How AI will bring new apps, experiences to maps

Overview

Are we at a crossroads in mapping innovation? In this episode of Today in Tech, host Keith Shaw speaks with Marc Prioleau, executive director of the Overture Maps Foundation, about how artificial intelligence, open data, and community-driven efforts are reshaping the future of digital maps.

Discover:
* Why map data is so difficult (and expensive) to maintain
* How open-source mapping can challenge tech giants like Google and Apple
* Real-world applications of spatial data β€” from disaster response to AI travel guides
* The role of generative AI and machine learning in map creation and validation
* The surprising importance of reducing mapping bias
* How augmented reality and digital twins could redefine how we explore cities

Whether you're a geography geek, a tech leader, or just someone who relies on your GPS daily, this deep dive into modern mapping will change how you view the world β€” literally.

πŸ”— Learn more: https://overturemaps.org

πŸ‘ Like, subscribe, and share if you're passionate about tech innovation. Follow TECH(talk) for the latest tech news and discussion!

Register Now

Transcript

Keith Shaw: Maps have been around since the beginning of exploration, but digitization over the past few decades has brought tons of new applications that help people get from point A to point B.

But are we stuck at a crossroads with the technology, or can artificial intelligence and other innovations drive us forward with newer, cooler mapping? We’re going to try to navigate through all of this on this episode of Today in Tech. Hi, everybody. Welcome to Today in Tech.

I'm Keith Shaw. Joining me on the show today is Marc Prioleau, the executive director of the Overture Maps Foundation. Welcome to the show, Marc. Marc Prioleau: Thank you. It's very good to be here. Keith: Did you notice that I got a bunch of mapping puns into the introduction?

Marc: Well done.

The sign of an experienced broadcaster.

Keith: First off, before we get into some of the main topics, tell us about the origins of the Overture Maps Foundation. As executive director, what are the goals of the organization? How do you aim to benefit both companies in the mapping space and end users in general?

Marc: Yeah, yeah.

So Overture was started in December of 2022, so we’re about two and a half years old. It was founded to address some key challenges. Obviously, as you mentioned, mapping has always been important to society, but it’s become especially critical in an increasingly digital and mobile world.

Mapping is vital because it ties what you're doing to a placeβ€”and place is something we all exist in. Many companies and organizations have a big interest in mapping, but it remains incredibly difficult to do well. The reason for that difficulty lies in the data.

If you think about how map stacks work, what you’re seeing is software rendering data. That data consists of digital descriptions of the physical worldβ€”everything you see around you, rendered into some digital form, and then presented as a map, a route, a plan, or guidance and analysis.

The problem is that the world is vast and constantly changing. So if you want to build a digital model of the world, you have to map it, then remap itβ€”and then remap it again. That’s why we started as an open data project.

It became clear that, at least for some types of map data, many people would prefer to treat it as a shared asset. Rather than having every organization build the same things independently, we can collaborate to create foundational data together.

Then, each group can add their own unique layers on top of that. That idea was a big part of the origin of Overture. Keith: Right.

And when most people think of maps, they probably think of the two or three major private companies that dominate the mapping world. Just look at your phone and you’ll know exactly who I’m talking about. Marc: Yeah.

And part of the reason a few companies dominate is because the data is so difficult. With all due respect to my software engineering friends, the software is complex and impressiveβ€”but the data is really the limiting factor.

It's what makes mapping unattainable for most companies unless you're willing to spend billions of dollars every year, indefinitely. When we looked at that challenge, we saw another trend: the evolution of digital maps.

Since the launch of the iPhone, for example, navigation apps have gone from simply getting you from point A to point B to now providing minute-by-minute accuracy, lane-level guidance, and context for complex intersections. All of that requires even more detailed data. And it’s not just navigation.

Every area of mapping is seeing this data challenge grow alongside rising user expectations. So the problem keeps getting bigger. One company collecting and maintaining all that data alone just isn’t feasible.

Open data provides a way to structure that data so we can keep pace with what's being builtβ€”whether that’s applications, routing, or analysis. Keith: Yeah.

