Subscribe
Share
Share
Embed
How do we begin creating the future we want, today? Design This Day takes you on a journey to our future world. Futurist Devin Liddell sits down with visionary leaders from some of the biggest names in tech and innovation. Each episode features a brilliant mind who is building the opportunities of the future before most people even know they exist. What will living in microgravity in space look like in the future? Can driverless vehicles go off-roading? What unexpected roles will robots play in our future workplaces and homes? We explore the role that design plays in shaping our future – with the big thinkers and doers who are creating tomorrow, today.
Design This Day is an original podcast brought to you by Teague.
About the Host: Futurist Devin Liddell
Devin Liddell is the Principal Futurist at Teague. With over two decades of experience in innovation and design strategy, Devin has worked with industry giants like Boeing, Intel, and Nike, helping organizations anticipate changes across both near and far-term horizons to create their preferred futures. Devin is a frequent contributor to Fast Company.
Get in Touch:
Have a complex problem that needs solving? Have a great guest or topic idea? We want to hear from you. Visit us at teague.com or send us an email at hello@teague.com
Ed Doran: A giant data center for a hyperscaler is a little bit like an F1 car. It's a triumph of engineering, but even Max Verstappen isn't driving that to Trader Joe's to get his groceries at the end of the day. I mean, I bet he could. I can barely navigate that parking lot, but I bet he could. So the question is, how do you take all of those accomplishments, all of that exquisite science and engineering and extend it into the real world?
Devin Liddell: Welcome to Design This Day, a podcast about the futures we want and the people working right now to make those futures real. I'm your host, Devin Liddell, a Futurist at Teague.
This episode, we're joined by Ed Doran. Ed started his career as a scientist, but he's spent the last 20 years as an executive at some of the biggest tech companies in the world. He's worked with Yahoo, Google, and he spent 13 years at Microsoft where he co-founded Cortana, Microsoft's AI voice assistant.
Today, he works as the VP of Strategy at the Edge AI Foundation. And through all of these roles, he's found a really unique specialization. He's an expert at the translation layer, helping projects get from the risky, uncertain research phase and into the real world. A lot of great ideas don't make that leap into reality. They get lost in a valley of death.
Ed Doran: You could imagine two people standing with a chasm between them. You wouldn't walk up with this beautiful, perfect product and then just hold it over the cliff and let go. It would never cross. Even if people on the other side wanted it. Let me walk you through the canonical life cycle of new inventions that make it to scale. Number one, that's a bad idea. You will notice something. You will figure something out. You will take this beautiful idea of yours to someone and they will look at you and go, "That is a dumb idea." They will. "No one's ever going to pay $600 for a phone. No one's ever going to get in their stranger's car. No one's ever going to stay at a stranger's house. Not going to happen." That's a bad idea. So then you go and prove, actually, this is what people really want.
This is the need they have. Here are the clunky ways they're trying to solve that need that don't work. So then phase two, okay, it's a good idea, but you could never build anything that works. So then you build out prototypes. And now with vibe coding and with rapid iteration with open source hardware and some open source capabilities, you can actually build out prototypes quite quickly from the hardware, software perspective. So now it goes to, okay, it works, but you can't build that to scale.
At this point, you've built that bridge enough that you're starting to reach the other side. So you have to have a scale compatriot who says, "Okay, I get it. I would build that differently if I had to serve a hundred million, 500 million, a billion people, but this is what that could look like." And then you'll get, you're on the other side and they'll say, "That's fine, but no one's ever going to pay for this." So then you have to be friends with the business folks and the general managers, and the BD folks and say, "Actually, this is what people would pay. This is what people are paying for now and not getting this." And so you go from one side of the cliff, "That's a bad idea," to, "That can't be built," to, "That can't go to scale," to, "That can't make money," to the last step, which is you've built that bridge and crossed the chasm and they're ready to go. And they look at you and they say, "Huh, that's cool. That's my idea now."
And they run with it. Now at first, as a creative or an exploratory person, your heart weeps and goes, "Aha, my baby is gone." But that's what you want. You want them to adopt that idea. You want them to pick it up and take ownership and feel responsible for it. You didn't just walk up to the chasm and toss it over the edge and hope it landed. So you have to go through that life cycle, which many creatives will recognize as a lot of nos, but you have to be willing to get through the nos to get there.
