Cisco is well known for its data, networking, security, and collaboration products. On today’s episode of the Me, Myself, and AI podcast, Cisco’s president and chief product officer, Jeetu Patel, joins host Sam Ransbotham for a discussion about artificial intelligence, a “megatrend” Jeetu sees as perhaps more significant than the development of the internet or the automobile because of its ability to build on past technological advances.
Jeetu and Sam discuss how to manage AI and how to staff for it — Jeetu argues that replacing less experienced or younger workers with technology deprives organizations of key perspectives and new ideas, and instead advocates for developing reverse-mentoring programs inside organizations.
Subscribe to Me, Myself, and AI on Apple Podcasts or Spotify.
Transcript
Allison Ryder: We talk a lot about AI being overhyped. Today’s guest believes … it’s not. Continue listening to learn how his point is connected to tech infrastructure and security.
Jeetu Patel: I’m Jeetu Patel from Cisco. And you’re listening to Me, Myself, and AI.
Sam Ransbotham: Welcome to Me, Myself, and AI, a podcast from MIT Sloan Management Review exploring the future of artificial intelligence. I’m Sam Ransbotham, professor of analytics at Boston College. I’ve been researching data, analytics, and AI at MIT SMR since 2014, with research articles, annual industry reports, case studies, and now 12 seasons of podcast episodes. In each episode, corporate leaders, cutting-edge researchers, and AI policy makers join us to break down what separates AI hype from AI success.
Hi, listeners. Thanks for joining us again. Today, I am talking with Jeetu Patel, president and chief product officer at Cisco. Jeetu, thanks for joining us.
Jeetu Patel: Thank you for having me, Sam. It’s great to be here.
Sam Ransbotham: As we talk, Cisco equipment is probably behind 90% of the infrastructure that we’re using, but some listeners may not be aware of all that Cisco does. So let’s start off there. Jeetu, can you give us some background on Cisco and, in particular, your role?
Jeetu Patel: Sure. I’ll start in reverse order. I run products at Cisco. So all the products that you use from Cisco, whether it be networking products, whether it be security products, whether it be data products, or collaboration products, those typically are ones that I’m in charge of building and taking to market, of course with a very, very capable team. Essentially, the way you should think about Cisco is we are the critical infrastructure company for the AI era. So all of the plumbing that’s required to make sure that people can connect, [that] people can stay secure while they’re connected, and that people can make sure that they have the data platform, those are the things that we provide to the market.
Sam Ransbotham: I saw a quote from you saying, “Think of us as the picks and shovels [company] during the AI gold rush era.”
Jeetu Patel: Right.
Sam Ransbotham: For listeners who may not be aware, there’s a famous — or infamous —statement that the people who made money in the gold rush were people who sold picks and shovels to the individual miners. Some of them made it big, but some of them did not. And I think your analogy is quite spot on there.
Jeetu Patel: One of the things that you find when you go through these kinds of massive — what I would call disruptive — platform shifts, where we’ve all been going down a certain path with a certain set of assumptions, and then the assumptions change because a whole new set of technologies emerge, which is what AI is, what you find is the infrastructure that’s required to go out and run those technologies needs to be rethought and reimagined. And anytime there’s any one of these major platform shifts, the infrastructure providers tend to make out pretty well because you have to change the entire plumbing of the apparatus that’s going to be used. So if you think about when automobiles were built, you now need to have roads, and you need to have expressways, and you need to have traffic lights, and you need to have a whole system in place for how automobiles actually get integrated into society.
It’s not just the automobile, but everything around it needs to change. If you think about AI right now, these data centers where these digital workers are going to live, they have to be completely reimagined, because the current data center does not have the power availability and the compute requirements and the network bandwidth to be able to fulfill and satiate the needs of what an AI system would need. So you have to kind of rethink and reimagine and, as they call it in technical terms, rerack the data centers because there [are] racks and racks of computers, and network and switching gear that actually have to be cooled in a certain way, and so on and so forth. That entire shift is what we are in the midst of. And Cisco is a natural benefactor of it because we provide the infrastructure for the AI era.
Sam Ransbotham: [Compared to] some of the analogies you’re making — we talked about the gold rush, we talked about the internet, we talked about cars — is artificial intelligence at that level of big deal, or not?
