ChatGPT. You start to feel like there’s a cognitive reasoning going on inside that model and this is what I’m calling continuous learning. If you look what we’re doing with like Zoom and Teams, now you know where we’re taking end-to-end data. With networking data, we’re training large models that can accurately predict user experience. That’s the transition or disruption that’s happening with these very large deep learning models.
Bob Friday: 19:01
And this is where I break things down into what I call simple math and fancy math. Simple math is like logistic regression, linear regression. These are things that have been around for a while and we’ve used them for decades to actually try to automate and learn things. What’s really changing and disrupting, you know networking in other industries is really around these deep learning models. These are big, large models that we’re training with tons of data to basically do something on par with the human right. And that Zoom team model is probably the example I’ve given in the past of you know, where a customer had a problem with some Zoom performance issues at a site, couldn’t quite figure out what was causing it. Once you got this fancy math in place, we were able to narrow it down to a misconfigured VPN gateway. This is that issue. Hey, not every user has the problem. It doesn’t happen all the time. These are the types of issue that AI fancy math is going to really help get to the bottom of these things.
Michael van Rooyen: 19:53
Helping customers resolve issues quicker. Better user experience. One for you is off the back of that. You’re working through your fancy math, simple math, and there’s a lot of thinking there. Bob, knowing that you’re a wine connoisseur, tasting wine often involves appreciating the complexities of that and the nuances of that wine. I guess it’s much like understanding sophisticated ii. Does your approach to tasting wine help you tackle complex problems in the technology sector?
Bob Friday: 20:14
it’s kind of a stretch, but there is an analogy between wine and wireless. Yes, there is. My wine adventure started when I started hanging around with marketing people, all right, and they started talking about this wine and French oak, American oak, Hungarian oak Gabby Toast.
Michael van Rooyen: 20:30
And I was like come on, guys, it’s just wine.
Bob Friday: 20:33
This is wine. How much difference could this stuff make? So, yes, I started making a barrel of wine a year to see, to find out. Yes, the engineer in me A, b, testing, control this thing. So every year I would do an experiment where I take half the wine, I’ll put it in a French oak or an American oak barrel and I will tell you that, yes, all these little variables do make a difference. The analogy to wireless and AI is organic. Chemistry is actually a harder problem than wireless. It’s much easier to make a good wireless network consistently than making a good wine consistently.
Michael van Rooyen: 21:07
The process is similar, right.
Bob Friday: 21:08
A lot of variables. Of course, I think the analogy here is wired networks are like everyone’s taking algebra. When you get to the wireless, not everyone’s taking stochastic variables. Wireless is a very probabilistic environment. You know the standard joke right? If someone has a wireless problem, just wait a while and see if it goes away. If you have a switch problem if someone calls up and tells you the port’s broken, it’s not going away.
Michael van Rooyen: 21:31
It’s going to be broken. Great tie-in there around the wine. It’s also the human variable as well, people’s perception of wine as well. Like wireless right. Is it actually really broken, Is it?
Bob Friday: 21:40
not. I think we had a little test right. We did Australian wine versus Napa wines Some of yours, yeah, yeah, yes, yes, and I think the Australian wine you know. Now I don’t know if there’s a little bias going on there, but maybe.
Michael van Rooyen: 21:51
Well, you had three Australians in the room, so potentially Just touching on a couple of other points before we wrap up. Bob, you know, with increasing number of connected devices, continuous explosion of devices, securing these networks is becoming more critical. What is your thinking around leading and using AI to really address these concerns, around security, and particularly when we think about explosion of IoT devices?
Bob Friday: 22:12
So I mean, I think you know if you’re in the security space you’ll hear about EDR, endpoint resolution security.
Bob Friday: 22:18
You know you hear about something called XDR. Now, you know, when I look at the XDR framework it looks very similar to what we’re doing with Marvis right now, yes, where I’m taking a lot of data from clients, network applications to predict a user experience. You can see the security guys will be headed towards the same sort of model, the same data that I’m using to predict user experience. It’s the same data most security guys are trying to use for predicting risk. So I think when we look at security, we’re going to be going from kind of a model similar to where we take all the data and stick it into some big data lake or warehouse and try to make sense of it. I think we’re going to get much more selective on the data coming back, bringing the data back that we need to make you know same user experiences or risk score. I think we’re going to see that transition going forward, that there’s going to be a convergence of network management, security and bringing these two worlds together, absolutely, even our organization.
Michael van Rooyen: 23:09
We’re thinking a lot about that, even from a if I ever take the professional services lens bringing the network and security guys together. The likes of SASE, which is obviously off the back of SD-WAN and Agartana Stand. You know how we start bringing those, not only the consulting capabilities together the security network used to be so separate, now they’re so blended capabilities together. A big security network used to be so separate, now they’re so blended. And then how do we also bring that to the operations center? You used to have a knock on a sock. I think you’re going to see those really blend together as one outfit solving customer problems, securing them, good experience, etc. Knowing that you’re so close on really leading the AI strategy in Juniper, what do you think the next breakthrough in our applications or digital infrastructure is, in your opinion?
Bob Friday: 23:46
Yeah, like I said before, I think what we’re going to see is more of this deep learning in networking and other industries. I definitely you know what we saw with ChatGPT and these large language models.
Bob Friday: 23:56
I think we’re going to start seeing that applied into the networking space and I can tell you people are working on that right now. I tell people from a conversational interface, from a user interface perspective, I think we’re going to see natural language become the next generation of user interface In networking. We’ve kind of started all our careers off CLI. We moved to dashboards. I think natural language interfaces are going to become the standard way for you to actually interact with your network and get data out of it.
