不動産・歩ける街・気候変動対策のためのAI ― John Renne教授との対話

What if two hours of work could replace a multi-year, half-million-dollar research grant? Today's guest, a professor at a prestigious university, says AI has made that his new normal. Welcome to episode number 1039 of the Super Data Science Podcast. I am your host, John Crone. Today's guest is Dr. John Rennie, professor of real estate and urban planning at Tulane University, where he directs the real estate development program. John is a world-leading expert on walkable communities, and even though he was already a tenured full professor, he went back to school for a master's in AI. In this episode, John explains how AI let him scale up his urban planning research by orders of magnitude, how computer vision is playing a role in real estate, and how his startup uses AI to tackle the scourge of students' AI-ridden essays. Enjoy this one. This episode of Super Data Science is made possible by Anthropic, Dell, NVIDIA, Origin, and Palo Alto Networks. John, welcome to the Super Data Science Podcast. What a treat to have you on the show. How's it going?

Great, John. It's wonderful to be here. I've been a long-time listener, so it's a real pleasure and honor to be a guest on your show. Yeah, that's how we actually know each other, I suppose. So through you listening to the podcast, you reached out to me about a year ago, and then I did an hour-long presentation. I did a half-hour presentation plus half an hour of Q&A with you on stage at a Tulane University event in the Northern Hemisphere spring of this year. And that led to you asking me to teach with you an innovative new course at Tulane. And so now I'm adjunct faculty at Tulane. We're going to talk about that course later on. But I really appreciate the connection. I really appreciate you reaching out. Yeah, you know, it's interesting, the power of kind of, you know, telecommunications, you know, it's like I learned about you because my professor assigned your book, Deep Learning Illustrated, and then recommended your podcast.

And that was about five or six years ago. And I started listening and learning a tremendous amount because it was brand new to me. Everything was brand new to me. I don't have a background in, you know, computer science until, you know, until I started this journey as a second master's. But, you know, as an avid listener, you know, I said I was at Tulane and we were planning a AI conference. And I said, well, we got to have John Crone as a speaker. And you certainly did not disappoint. So thank you for coming to New Orleans. It was great. Also, my first trip ever to New Orleans and had a great time as well. Stayed for a weekend after. It was not the Jazz Fest. That was like a week or two later. It was a French Quarter Fest. Yeah. A lot of people don't know the French Quarter Fest is better than Jazz Fest in many ways because it is free. Jazz Fest is expensive. I highly recommend Jazz Fest as well and Mardi Gras. Yeah, and it's not like it was small.

It's hard for me to imagine that Jazz Fest would be bigger than French Quarter Fest because everywhere you went in the French Quarter, there were stages. There were dozens of stages with music playing all day. And because it was free, you could just go around anywhere you wanted and just take a punt. Yeah, really cool experience. But let's talk about how just, you know, you're a professor and you were just talking about how you recently had a professor. So it's kind of an interesting thing. So you are a professor of real estate and urban planning at Tulane. Yeah. And you direct Tulane's real estate development program. But while you were already an academic, so established for decades as a university professor, you know, going climbing up the ranks assistant, associate full professor, you decided at some point during your career that you needed to go back to school and do a master's in AI while keeping your full-time role as a professor. So tell us about the thought process behind that and whether kind of everyone should be doing that if they're interested in AI.

You know, it's interesting. I don't know if everybody needs to do it just because the way in which learning is happening these days, you know, there's many different ways to learn. And I didn't immediately expect or think about doing a second master's. What happened was I was supervising. Ph.D. students, master's students, working with junior colleagues from other universities who were assistant professors, you know, kind of new to the field. I like you said, I had been at that point. I had been I was a full professor tenured and had been, you know, basically working for 15 years in those roles. The time I was at Florida Atlantic University. And COVID hit, and I, you know, kind of had this realization that my data science skills and my statistical skills were really, really out of date. I was using SPSS mainly, and I was also, you know, learning other programs like a program called HLM 7.0 and, you know, a little bit with a few other, you know, kind of, you know, major, major statistical packages.

I didn't know how to code at all, and I was asking around. A lot of my colleagues were using R, and some of my students, the younger ones, had learned Python. And they were just telling me, you know, Python is the way to go. And so I said, you know, I want to be able to be competitive, not be competitive, but to, you know, have the same skills that a freshly minted Ph.D. student coming out was having. And I really feel like I need to learn Python. So I started looking around at these different online certificate programs, and I was about to do one of, like, I think Stanford has a pretty popular online data science program at the time. And I decided I was about to start that. And around the same time, Florida Atlantic University launched the first master's in AI, which is in their computer science department, in the state of Florida. So it was the first master's program offered, you know, in the state of Florida.

And I didn't, you know, I mean, I kind of knew this, but I didn't think about it. I was like, well, I could take classes for free because I was working there. And I could take up to six credits, which is basically two classes a semester. I don't recommend doing two classes a semester. I tried that once. And, you know, my wife was like, I don't know if that's a good idea. And, you know, I had to tell her, yeah, you were right, you know, like most things. It was too much. It was too much. But I did one class this semester, and that was perfect because I really enjoyed learning Python. And so I took two undergraduate classes in Python to be able to be qualified to apply for the master's program. And then once I got into the master's program, the first class I took was intro to data science. And then, you know, and then kind of I went on from there with, you know, a bunch of classes. I'm still just wrapping up right now, literally as we speak, wrapping up my master's thesis. So I did this very slowly because, in all honesty, I didn't really need it for my career.