Jeetu Patel: I think it might be bigger. The way I’ve seen these kinds of shifts happen: Imagine if Amazon.com got built in the 1600s. It would be an epically failed company, because you didn’t have the internet and you didn’t have the underlying infrastructure on top of which Amazon could be built, because you didn’t have the shipping and the logistics infrastructure and the internet and all of those pieces. In a similar vein with AI, we have the benefit of having all of the infrastructure that’s been built out to date. When you have something that’s built taking advantage of all of these things, by definition, each one of these subsequent, major, transformative waves tends to be bigger in impact than the previous ones, just because it was built on top of shoulders of giants of the previous innovations that had happened.
So I would say this is probably the most consequential set of inventions that we will have seen in our lifetime, for sure, and probably arguably in humanity. The thing to keep in mind is the pace of innovation and the slope at which it happens actually makes it impossible to predict what’s going to happen three years from now because we’ve compressed the time. Scientific progress will probably compound by 1,000x. And so you’ve compressed this timescale. … We’re in a warped situation right now where humans can’t make sense of this because everything’s moving so fast.
Sam Ransbotham: Actually, there’s so many things to talk about there. I think your analogy to Amazon and the 1600s is interesting because, for example, we had neural network designs back in the 1980s. We just didn’t have the compute infrastructure, the data, the telecom to pull it off. So it took all these things coming together. And I think what you’re saying is that all these things are together now for us to build from.
Jeetu Patel: Actually, the biggest thing that we have that we didn’t have then, because this would’ve been hard to go invent, is AI has been around for a while, but when did it actually take off? It took off on Nov. 30, 2022. What was so significant about that date? That’s when ChatGPT was launched. What was so significant about ChatGPT? It was essentially what they call a large language model. And a large language model was a model that actually understood human language, rather than the machine having to be rigid and the human having to learn the machine’s language.
It became the other way around. How did that happen? That happened because we had petabytes and petabytes of publicly available data on the internet that you could use to train these models so that these models would then know what to do with it. So if you didn’t have the internet, you would’ve not had AI because you would’ve not had that level of data to then train the models on. It goes back to our point of each one of these inventions or revolutions is built on the infrastructure provided by the previous revolution. And it’s very evident in this case.
Sam Ransbotham: One of the last conversations that I had with my grandfather was how much life had changed from no electricity, limited indoor plumbing, no airplanes, no space travel, no computers, to when he passed — how radically different that was. And I remember talking with him at the time, saying, “Oh, wow, you lived through a bunch. I can’t imagine things changing as much during my lifetime.” I may very quickly be eating those words.
Jeetu Patel: I think we get into this kind of cycle. … Humans are not very good, in general, at imagining the exponential outcomes. We’re very good at imagining linear regression but not the exponentiality of the outcome. So what ends up happening is we think about exponentiality in a single dimension, not multidimensional. … Years ago, I had a chance to sit down and talk to Ray Kurzweil, who’s one of the scientists at Google. This was 20 years ago or something. And I was interviewing him for something, and we were talking about this notion of perpetual extension of life: Can a human live long enough to live forever? He had written this book where his thesis at the time was [that] if you live until you’re 40, we will have the science and technology to allow us to live in perpetuity.
My topic of communication with him was around the social implications of that. What happens if seven generations live simultaneously or 10 generations live simultaneously? That’s going to be really hard because we’re not going to have enough room to put everyone, and we’re not going to have enough crops to go feed everyone. And he’s like, “You know, this is the problem with humans. … We can’t think in exponential terms because we think in a single dimension of exponentiality,” which is, if seven generations or 10 generations lived simultaneously, what would happen with everything else being exactly the same? But the reality is you might have skyscrapers that might be 2,000 stories high, and you might have a crop cycle that takes three days.
And so those are all things that would also simultaneously evolve so that they can accommodate the constraints that get created because of the developments that happen in certain areas. I feel like that’s the same over here, when people say, “You know, I think humans are going to be sitting on a beach, have nothing to do, and AI is going to do everything,” I just chuckle a little bit because I just refuse to believe that humans are designed to be obsolete.
So we will continue to find ways to add value and think creatively. That doesn’t mean that we’ll be doing what we do today. It might very well mean that all the jobs that we do today might not be the jobs that we have, but that also doesn’t mean that we’re not going to have jobs. … The desire for a human to be productive and add value to society doesn’t go away because something got automated. You just create higher-order bits that you’ll then be able to go focus on that you were not able to focus on in the past.