And since we're naming namesβ€”Google and Apple are probably the two big onesβ€”how much of the world is actually mapped at this point? Have they gotten close to full coverage? I’m going to take a guessβ€”maybe 85% of the world is mapped? That’s my estimate.

What do the numbers actually look like?

Marc: I’ll give you the answer you’re looking for, but I’ll also tell you why the question is more complex than it seems. It’s hard to know for sure, but one way to look at it is through open data.

There’s a project called OpenStreetMap that just celebrated its 20th anniversary. It’s a community-driven initiative to build maps, and they estimate they’ve covered about 80 to 90% of global roads. So your guess was actually pretty good.

The way that’s calculated is by comparing public figures on how many miles or kilometers of roads exist in a region against the roads they’ve mapped. Generally speaking, in urban and developed areas, the coverage is close to 100%. In more remote regions, it's much lower.

But here’s why the question is tricky: maps aren’t just roads. Increasingly, they include buildings, places, addresses, and boundaries. And as maps power richer applications, the associated dataβ€”like lane details, speed limits, signage, and turn restrictionsβ€”becomes even more essential.

So the question isn’t just β€œis it mapped?” It’s also β€œhow richly is it mapped?”

Keith: That reminds me of being back in elementary school when we first learned about maps. I think most people, when they hear β€œmaps,” probably just think of GPS on their phones and basic street-level directions.

But there’s so much moreβ€”topographic maps, geographic data layers, and all this other information that people either forgot about or never realized existed. Marc: Exactly.

Maps are made for humansβ€”something we look at to interpret and act upon. But there’s also a growing amount of spatial analysis that machines handle. Take flood mitigation, for example. Or autonomous vehiclesβ€”that’s a completely different kind of map.

It’s not for humans to read, it’s for machines to interpret and make decisions from. Even basic maps differ depending on purpose. A driving map is different from a bus map. A shopping map is different from a tourism map.

So there really is no β€œone map.” There are some shared physical realities, but each map is unique to its use case. That’s why open foundational layers make so much sense. Everyone needs to know where roads are, where city boundaries fall, and where addresses lie.

So let’s collaborate on those shared layers, and then let organizations build unique features on top of that.

Keith: When we first started discussing this episode, I wondered whether diving into the nitty-gritty of maps would even be interesting. But the question that really got me hooked was: what kinds of applications are emerging that use mapping dataβ€”both from the private sector and open-source efforts?

You listed some really cool examples during our prep call. Can you share what’s exciting you the most about where this is all going? Marc: Absolutely.

If you think about maps in terms of spatial dataβ€”well, that’s just one of the constants of life, right? Time and place. Where was I when something happened? Where is something going to happen? Where do I want to go?

So if you view maps as a universal framework to define and measure things, the use cases become almost limitless. Yes, everyone thinks of driving directions first. But there’s also local search, trip planning, enterprise applications like determining store locations, analyzing foot traffic, or studying demographics.

Global development organizations want to understand economic growth, which ties into infrastructureβ€”roads and buildings. Humanitarian response teams need to know where people are, where resources are, where to send aid during floods, fires, or disease outbreaks. We’ve even worked with the World Health Organization on Ebola outbreaks.

They need to map where people liveβ€”often inferred by building locationsβ€”and where vaccines and hospitals are. Nearly everything ties back to place. That’s why spatial data is so valuableβ€”and why we need to keep improving how we collect, manage, and share it. Keith: Right.

Here's an example from my own life. The company I work with is searching for a new office building. We had to vacate our current location, and now the team is looking at different options.

They’ve got all this data on where employees live, and they’re using mapping and spatial analysis to figure out the most convenient building locationβ€”so that we don’t end up forcing 80% of our people into an hour-long commute.

That’s not just about where people liveβ€”it’s also pulling in commute times and access routes. Marc: Right.

And in some organizations, that location might be weighted toward wherever the CEO lives.

Keith: I can’t believe a company would do that! Who would ever...

Marc: (laughs) But seriously, think about real estate. One clever tool I saw asked you to enter your work address, and then it would show you all the neighborhoods within a 25-minute commute during rush hour. It’s not just about distance anymoreβ€”it’s about real-world travel patterns.