And then something really interesting happens, is once it's at scale and you're serving a billion people, you start to hit edge cases. You start to hit things that are weird. And I personally would argue, and Devin, you could push back with me on this. I don't think you have a successful product until people use it in unexpected ways. And when that happens, guess what? It just came back to you because now the question is, should you be designing the next generation of that product or should you be just designing a completely different product? And then the cycle starts again. It's hard, but it's fun and it's worth it.
Devin Liddell: Saving a product from falling into that chasm means you have to bridge two worlds. On one side, there are the scientists, technologists, or engineers who are designing these new products. And then on the other side, there's the business team, the product managers or senior leaders who will take it to the real world. So Ed has to understand both teams to be able to bridge that gap.
Ed Doran: If you think about that research side, it's about patience, it's exploration, it's inclusive thinking, it's design thinking. It's really a wonderful place to try to figure out what's coming next, understand the data, but see through the data to where it's going. On the other side, the productization side, and I've been on that side too at Google and Microsoft, it's really about, okay, now that I have this object, how do I scale it up to maximize the value I can deliver to the world?
And so you think about things like efficiency, stability, resistance to challenge, security, policy compliance, et cetera. And so you end up with very different skill sets, very different even personalities, very different processes. And if you try to make one play the other's game, it seldom goes well. So it's a little bit like hiring artists to be accountants and vice versa.
There's the rare few who can do both and we love them, but they're very, very rare. And so it helps to have someone who can carry things across both worlds. They're both very passionate about delivering value. It's just a question of how. Creative side is what's coming next? Where's the open space? What can we rethink? And the scale side is for what we're targeting, how do we deliver it really well, reliably, in a trustworthy way?
Yes, they're distinct, but I think they're complimentary because honestly, the best innovations happen in the middle and you don't see them coming. So at one point I was in grad school and a good friend of mine said, "Okay, I'm doing this brand new hardware thing and we're doing some new statistics on top of it. Can you take a little time and help me out and get this done and I'll pay you?" And I was like, "Okay, great.
When is it?" "Well, we need July." "No problem." So I sign up and then he's like," Oh, by the way, I forgot to mention, it's in the middle of the ocean and it's on this ship." I'm like, "Well, that's a pretty important piece." So I spent July in the Northeast Pacific on the NOAA ship, MacArthur. By the way, beautiful ship, beautiful, elegant, rides like a cow. So it just bounces everywhere 24 hours a day.
At 6:00 PM, we would go work on hardware and statistics, it didn't finish at 6:00 AM every single day. But at 6:30 AM, you would find myself, my collaborator, and one of the deckhands sitting inside the walk-in freezer with one three-gallon tub of Neapolitan ice cream and three spoons just eating ice cream and talking about what we saw. And the ops guy was talking about the operations. The hardware guy was talking about hardware. I was talking about stats. And that's where the really interesting thing happens. It's three spoons and a giant tub of Neapolitan ice cream. I would encourage you to find those moments because that's when really, really non-linear, really exciting ideas come from.
Devin Liddell: The Neapolitan tub of ice cream portrait of activity is really spectacular because what you're describing is not actually a multidisciplinary approach. What you're describing is an interdisciplinary approach. All the folks gathered around that ice cream are having ideas that are actually technically outside their discipline, but because of your collaboration, everyone's contributing in a way that sort of supersedes those disciplines.
Ed Doran: Oh, 100%. And as an organizational leader, your job is to create those little nodes, places where the scale folks and the creative folks come together. Those places where people can freely share their ideas. And yeah, it's a tub of ice cream. 100%.
Devin Liddell: I love that. I've done a lot of ethnography across my career as well. And sometimes I think those moments actually end up being the things you take away the most compared to actually the product itself.
Ed Doran: Yeah, I think you're totally right. And we've done really interesting work with Teague in the past and with other partners because I think you need that empathetic view of things. Because you need to understand the data, you've got to do the quant work and the stats. I know that more than anybody, but you need to figure out where it's real. So at one point we were trying to wrap our heads around the next opportunity for taking compute with you into the real world. And I ended up in the studio apartment of this Brazilian woman with my translator. And she was sitting on her bed and she said, "I don't really like my phone. I don't use it that much. I'm really a people person. That's my life." And I said, "That sounds interesting." And I scanned over and next to her bed was an elaborate station to charge her phone, to position her phone, to use her phone while she was lying in bed.