Bob Friday: 24:21
I think the other part we’re going to start to see is that Zoom, teams, model, continuous learning, where we’re going to start having more data continuously coming in and models are going to be constantly updated. On learning your network right. You know you think about these models that predict Zoom and Teams and application experiences. Those models will be totally updated nightly you know, or hourly.
Bob Friday: 24:43
So as your network changes, the model will start to pick up on that. If there’s something happening in the network that starts to change the predictions, you’ll start to get that type of visibility.
Michael van Rooyen: 24:51
So really moving into predictive network engineering, management, troubleshooting, really moving from that reactive to proactive, to predictive type services.
Bob Friday: 24:59
I’m actually working with a professor at a university. Right now we’re looking at if I can build models that can actually predict some user experience. On top of that, you can actually start building reinforcement learning models. You can almost treat the network like a game, right? I basically want to learn how to play the networking game and I want to optimize user experience right and understand if I adjust channel power, these features. How does this feature adjust this experience right? Yes, and that’s going to be the power of hey. Once you get these large models, what can you do with them?
Michael van Rooyen: 25:29
You touched on a good point about the interaction and one of my earlier episodes. I was talking to Kevin Block, who was an associate of yours at Cisco, who we’re talking about a missed and what you created there, and he drew the analogy that really English is going to be the new programming language or is currently the programming language. You know people have to learn older skill sets in particular programming languages, but you know english is going to become so important about that. You know prompt engineering.
Bob Friday: 25:50
I’m sure that that’s your thoughts as well, yeah, I mean I think we look at these big models, texas, german, I mean, what people are thinking? Guys hear people talk about hey, do I really need software engineers if I can have models actually write code for me?
Bob Friday: 26:00
yes I think people are thinking the same thing around business intelligence Do I need to learn SQL? Can I basically just tell the database what I want to know, right? So I think, yes, we’re going to be seeing that happening more and more, where the average person is going to be enabled. You don’t have to become a software engineer to write software per se. You don’t have to become an SQL expert to actually write queries into your database. You can basically just say, hey, I want to know what’s the correlation between how many clients were on this network last week or this time period? You should be able to just talk to your network. I tell people Star Trek, right?
Bob Friday: 26:34
That’s right, this is going to be the next generation of talking computers.
Michael van Rooyen: 26:38
It’s absolutely amazing. And again, I was listening to them talk about ingestion of data and you’ve talked about it a bit already. But where about ingestion of data and you’ve talked about it a bit already but where are we expanding that data to? We’re talking about graphs and all sorts of other data. We’ve got, obviously, a lot of text. We’re doing some video ingesting all these variables. I think that’s your point you keep making, which is we’re just going to continue to grow these models and grow this inference on the amount of data we’ve got to leverage that for answering questions, et cetera.
Bob Friday: 26:58
I think the thing I’ve learned in working with the AES the model itself is probably the simplest piece of the puzzle. Any good college graduate can train a model. It’s more about the training process than the data itself. If you look at ChatGPT, there was some magic around the transformer model that Google created back in 2017. But a lot of the magic in ChatGPT was really around the training. What makes ChatGPT so much different than the large language models? That gets down to the data and the training process.
Michael van Rooyen: 27:27
Yeah, agreed we’re coming to the end of this episode, bob, and I just had a one question for you, which is really an open question, not specifically related to mist or cisco or any of your previous history. Tell me about the biggest technology change or shift that you’ve personally been involved in or are seeing or have seen?
Bob Friday: 27:43
that’s the problem with living so long now you know, if I look at the 40-year adventure yes, you know from where I started right, yes, you have to remember back when I started high school. We’re still using slide rules so there was the calculator.
Michael van Rooyen: 27:57
That was a pretty big change. Wow, wow, that’s a good one.
Bob Friday: 28:00
Yeah, that led to everything yeah, there was a cell phone thing that happened. Yes, that was a pretty big change, that is a massive change yes.
Bob Friday: 28:06
It was at Metricon, ricochet. There was this internet thing that came around. That actually transitioned back. Then we went from being a utility reading meters company to building networks for laptops, right. So that was a major transition. Then we had this Wi-Fi thing pop up in that 1997 time frame. If you look at my career, my whole career was built on unlicensed spectrum, right? Yes, you know whether it was metricon, ricochet, wi-fi, ble locations. So all these technologies are transitioning, I would say probably the only thing that is amazing and then there’s the ai, chat, gpt stuff that’s come around is the speed at which things are changing.
Michael van Rooyen: 28:45
That’s a good point.
Bob Friday: 28:46
Right. I mean I think over my 40 years is keeping up with the transition. That’s right. I’m a software engineer, a wireless engineer, a system engineer, and my standard joke now is, like you know, 40 years ago, if you bet a dollar that in my career as chief AI officer, you guys would be all billionaires right, oh, totally yeah. It’s like you know who would ever thought that 40 years later I’d be chief AI officer.
Michael van Rooyen: 29:08
That’s a good point, Bob. It’s a great history. It’s amazing what you’ve seen and developed for not only the industry and users. I really appreciate the time again today. Fantastic to see you and thanks again.
Bob Friday: 29:18
Yeah, thanks for having me. It’s always fun. Hawaiian and wireless go better together every day. That’s exactly right. Thanks, Bob.