That gives you a shape, a polygon, of where it actually makes sense to look for housing. It's a more personal version of the example you just gave. Keith: Right.

And a lot of cities are exploring smart city concepts. We just did a show about quantum cities, and they’re sitting on a ton of data. Are you seeing more companies or municipalities doing things like digital twins or city overlays?

Marc: Very much so. Cities are in a kind of race right now. Technology has advanced quickly, and cities are looking for ways to use it for resource planning, smarter development, better transportation optionsβ€”you name it. But the problem is legacy systems.

I spoke with the CIO of a major U.S. cityβ€”I won’t say which oneβ€”and he told me that three different departments in that city maintain their own separate street maps for their specific needs. That’s a lot of redundancy. He asked if we could help with open data.

I said, β€œSure, I can help technically... but the political side? That’s trickier.” Still, that’s the point of open data. These departments may believe their use case is uniqueβ€”and it might beβ€”but the roads themselves are the same.

Instead of duplicating effort, let’s agree to collaborate on the foundational layers: roads, addresses, boundaries. And then each department can build their custom applications on top of that. The power of an open project like Overture is that no one organization controls the data.

We’ve got about 40 member organizations now. The data is open, free to use, and maintained by the community. And here’s the key: a road’s location isn’t a competitive advantage. It’s public.

I live on a roadβ€”it’s been there a long time, and it’ll probably still be there in 50 years. Maintaining that data isn’t where companies make their money anymore.

So instead of competing on that, let’s pool our resources, make that base layer as good as possible, and let everyone build their unique layers on top. Keith: Right, right.

And that leads to my next question, which you’ve already started answering: What are the benefits of moving toward an open-source mapping model? But one thing we discussed in our prep call was the issue of mapping bias.

Can you explain what you mean by that, and why it’s better to minimize bias in mapping? Marc: Definitely.

Mapping bias is especially important to think about in an open project like Overture. Bias happens when the mapperβ€”whether it's a person or an organizationβ€”decides certain types of data are more important than others.

For example, if you and I walked down the same street full of shops, I might pay attention to certain places while you notice others. If we each mapped the street, our maps would reflect different priorities. Now scale that up.

Overture was started by Microsoft, AWS, Meta, and TomTomβ€”four big tech companies. But as we’ve grown, we’ve brought in a more diverse group of members. That’s important because it allows us to represent different needs and priorities.

For instance, humanitarian organizations working in the Global South need very different data than tech companies based in the Global North. They may want to map land use for environmental or economic development purposesβ€”data that automotive companies might not care about.

The challengeβ€”and the opportunityβ€”is to bring more people into the project who can say, β€œHere’s what’s important to me.

Here’s what I need mapped.” When they bring those needs and the corresponding signals, we can prioritize and gather that data.That’s the real strength of an open platform: anyone, regardless of sector or geography, can use these base layers and then attach their own data to them.

Ownership details, building height, 3D models, tax dataβ€”whatever is useful for their application. And we’re building Overture to make that process open, stable, and universally accessible.

Keith: And of course, we haven’t gotten through a single episode of this show lately without talking about generative AI or artificial intelligence. So, this is that part of the show. You can’t escape it!

What kind of impact are you seeing around generative AI or other machine learning and deep learning tools in the mapping space? Don’t talk about apps yetβ€”we’ll get to that. What are you seeing on the back end?

Marc: No surprise hereβ€”we’re still at the early stages. But we’re already seeing some very promising developments.

Someone said something recently that really stuck with me: β€œAI is as bad as it's ever going to be right nowβ€”it’s only going to get better.” I think that’s a great way to look at it. At Overture, we think of AI’s role in two ways.

First, how it helps end users access and use maps. That’s the part most people think about. But secondβ€”and more important to usβ€”is how AI can help us build, validate, and improve mapping data. That’s the back-end side we’re heavily invested in. Historically, map data came from surveys.

Go all the way back to Lewis and Clarkβ€”or Spanish and European explorersβ€”they were out in the field, observing and recording. And up until just a few decades ago, the best maps were the ones most recently surveyed.