It was the first thing she touched in the morning, last thing she touched in the day. I looked at it. There wasn't a speck of dust on that charge station. Now, if you don't look at that empathetically, you say, "Okay, well, she didn't tell me the truth. I'm going to move on. That's it." If you look at that with a design mind, with an empathetic mind, you say, "What she's really telling me is what she wants. She wants not to be tethered. She wants not to be interrupted. She wants not to have to build a whole set of infrastructure to live her life."
And so you look at what she's saying and saying, what she's really telling me is, "I would like to have this support, this companionship, this enhancement to my life without all of these trappings of battery life and cords I have to plug in and my phone that could get stolen."
And if you look at it that way, you see a lot of the conversations you're hearing even today around wearables, glasses around pods, around pods of cameras, that's the direction they're going in. And that's that shift from, don't make me interrupt my life to stop and look down at my hand to work on my phone, integrate into my life and enhance it in new ways.
And so I think you're 100% right about those ethnographies, about thinking more deeply about this. And I think the empathy is the difference between missing what she was saying and thinking she was just not telling you the truth, and understanding what she was telling you, which is what she wanted in the future. But I still think companies are trying to figure that out right now.
Devin Liddell: Yeah, absolutely. Yeah. We used to use, sort of a three-word, a three-pronged kind of approach to build around. And so like honest, disruptive and loved. Honest in the sense of really born of empathy. Actually born of insights you cannot get without observing people in the real world doing what they do.
And then disruptive in the sense of actually creating something transformative that changes the way people do things. And then loved in the sense of we're actually not after loyalty. We shouldn't be after loyalty. We should actually be after real affection for, "Hey, I love this thing because it helps me be my best self by using it."
Ed Doran: Completely right. Back and forth between the scale side and the exploration side. Honesty is bringing things back to the scale side and saying that thing we thought, it's not really what they want. It's not really what they're hoping for. When I worked on mobile phones and this digital assistants, like Cortana, we agonized over GPS. I built these long, complicated scenarios of, okay, I'm going to drive home from work and I need a gallon of milk. So I got to know my route home from work and I got to know which stores I want to go to. And am I going to turn right off the road, not left so I don't put my life in danger? Blah, blah, blah, blah, blah, blah, blah. I swear to you, not a single time in all of the discussions we ever had around GPS, did anybody raise their hand and say, "Dating. People are going to want GPS for dating on their phones," that's what they care about. And that is 100% a key scenario for a mobile phone. So to your point, don't just be honest about what you learned, what you saw, but be honest back into the organization.
Devin Liddell: Cortana was Microsoft's AI-powered assistant before Copilot. Ed co-founded Cortana back in 2014. They took the name from an AI character in Microsoft's Halo video game. So in a way, this is another moment where science fiction became science fact.
Ed Doran: So when I came into that opportunity, I was brought to Microsoft to help build new products and new things for the future. I started off by a rather controversial product, which was yes, x86 is great, which is what Windows was founded on, which Microsoft was founded on. I founded the Windows on Arm product. I think the future is going to be on these Arm devices. I think these Arm devices are really interesting and they're going to change our interaction.
At the time that was heresy because laptops are wonderful and desktops are the best. And you're looking at the capabilities of Arm and you realize that actually mobile is going to be more interesting. So the question was really, if I was going to get advice on a mobile device, what'd I want that to be? And very quickly you come to the conclusion that you don't want it to be purely text.
You have a life to lead. You have things to do. You have a car to drive, you have people to interact with. You would want a voice interface. And so from there we jumped off into, well, what kind of things would be helpful to you to allow you to live your life without interrupting your life?
And so that brought us to these questions of how do you build trust? How do you decide what it's going to do for you? How do you decide to change its actions when you disagree, et cetera? And there was a whole bunch of research we did there. We did a really interesting study with the assistants of movie studio heads. So we thought, "Okay, let's find an assistant that's going to have kind of a Miranda moment here." Those of you who've seen The Devil Wears Prada. And it was fascinating.