Sam Ransbotham: Exactly. Just to make sure that I can get this thrown back in my face later, I’m predicting massive increases [in] employment, not decreases. No one since the internet is doing less than they were before the internet, despite all the progress possible. I can’t believe that’s not going to happen again. I think we’re headed toward the opposite direction.
Jeetu Patel: I think you’ll see some displacement of jobs temporarily, which we should not take lightly, because I think it’ll cause human suffering. But that does not mean that that will be the state in perpetuity. What you have to keep in mind is that displacement period, if we get ahead of it, because of the pattern that we’re starting to witness, we might be able to actually get society retrained in a more efficient way than we might have done in the past with previous disruptions. And that might be a responsibility that the tech community, the collaboration between the public and private sector, should have as a good, beneficial outcome, because I think there’s going to be more and more of a need for collaboration across different sectors.
This is one of those areas where I tend to be, in the long term, an optimist without being naive about the short-term implications that this might have. Even [the] mid-term implications it has around safety and security, and the downside effects could be profoundly consequential that we have to keep in mind. But I refuse to believe that we’re going to be obsolete or that we’re not going to have value to add. It just seems unnatural.
Sam Ransbotham: I think that’s well put, that we can have a positive aggregate effect but still have lots of heterogeneity in how that average plays out across society, and being naive about that is going to hurt us in the long run. You mentioned security, though. And when we were talking about reracking and infrastructure changes, we quickly slipped toward routers and modems and telecom and hardware-oriented things. But one of the things that I think you’re very focused on in terms of infrastructure is the idea of security, and how does that become a first-class player versus the thing that’s derided as hampering productivity.
So we’ve always had this sort of productivity-security trade-off. I think that you’ve mentioned that we may not be making that trade-off anymore. How can we help security be part of the infrastructure?
Jeetu Patel: Firstly, I think in this particular age, security is going to be a prerequisite for successful adoption of AI because if people don’t trust these systems, they’re not going to use them. That’s very different from in the past, where you would think about security as a necessary productivity impediment. That’s no longer the case. Now it just happens to be a prerequisite for successful adoption of AI.
But I actually feel like it’s probably worth taking a step back and saying, “Where are we still thinking very linearly?” Where I feel AI is underhyped the most is the fact that we still keep thinking that this is just a productivity game. Humans are going to get more productive. Things are going to happen cheaper, faster, better. I actually feel that’s only the first-order effect.
The second-order effect is [that] you will actually start to see these AI models. It’s not even clear if large language models will be the ultimate destination. You’ll have large world models. You’ll have physical models. All of these things will start kind of combining together. But the new paradigm, whatever it ends up being at some point in time, and the existing ones will start to create original insight that did not exist in the human corpus of knowledge. It won’t just be an aggregation mechanism. It won’t just be where you take a multitude of different perspectives. This is not just a better search engine, where humans had some data, and you indexed that data well, and you were able to go out and put it into a clean paragraph. It’s going to start creating original insights that didn’t exist in the human corpus of knowledge.
And when that happens, the thing that changes is [that] you are now able to imagine solving problems that you could never even dream of solving before. And that is far beyond just going out and optimizing for productivity. I feel like that’s the most misunderstood part of AI: “Oh, I’m just going to get more productive.” Productive use is going to be like 10% of the equation. Going out and doing things you couldn’t do before, and solving problems you couldn’t solve before in different ways that you couldn’t even dream of solving before, is probably going to be the 90% factor.
Sam Ransbotham: I like that idea, to observe the productivity effects first, because, as you say, they’re first order, but how do companies, how do people start thinking about what they can do with that 90%? Productivity is pretty tempting. I mean, I like greater productivity. This can be hard for me to turn away from. Or maybe turning away is me sort of making it a Hobson’s choice, where you have to do one or the other. But how do people start thinking about this 90% or these more than productivity options? I don’t think I’ve ever talked to anybody who thought AI was underhyped. And I think we’re on record for saying that here.
Jeetu Patel: I think where it’s overhyped is where the human obsolescence becomes almost a foregone conclusion in some people’s minds. I think that’s where it’s overhyped. I feel like human instinct and human judgment [are] still pretty hard to go out and replicate in the machine’s ability to do things because we don’t make most of our decisions based on data. We make a lot of our decisions based on gut. And that gut is hard to go replicate. Typically, people say, “Listen to your gut,” because there’s a reason for it. There’s an instinct that’s palpable.