Then phones became location-aware, and suddenly we had sensorsβ€”billions of themβ€”feeding us information. This gave rise to the concept of remote sensing: using satellites, phones, or other devices to observe and infer data. That’s been the major evolution in the past 15 years.

Now we’re entering a new phase, where large language models and AI agents can help create or validate data.

Let me give you an example: if you asked a language model, β€œDoes Keith’s Hamburger Restaurant in Elko, Nevada still exist in 2025?”—the model might search through reviews, social media, or other indicators and say, β€œYes, we’ve seen activity suggesting it’s still open.” That’s powerful.

In the past, you’d have to send someone physically to verify that. Now, with agents repeatedly checking various sources, we can pull together signals to streamline and validate data. I think that’s going to revolutionize how we build maps.

Keith: Are there other innovations you’re seeingβ€”or have you covered most of them? I think you mentioned voice guidance during our prep call.

Marc: Right, now we’re talking about what users will actually experienceβ€”what people see and interact with. We focus a lot on the back-end, but the front-end is where the magic feels real to most users. Take local search. That’s one of the most common mapping applications.

You’re looking for a restaurant, right? But the way we phrase our searches is limited by the tools we have. I was doing this with my family recently. We asked, β€œDo we want Italian? Mexican?

Thai?” But what we really wanted was something casual, not too expensive, with a fun atmosphere. That’s not something you can type into most search engines effectively. But that’s a perfect prompt for AI.

A model could understand your contextβ€”who you’re with, what mood you’re in, how much you want to spendβ€”and give you better recommendations. That’s a major leap. AI is helping bridge the gap between what we’re thinking and what we can ask. That’s huge.

Marc: You’re also seeing this in AI-powered travel guides. It’s not just β€œWhat’s a good hotel?” anymore. It’s β€œPlan me a three-day trip to Sicily,” based on my preferences and interests. And those answers can look very different depending on who’s asking. Cities are exploring AI, too.

They want to optimize traffic flowsβ€”not just for individuals but across the entire system. Today’s routing apps optimize you. But what if we could optimize the whole city? And disaster response is another area. I live in California, and we’ve had massive wildfires.

Emergency response requires pulling data from local, federal, and private sources, all in real time, while the media is standing outside demanding answers. AI can help combine and analyze that information much faster.

Keith: Yeah, I want to go back to a couple of those examples you mentionedβ€”like the AI-powered travel guides. That can be both a blessing and a curse. Sure, it’s great if the AI knows what I like and helps me plan my trip.

But I’m thinking about tour guides, for example. Sometimes, the best part of a tourβ€”like a Boston Duck Tourβ€”isn’t just what you’re seeing. It’s the person telling you stories, cracking jokes, making it fun. Wouldn’t an AI-generated experience be kind of bland or repetitive? Marc: (laughs) Yeah.

The joke I’ve heard is: an AI walks into a bar, and the bartender says, β€œWhat’ll you have?” And the AI says, β€œI’ll have what they’re having.” It tends to aggregate and synthesize existing content rather than create something novel.

Personally, I’m a big believer in meaning and purpose for human beingsβ€”especially in roles like tour guides. The value is in the personality, the insight, the humorβ€”those human elements that make the experience memorable. That’s not to say AI won’t revolutionize a lot of the foundational work. It will.

For example, at Overture, we’re trying to build open map layersβ€”where are the roads, what do the lanes look like, that sort of thing. That work is incredibly hard to scale, and AI can help automate and streamline those processes significantly.

But the more commoditized things become, the more value will shift to personalization and storytelling. That’s where humans shine. Like you saidβ€”get the right guide on a Duck Tour, and it makes the whole trip.

Keith: Yeah, and we also talked about how cities might be able to use AI to reconstruct the past. If you have enough historical data and combine that with modern mapping, you could recreate what an ancient city looked like.

That would be amazing in Bostonβ€”being able to hold up your phone and see what the city looked like in 1776, or even earlier. I mean, in the Fenway area, you’d be standing in what used to be a swamp! That would be really coolβ€”especially for history lovers.

Marc: I love that idea. When I was a kid, we traveled a lot, and my mom had these books with overlaysβ€”plastic sheets you’d flip to reveal how ancient ruins looked in their prime.