And what we found when we talked to them is, "Listen, if you really want to do a good job, you need to know what they like, what they don't like, and what they say they don't like, but they secretly like." By the way, a ton of vegan movie studio heads love In-N-Out Burger. We won't name names, but there you go. And if you think about that today, it still applies if you're talking about Copilot or talking about something else. You make deliberate decisions to understand the person that you're helping, but you also have boundaries where you're actually getting implicit understanding of what's important to them.
And so we worked really hard to understand, and I think we're still working really hard to understand what's important to you so we know what emails to service, recommendations to get, what you don't like, or even the implicit things that you don't say out loud, but they still strongly influence your opinions in your life.
And there's even some scenarios that I think I'd love to see today that I've yet to see today. So for example, we did a bunch of research and we actually came up with one called Vegas mode. When you invoke Vegas mode, that means you have to forget every single thing I did and have to do in the last 48 hours.
Devin Liddell: Oh wow. Okay.
Ed Doran: You're shopping for an engagement ring and you don't want those ads coming back up and your significant other hearing about it when you're interacting with other devices. So let's keep that between us. But it does sort of touch on the point that we're seeing quite a bit right now around trust. What do you trust that's between you and I? What do you trust that's between you and I and other people? And what do you trust that's for no one? And I think right now we're at a point where all digital assistants are thinking their way through that. How do I build trust? How do I bound trust? And if I have to, how do I re-earn trust? That's a really interesting domain.
Devin Liddell: So Vegas mode was a mode within Cortana. It did not ship, but you're saying you would ship it today.
Ed Doran: Oh, 100%. I mean, think about when you start to build trust with an assistant, right? There's some things that are just like, okay, this is probably not going to buy us things going forward. So let's imagine your friend invites you to go to the Ren Fair. You've never been to the Ren Fair. Why not? You're going to go. Fantastic. Let's go to the Ren Fair. You do it and you decide, "Nah, it's not really for me." Sorry to our Ren Fair people out there. I can feel the hands reaching out to keyboards as we speak. I'm sure they're lovely. But you're like, "I don't want to get a bunch of suggestions in the future for the next Ren Fair. I don't need a flag in for my need in my house. I'm good." So I do think some form of Vegas mode would be valuable no matter what assistant you're working with, because you change because life evolves.
And if we build these digital assistants with very static views of, oh, you once watched a Chargers game, you must be a Chargers fan. We're going to essentially seem less and less relevant and erode the trust that we have. So yes, I think I would definitely ship Vegas mode again.
Devin Liddell: Cortana was quite groundbreaking in a lot of ways, not the least of which was that Cortana had a very specific tone of voice. And I'm curious if you have any thoughts on both how you arrived at sort of the appropriateness of that tone of voice, like a really energetic and confident voice, and kind of like sardonic in some ways. Is there any backstory about how you arrived at the appropriateness of that tone of voice?
Ed Doran: It's a great question. We did a ton of research on what kind of assistance people wanted. Everything from what did pilots hear when they were in the cabin and they needed good advice? What do you expect when you have an assistant at work? How do you interact with your good friends? I think at the time we tried to choose something that would serve a number of those scenarios. If your assistant is the connector across expert systems, then you're probably going to build a ton of rapport and that rapport is going to adapt to the person that you are.
Some people need reassurance. They want a calming voice. Some people want banter to brighten their day. They're going to want a little sarcasm and a little bit of fun. And I would even argue that that depends on where you are in the life cycle. So the first time you ever meet someone, you're at a party or whatever, and they walk up to you, they're not going to talk to you like they're your best friend.
They need to earn that right. So they're probably going to talk to you in a polite, socially acceptable, expected way. But over time, they're going to start to have a deeper understanding of you and what is funny and what's not funny and what's acceptable and what's not acceptable. And that combination of tone and trust go hand in hand. And that's a lot of the conversation that's happening today. And I think done well, it's fantastic. Done poorly, it's a real problem.
So then the next conversation is how do you recover from that? If you've ever walked up to someone and told them a joke, not realizing a moment before they got bad news, you recover. You pause, you put your hand on their shoulder and you say, "Oh, sorry, I think I misread something here. Let's talk."