But to go back to your question of what should companies be thinking about. By the way, I think I highly encourage the productivity argument. I think everyone should go out and think about productivity. And they should continue to keep kind of powering through that. It’s going to be a great benefit that we will all be recipients of. Where I think the unlock truly comes in is by actually trying to make sure that we challenge the conventional norms of thinking and ask ourselves the questions, “What problems have we been conditioned to think that we can’t solve?” and “Are those unsolvable moving forward just the way that they were in the past?” I feel like you’ll actually start to find very different answers.
I feel like we need more and more questioning … of the status quo. And the way to do that, in my mind, the one single thing is going to be the exponential difference. You have to unlearn as much as you are learning. Unlearning requires that you actually inject new talent into the system at a very rapid pace, and then give creative freedom to that talent so that you are actually getting mentored by them just as much as you’re mentoring them.
This was a recent conversation I’d had at a conference I was at, where people were like, “Entry-level jobs are going to go away,” and “We’re just not going to hire early-career people,” which in my mind seems like the stupidest idea that a company could pursue. If you actually don’t hire new people to come in, you have essentially given up on [the] injection of new talent and new ideas into the thought process. So this kind of baggage of experience will always hold you back. You know a lot of things, and you might not be, as a company, good at unlearning, and you have no one else to actually instigate that unlearning and catalyze that unlearning by asking questions, because they didn’t have the baggage of knowing. I do feel like the mix, the continued infusion of talent [that’s] early in career is going to be so important for companies to be able to get the most out of it, because we have to understand instinctively how to go out and use these tools, which are in service of humans, in a very different way than the way that we might’ve used them in the past.
And right now, frankly, if you take someone who is a 20-year-old and a 28-year-old, and compare the two of them and how they use AI, it’s night-and-day different.
A 28-year-old might actually use it for productivity. They’ll go out and they’ll ask it some questions because they’ve got some answers to get. Then they’ll move on. A 20-year-old thinks of it like a companion and a brainstorming partner. And they might actually talk to it and brainstorm with it so that they can come to an ideation. They’re not looking for answers. They’re looking for substantive volleying back and forth, and brainstorming. I learned that from the interns that would come into Cisco. So I think that’s the aspect that I think we have to keep in mind, that as a company, we have to make sure that we keep challenging and disrupting ourselves before someone else disrupts us.
By the way, innovation is not limited. This is the one thing that people kind of bucket into these completely unproductive ways, which is you’re a small startup, you innovate really fast, you become large, you stop innovating. I think it’s nonsense, because innovation is not like something that’s limited to a certain group of people. Anyone can choose to innovate at any point in time. You just have to have the right mental model and mindset. What you have to do is fight the temptation for bureaucracy being something that you succumb to. So challenge the bureaucracy and allow people to come in [who] challenge the status quo, and … by definition, the byproduct of that is going to be invention.
Sam Ransbotham: The unlearning idea really hits home for me. I think we’re going to have to cut that from the episode because I don’t want my kids to hear it and then think that I am not full of wisdom and that their ideas are important. But it appeals to me, because many of the things that I think that got me to where I am in my career are not necessarily the things that seem like they’re going to keep me going through the next phase of my life. And so that unlearning makes sense, but at the same time, I have trouble knowing what things I should unlearn and what things I shouldn’t unlearn, and I feel like companies have that same problem. They got successful through some strategic core competency. The idea that “Oh, yeah, we unlearned everything” seems too global to me. So how do we decide?
Jeetu Patel: No, I don’t think you have to unlearn everything. I think humans build on top of each other’s learnings. One of the most important inventions that ever was created was a printing press because we were able to communicate the learning from one generation to the other in a way that was very concrete. The combination of language, script, and the printing press, and [the] ultimate level of desire to share your knowledge with others, which is instinctive to us, and [the] ultimate desire to learn from other people’s learnings, which is also instinctive to us, was actually pretty valuable.
So I don’t believe that you should unlearn everything, but I believe that pairing up experience with inexperience is really valuable. And … make sure that there’s a bidirectional mentorship that’s occurring in your ethos of your organization, where the experienced people are coaching the early entrants — by the way, by early entrants, I don’t always mean young people, because I could be inexperienced in a brand-new domain. Sometimes I have to force myself to just go into uncomfortable spots and go into new domains where I can learn [a] new thing, that I can ask questions that might not be conventional wisdom.