You’d see the modern-day Coliseum in Rome, then flip a transparency to see the original structure from the days of Imperial Rome. The digital version of that is comingβ€”augmented reality. And it’s not just for consumers. We have members like Niantic and Meta working on AR applications.

Imagine standing in a city and seeing a digital overlay of what used to be there. That’s powerful. Another application we’re seeing is for city planning. Let’s say there’s a vacant lot, and the city wants to put up a new building.

Using AR and mapping, you could point your phone and see a projection of the building in real space. You can evaluate: does it block your view? Is it too tall? Does it fit the area? That kind of immersive planning is already happening.

Keith: Yeah, the closest thing we have to that in my hometown is the historical society posting old photos. They'll say, β€œThis was a drive-in from the 1950s,” and show a side-by-side.

But unless you remember exactly where that photo was taken, it’s hard to connect the present with the past. Marc: Exactly.

That’s a fascinating technical challenge. It’s not just about having the historical mapβ€”you also need to know precisely where you are, the orientation of your head, what direction you’re facing. That’s called β€œpose” in the technical world.

The mapping part is important, but so is aligning your physical experience with the digital overlayβ€”especially in wearables. One of the AR features I find most useful is when you come out of a subway station in New York.

You pull up your phone and try to figure out which way to walk. You head one way, realize the blue dot is moving the wrong direction, and have to turn around. I’m pretty sure all the New Yorkers are laughing at us.

Keith: Oh, they’re definitely laughing at us.

Marc: (laughs) But seriously, AR can help with all of that. And beyond tourism, there are real applications for city planners. Imagine a proposed developmentβ€”you could walk around and see a live 3D projection of what it would look like from the sidewalk.

That kind of visual context could really help with public feedback and decision-making.

Keith: I wanted to ask about privacy, data, and security from that perspective. We had a guest on the show a few weeks ago who talked about open-source intelligence and how easy it is to gather information on people.

It feels like if the world is moving toward open-source mapping, that might be ripe for abuse by bad actors. So how do you approach security when it comes to open maps?

Marc: That’s a really interesting question. I’m not an expert in open-source intelligence, but I do know people in the intelligence community. One of the keys to their craft is that no single source gives you everything. You gather little pieces from everywhere, and then you stitch them together.

So yesβ€”anyone putting information out there could be contributing to that broader puzzle. But in terms of what we map, we focus on things that are publicly observable. A road is public. An address is public. A building is public.

If it's something you can walk down the street and see with your own eyes, that’s within the scope of what we map. And yeah, could someone use that for intelligence? Sureβ€”but again, these are things already in the public domain.

Take an address for example: β€œ123 Maple Street” isn’t a secret. It has to be publicβ€”Amazon needs to deliver there, Uber needs to find it, your mail needs to arrive. What is private is who lives there. We don’t collect or expose that kind of information.

Keith: Okay, but you're not going to make it easier for people to add those overlays or link personal information to maps, right?

Marc: No, absolutely not. The only thing we might do is provide something like, β€œ123 Maple Street exists at these coordinates.” That’s public data anyway. But we don’t link that to a person.

Keith: AI will probably figure it out anyway. It’ll say, β€œ123 Maple Street? That must be Keith’s house.”

Marc: (laughs) Yeah, I suspect that with 10 or 30 minutes of effort, someone could find out where we live. That’s not even necessarily an AI thingβ€”it’s just the reality of how data is shared today.

But again, what we map are things that are already public: roads, city boundaries, buildings. For example, Boston’s city boundary needs to be publicβ€”otherwise, Boston and Newton might fight over who collects taxes where. We don’t map anything personally invasive.

Even in our schema, there’s no field for β€œname” or β€œresident.” That’s just not part of what we do.

Keith: So, no list of who lives where. Got it. And kudos on pronouncing Worcester correctly, by the way.

Marc: (laughs) Yeah, I’ve learned it’s not β€œWorchester.”

Keith: I want to end with a philosophical question for you, Marc. I mean this as a compliment, but you're probably the biggest map geek I've ever met. As we talk about all this amazing mapping technology, do we still need to teach kids about maps and directions?