That level of interaction is huge. And I think we need more designers asking the question of, "Okay, how do I interact with this system?" And that really draws from human interactions, human expectations, and that journey of trust that you're on.
Devin Liddell: One thing that's sort of remarkable is that a lot of what you just described would be described as soft skills, that ability to maneuver complex social nuance. You're describing the future of assistants and other sort of edge AI use cases where they will actually possess those "soft skills."
Ed Doran: I think they have to possess those skills and I would consider them critical skills. And that's where we need designers, we need sociologists, we need psychologists to help us do that. Otherwise, I don't think you can maintain the trust. So one of the big challenges I see right now in AI is that it gets talked about in big abstractions. So AI is going to be wonderful. It'll read all your emails for you. It'll write all your responses for you. It'll do whatever. But the world doesn't work in abstractions. I think the way to think about it is, think of AI as your new coworker. How many new coworkers have you had?
And think about them. Some new coworkers, you met that person and you said, "Oh, a breath of fresh air. I cannot wait to have this person work with me." Sometimes you meet with a coworker and you're like, "I don't know how they didn't suffocate themselves with their own pants on the way out of the house today." You judge how much you trust them.
And even for the ones that are great at one thing, you wouldn't use them for everything. Your car mechanic, you're not going to ask them to do brain surgery for you. And you take that lens, it actually makes it really helpful to think about AI because recently there were a couple AI products that were launched and they were designed to be sandboxed, a small thing to do whatever you want them to do and figure out how to build trust with them.
And some people just gave them everything. That's the equivalent of saying, "Hey, Devin, here's your new coworker. By the way, we gave them your bank account and your car keys and your house keys, and all your email, and all your text message threads and everything else. Let's see what happens." Now when I queue it up like that, you instantly say to yourself, "There's no way in this earth I would want that to happen."
But people did that with AI systems because they didn't just think of it as a coworker. So it's an interesting lens because I think it builds on your comments around psychology and sociology. How do I meet this person? How do I assess what they're good at? How do I build trust? But it also allows you to build on your experience to say, "How do I choose what they're good at? How do I trust them to do the things that I want them to do? And how do I want them to communicate with me as they proceed?"
And this is a conversation that we talk about quite a bit where we talk about not human in the loop, but human on the loop. Human on the loop is overseeing what are we working on? What is the outcome of this? How does this make sense given everything else that's in my life?
Devin Liddell: This came up a couple years ago. A newspaper in Europe actually asked me to prognosticate about a technology that I thought could be transformative in the next decade. And one of the things that I offered up to them was you could imagine actually fairly easily the confluence of some existing technologies. You mentioned wearables earlier, but you can imagine the confluence of present day in-ear devices, for example, and maybe some glasses.
You talked about glasses earlier, some smart glasses. You combine those with some machine learning and some AI help. And what you could arrive at is what I described as sort of like a back channel superpower, meaning that you and I are in conversation with one another, but I'm actually getting a back channel in my ear that's giving me sensorial powers that I wouldn't otherwise have. Meaning they're like, "Oh, it looks like Ed is interested in what you're talking about." Or, "No, it looks like Ed is actually kind of agitated or bored with what you're talking about."
The reason I bring that up is I'm curious from your vantage point, and this might be a segue into what you're up to at the Edge AI Foundation. Is there a interface that we're missing to accomplish some of the things you're talking about? When you talk about human on the loop, what are the ways in which we'll actually be interacting with AI models a decade from now that we don't yet have that we need to interact with this diversity of touch points?
Ed Doran: Great questions. Let me talk about two things that I think you just brought up because they're both super interesting. I think if we look back at the history of innovation, they're very seldom completely to the side. They're usually integrated with existing technology. I spend a lot of time working with and talking to people doing self-driving vehicles. That is a fantastic example of integrating old technologies. Right now we're actually retrofitting existing cars.
And as we evolve towards a bespoke technology, and I would argue we have defined very few of them. We have co-opted old interfaces. I was in a self-driving car the other day. It still has a steering wheel for some reason. It's there to psychologically comfort me as if I could grab it if I didn't like where this thing was going, but it doesn't need to be there. And so the question will be, how do we define the future interface for interacting with this? And I would love to see more design work on that. What is the interface for working with a truly autonomous vehicle? What should that look like? What is the interface we should have if I'm on the factory floor and I'm surrounded by AMRs, robots, cobots, humanoids, robot arms? What interface do I have? If I'm surrounded by human beings, I can make eye contact, I can gesture, I can say things, I can tell what's going on.