What I think is really important is conventional wisdom helps many times. And many times, conventional wisdom prevents us from exploring something that we want to explore, and thereby creates barriers that are unnecessary. And so that’s what we have to kind of undo. That’s the area that I feel like there’s opportunity for organizations thinking differently. Don’t just mentor your interns. Have a reverse mentorship program as well. So if you’re spending an hour with an intern, make sure that one of your key objectives is — for 30 minutes of it — make sure you’re getting something out of it, not just them. It’s not one-directional. It’s bidirectional.
I think I’m practicing what I’m preaching in this conversation that I had with you because I’ve tried to always go into areas that I knew nothing about. And I found that keeps me curious, that keeps me motivated, that keeps me learning. And it also allows me the permission to ask silly questions, which then free me from the burden of experience sometimes that I have to have in certain areas. And then in other areas, [the] number of years in the system teaches you patterns that as long as you don’t think those patterns are non-shiftable, then you actually benefit from them. I always think [that] “strong opinions, loosely held” is a good model, which is completely flippable with new arguments and new data.
The reason I got into computers was because some uncle told me the night before that it seems … better than going into business management. So I took computers as a class. And then I got into that and got interested. And from there, [I got] into consulting. And then I did consulting for a long time. I started my own business. And then the consulting thing got boring after a while because I didn’t feel like I had exponential scale in that business. And I wanted to learn scale because I was fascinated with scale. So I got into software.
From software, one thing led to the other, and I got into products. And products were kind of fun to use because you started doing things in the cloud. And then the cloud became fun because … now you’re starting to benefit from AI. The formula that I’ve used is “don’t ever fight a megatrend, and use it as a tailwind.”
And know the difference between a megatrend and a hype cycle. So if you can go out and effectively deduce what is a megatrend [and] what is the hype cycle, not fight the megatrend and ignore the hype cycle, you have an advantageous position in society. Now, by the way, the instructive question is, How do you know the difference? In my mind, there’s a simple formula, which is if it requires a Ph.D. for someone to explain what the benefit of something is, it is a hype cycle. If it’s something that is instantly obvious in what the benefit could be, where you could imagine five steps forward, it is likely a megatrend.
Sam Ransbotham: We’ve had a very conversational-oriented [discussion] so far. Let me switch to … rapid-fire questions. Just give me the first thing that comes off your mind.
What’s moving faster about artificial intelligence or slower than you expected?
Jeetu Patel: What’s moving faster is the rate of change. And what’s moving slower is the use cases that organizations are actually starting to find tangible value from. I think the technology is moving fast. The adoption is moving slower.
Sam Ransbotham: What’s been the best use of artificial intelligence so far for you personally?
Jeetu Patel: Research. Getting dexterous in a particular domain in a fraction of the time of what it could be in the past is something that I don’t think I would’ve been able to do [in] my job currently and have taken that job on and been able to get up to speed as fast if it weren’t for AI. I am a direct benefactor of AI. My family would not be fed the way it is today if AI wasn’t around. It’s that simple.
Sam Ransbotham: What do you wish that AI could do better? Or what frustrates you about AI?
Jeetu Patel: I think we are still in a very kind of chat-based interface. Yet I think we are squarely entering into the next phase, which is agents being able to conduct tasks and jobs fully autonomously. I still do a lot of things I hate doing in my day that I think at some point in time AI will take off my plate. I don’t think we’re quite there yet.
Sam Ransbotham: Amen to that. Has using artificial intelligence made you spend more time with technology or less?
Jeetu Patel: More.
Sam Ransbotham: More?
Jeetu Patel: Because I’m just curious. I spend from 9 to midnight every night, almost, in my learning mode, which is something I never really did quite that religiously before AI.
Sam Ransbotham: Well, it’s been fascinating talking with you and learning. … I like the phrase about never fighting a megatrend. And if you’re right about AI being underhyped, which I think you’ve made some cogent arguments for, then we’re in the middle of a real shift. Thanks for taking the time to talk with us today.
Jeetu Patel: Thank you for having me.
Sam Ransbotham: I hope you enjoyed the conversation today. In two weeks, I’ll be joined by Kathleen Peters, chief innovation officer at Experian. Please tune in.
Allison Ryder: Thanks for listening to Me, Myself, and AI. Our show is able to continue, in large part, due to listener support. Your streams and downloads make a big difference. If you have a moment, please consider leaving us an Apple Podcasts review or a rating on Spotify. And share our show with others you think might find it interesting and helpful.