We still teach geography, sure, but I wonder whether knowing how to get somewhere is an innate skill people still need. Are we going to rely on tech too muchβ€”like how cursive writing has basically disappeared?

My third child is about to start driving, and I’ve gone through this twice already. I always try to get my kids to be curious about where they're goingβ€”like, what happens if GPS doesn’t work, or the phone dies?

I’m reminded of that Robert Frost poem, β€œTwo roads diverged in a wood…” If GPS had existed back then, he probably would’ve just taken the road more traveled.

Marc: Yeah, I think what we’re starting to seeβ€”more and moreβ€”is the unintended consequences of technology. You see it with smartphones, with social media, and I do think it applies to mapping too. When I was a kid, we’d pull out a paper map.

And there were two types of people in the worldβ€”those who could fold it back up, and those who couldn’t. But those maps gave you a bigger-picture understanding of place. You could look at the San Francisco Bay Area, see the peninsula, the water, the bridges, the cities.

You built a mental model. But now? I was in Kansas City recently. I landed, opened my phone, typed in an address, and drove 20 miles without having any idea what Kansas City actually looks like. I didn’t know if I was going north, south, east, or west.

That sense of spatial awareness is lost when you’re just following turn-by-turn instructions. When my kids were learning to drive, I’d ask them, β€œDo you have a mental map of this area?” Because I think we lose something when we don’t.

Keith: Yeah, I still love going on Google Maps or Google Earth and zooming out to see where things areβ€”especially if it’s a place I’ve never been. Maybe that’s a Gen X thing. We grew up with atlases and road maps.

Marc: Google Earth is actually a fantastic exploration tool. It’s not the main navigation app, but it’s great for zooming, panning, and seeing the terrain. I remember during the first Gulf War, I think CNN used Google Earth to show people where military activity was happening.

It helped people visualize places they’d never been. More maps are now adding terrain, 3D buildings, and those kinds of features. That’s great. But what’s really valuable is developing a sense of contextβ€”understanding where you are in relation to everything else.

Sometimes, GPS takes us so directly to a solution that we miss all the interesting stuff around it.

Keith: Yeah, I’ve got a friend at the Massachusetts Department of Transportation, and I’m always pestering him about road design. Like, why does this street go this way? Why doesn’t it go straight?

And he’ll just say, β€œLook at the map.” A lot of times, it’s because there was a hill or some obstacle, so they went around it instead of through it. And suddenly it makes sense.

Marc: Yeahβ€”the interaction between geography, topography, and infrastructure is so fascinating. I thought in Boston it was all based on where cows wandered?

Keith: Oh yeah, especially downtown. A lot of the roads were originally horse and carriage paths in colonial times. That’s why Boston’s downtown street grid is so chaoticβ€”it’s just layers of history built on top of each other.

Marc: Boston’s a perfect example. It’s got a river running through it, which you can only cross at certain points. So if you’re trying to get from MIT to Boston Common, you’ve got to pick the right bridge.

If you grew up only using GPS, you’d be like, β€œWhy can’t I go in a straight line?” But a good map gives you that context.

Keith: That’s how we confuse the tourists. It’s like New Yorkers watching people walk out of the subway and head the wrong way.

Marc: (laughs) The Big Dig still confuses me. Even after all these years, I’ll drive through and pop up somewhere, and I have no idea how I got there.

Keith: Yeah, same here. It’s fun living in an old city.

Marc Prioleau, thank you again for being on the show. We talked way more about maps than I expectedβ€”but I think you’ve officially turned me into a bit of a map geek myself.

Marc: We all have a little bit of map geek in us. You’ve just got to lean into it.

Keith: For anyone who wants to learn more, where should they go?

Marc: I work with Overture Maps β€” you can find us at overturemaps.org. I’m also on LinkedIn and the usual social platforms. Keith: Awesome.

Thanks again, Marc. That’s going to do it for this episode of Today in Tech. Be sure to like the video, subscribe to the channel, and leave your thoughts in the comments if you're watching on YouTube. Join us every week for more episodes.

I’m Keith Shaw β€” thanks for watching! Β