But what is that now? What is the interface when you're the person who's sitting on top of the factory and you're trying to orchestrate all of these types of devices? This is all new ground. If I was a designer today and I was thinking about the future, I would realize that there is a vast ecosystem of intelligent devices that need a new way to interface in order to bring those senses that they have that I don't, to bear.
They know when a machine has too much vibration and it's going to break. They need to tell the shop foreman, but how will they tell them? They know that car is coming down the road, but it's hidden behind a building. How do you tell the riders about it, or do you? They know that they're going to change altitudes in the plane to get a smoother ride. How do they tell you? This is a great opportunity for not only to build new interfaces, but to express the value of AI, which is something we desperately, desperately need to do. It's a huge opportunity for designers.
Devin Liddell: These are the kinds of questions that the Edge AI Foundation where Ed is the VP of Strategy are asking. Their focus is moving AI off the cloud, away from data centers and to the Edge, meaning the AI is inside the devices themselves.
Ed Doran: So we've done a bunch of work in the cloud, right? At my time, both at Microsoft and at Google, and even now. The cloud is an exquisite triumph of science and technology. These are data centers running incredible models we would've loved to have had years ago, but that doesn't mean they're right for everything that you want to do or everywhere you want to go. It's a little bit like saying, a giant data center for a hyperscaler is a little bit like an F1 car. It's a triumph of engineering, but even Max Verstappen isn't driving that to Trader Joe's to get his groceries at the end of the day. I mean, I bet he could. I can barely navigate that parking lot, but I bet he could. So the question is, how do you take all of those accomplishments, all of that exquisite science and engineering and extend it into the real world?
And that takes a bunch of new opportunities. That means things like on-premise, where you take all of those data center capabilities and models and you put them on-prem. That can be a pristine surgical suite that helps the hospital keep their patient's data safer and more private. That can be a container sitting in the Australian Outback outside a mine dealing with all the heat and vibration and dust, but still running the entire operation.
If you thought about a big mining operation, it's like a city underground, and safety is paramount. It's a submarine. We've literally talked to a retired admiral who talked about, "I need AI to do my job, but I can't surface my submarine every single time I want to ask a question." So the cloud is not an option. So there's all these flavors of on-prem. There's all these flavors of vehicles from things deep in the ocean to riding on the surface of the earth, to flying through the air, to the satellites screaming over your head.
Those are all edge AI in a vehicle. There's things like robot arms, humanoid robots, AMRs. There's plenty of other types of bespoke AI devices. There's the smart device that's sitting in your pocket right now. There's more and more AI in your laptop and in your phone than ever. And there's even intelligent sensors distributed out in the field. And if you move AI out in the field, you reduce the amount of bandwidth you need, you reduce the latency to get things done, you reduce the power necessary to get things done.
You can do really unique scenarios, like flood awareness, wildfire prevention, earthquake awareness, et cetera. So I think of edge AI as the next generation of AI. We build on all of the lessons that we've built up in these data centers and we interact with them to optimize, but we bring AI into your real life.
Devin Liddell: And to riff on that for a second, as AI shifts more and more to the edge and there's this proliferation of sort of AI-enabled, AI-empowered devices, there is a sense that it will be harder to parse where AI begins and where it ends, meaning that it will be in sort of almost everything. Is that a fair way of sort of understanding where you think this is headed?
Ed Doran: It's a great question. I think the really interesting question for us to ask, you and I and your audience is, where should AI be? And what should AI be doing. So have you ever had that person in your life who won't shut up? They just talk, talk, talk, talk, talk, talk, like, listen-
Devin Liddell: That might be me, by the way. So it be me.
Ed Doran: Well, yeah but you're delightful. Someone who isn't delightful, like that neighbor and you're like, "Listen, Bob, I'm so glad your colonoscopy is going well, but I really don't need to hear about that today." Apply that to AI. What job should AI do and when should AI be connected? So if AI is something in wearable, do you need to share every single thing about your personal health or your personal life to the cloud? Probably not.
You probably only want certain things shared. And so edge AI keeps that information with you instead of oversharing it more broadly. If you're putting sensors in the field because you're counting endangered species or you're tracking wildfires or you're avoiding floods or earthquakes, you need those batteries to last a really long time. So you want edge AI because it'll only pop up and send an alert when it's absolutely necessary. So the thing about edge AI is, because of the constraints that it works under, it is fundamentally more judicious about what information's collected, what information is processed, and what information is shared.
And I think that's a really good way to think about things because that's why it feels like the next generation. The thing you brought up a moment about connectivity is, we're actually moving into a point where connectivity is not only no longer assumed, disconnected states are there by design. So if you're working in defense, disconnected silence is survival.
There's no way to know the sensors are there. There's no way to shut them down. If you're deep in the ocean in a submarine, disconnected is necessary for this scenario. If I'm walking down the street and 3000 pounds of iron and glass in the form of a car is speeding towards me, I would like to know that if it loses connectivity, it's still going to keep me safe. So there's an entirely new opportunity for design thinking around what information should I be collecting for my AI?
What decisions is my AI making? And when does my decision connect and share that information or those decisions outside of the device? And that I think is going to deliver a ton more value to the world with the next generation of AI.
Devin Liddell: A quick question on that front though, by the way, what are some of the technical challenges of moving AI to the edge? I mean, I imagine that this is not as simple as just, oh, let's move it over here. So what are some of those key challenges?
Ed Doran: Everything. So I have people who are working on the hardware challenges. Can we make sure this hardware is secure by design? So if I build up a whole bunch of IP and put it in by drone to go out and count endangered species on the side of a mountain, can someone just knock that out of the sky and then take a hammer, knock out the chip and then walk away with my IP? So security by design is part of it.
Energy management by design, which is how do we be as energy efficient as humanly possible? Model design, what kinds of models should we run? How big do they need to be to actually accomplish this task? We have people building new types of models and new ways to use models. We have people building new ways to orchestrate things. So if you are the person who's running the hospital, how do you keep track of collecting all the data that you want, but not sharing any of it outside the hospital to protect people's privacy?
So whether you look at the hardware piece, the software piece, the services piece, the model piece, all of that is evolving incredibly rapidly right now. But that's when you want design thinking. You don't want to wait till everyone's done and then say, "Hey designer, can you come make this right?" You want them on that journey so you can have those conversations about how do you interact with it? How do you control it? How do you make sure this is delivering value?
Devin Liddell: That's fantastic. I love that. And now in honor of Teague's 100th birthday, it's time for the lightning round. What are the most under-hyped and over-hyped innovations of the last 100 years?
Ed Doran: All right, this is a controversial answer. I personally think autonomous vehicles are wildly underappreciated. People don't appreciate how good they've gotten. The last data I was looking at from one autonomous vehicle company was they were 13 times safer than the average human, which is massive. I think the use of AI for derivative products is grossly over-hyped. "What if we made this AI?" Well, that's not a very interesting question. That's like back when people said, "What if this was on the internet or what if this was mobile?" Who cares? What does it do for your life?
Where I think the long-term trend where things get really interesting is, when we start applying AI to things that are the three Ds, dull, dirty, and dangerous. So human beings are not meant to stand at an assembly line and put in three screws every four seconds for 10 hours.
That's not what humans are for. So replace it, move on. Let humans do good things. Dirty, we don't need human beings to crawl through sewage pipes or coke ovens or scrub the insides of industrial washers. It's not a good human job. Replace it. Increase people's quality of life. Or dangerous. People do very, very dangerous things. We had a project a little while back where people would have to go up into the mountains and hike around for months just to look at power lines and see where there might be problems for branches knocking out power. Go send a drone.
Devin Liddell: Fantastic. Ed, this has been an awesome conversation. Thank you so much for being here today.
Ed Doran: Oh, thank you so much for having me. It was a lot of fun. I'm looking forward to hearing some new updates from Teague.
Devin Liddell: Thanks again to Ed Doran for a fantastic conversation about where AI is today and where it's going.
If you find yourself facing a complex problem you'd like help with, we'd love to hear from you. Visit us at teague.com or send us an email at hello@teague.com.