I've really been doing it out of the ability to, A, because I enjoy it, and B, because it's giving me a tremendous boost in my knowledge, which is helping me in all aspects of my career, both inside academia and in several outside academia entrepreneurial projects I've been working on. And we're going to talk about this, too. And so, yeah, when you say you didn't need it, you didn't need it in the sense of it wasn't going to affect your tenured faculty position. You know, you already have a Ph.D. You already have 25 years of research in urban planning under your belt. It's not like a master's in AI is going to, in that, in your kind of your day job, quote unquote, give you some kind of boost. But it's given you huge amounts of new abilities, magical powers, really, as you and I have discussed before. I think it certainly helps.

I've been at Tulane now for nearly two years. and I had some connections to New Orleans because I spent the first 10 years of my career at what's now LSU New Orleans. Back then it was called the University of New Orleans. But I spent the next 10 years at Florida Atlantic University. And I do think that it did make me very competitive to be able to be selected for the position that I have at Tulane right now. Oh, really? Yeah, because I'm probably one of the first academics, you know, out there that was really talking about AI in real estate because I was learning AI. I started really learning AI before it became popular, not long before, but, you know, it was certainly before ChatGPT was released. And, you know, not a lot of people were talking about AI when I first started studying it for the first couple of years. And then obviously now it's, you know, it's the topic that everybody talks about.

So, you know, my cousin Nikki always tells me, you know, it's better to be lucky than smart, but it's even better to be lucky and smart. So I guess I was fortunate to get both. What inspired you about AI before the ChatGPT moment? Like, what was it about AI that made you certain that this was the future or something that really excited you about it? In all honesty, I was interested in data science. I have been analyzing for many, many decades, large data sets. And I love big data, increasingly, you know, trying to get as much data as possible. I was mainly using Excel, SPSS, and hitting limitations, like what I could do with that data. so like I've listened to so many of your episodes and and bought so many of the books that your guests have recommended I think it's Stephanie Moline's book on pandas and you know a bunch of

other ones that I've you know I was exposed to through listening to your show and so I learned all of you know so many I mean I got them sitting here next to me I learned so many of that And it's interesting now because it's so different with VibeCoding today than it was even like, say, a year and a half or two years ago. And I'm so grateful that I have the foundation in understanding things like Pandas and Geopandas and all of the network architecture and models for deep learning and machine learning. Right. Because I think nowadays it's so easy with vibe coding to just, you know, basically get behind the wheel, turn the key and drive the car. But now I really feel like to me, these models are not as black box because I really feel like I understand how the transmission works. And I understand, you know, how to understand how the brakes and, you know, the intricacies of the engine works.

And that gives me a tremendous amount of confidence in doing the work that I'm doing now, because obviously we're not doing pandas today like we were doing three or four years ago, you know? Yeah, you're not writing each character in your Pandas data frame queries or lines of code anymore. But it is still helpful, I imagine, in the now Vibe coding era to not just be, to use your car analogy, to be sitting behind the wheel. You're not just sitting behind the wheel. You can look under the hood and make sure that everything's running as you anticipated. Yeah, exactly. And I could look at the code and I understand what it means. You know, I imagine, you know, if somebody, you know, was kind of a mid-career professional like myself and doesn't understand Python and they're, you know, running these models and looking at the output and the coding, it's like, well, you know, if you don't understand how to read, you know, Python, it's really hard to understand whether or not your models are doing what they want to do.

because it's so dangerous with VibeCoding because it does it all for you. But I do think having that fundamental knowledge is so – I feel so grateful that I was able to get that knowledge before this current era. And I went from, like, a data science era, right, to the chat GPT era, right, and now we've entered the VibeCoding era. So I've seen this tremendous transition that if we were to look and think about, this could have taken decades, but it just happened just a few years. Yeah, it's wild to think what will be coming in the coming decade. I recently had Richard Socher on the show who has a bestselling book called The Eureka Machine. and he's talking about the next decade having as much progress as the previous century. It's unbelievable. Yeah, and we've talked about, you know, there's people who think that that's like a slow,

what they call a slow takeoff scenario. Well, it's interesting because I see it in my day-to-day work. So one of my hats that I wear is that I'm a real estate developer And the stuff that I see AI doing now with planning out floor plans and designing buildings, I mean, I would have never guessed that it could have ever done what it's able to do today. Same. I've said online before how, yeah, finishing my PhD in 2012 in AI, if people had told me in 2020, like in 2026, we'd have these magical abilities to be able to just dictate into our computer and have a web app built in minutes. I'd say that's crazy. There's no way we're going to have that. And it's interesting because I think when you started data science as well, when you started doing these courses, you know, a decade ago, everything was very narrow.

You know, you would build a model to do something so specific. You were like, okay, we have this training data set that's labeled in this way. And so it will very narrowly be able to help us out in this very specific situation. And so this idea of intelligence being so generalizable by taking an architecture like the transformer architecture and just scaling it up, not really having that many breakthroughs in terms of the science of learning, but really just saying, okay, this transformer architecture, this ability to attend over large sections of text or pixels, let's just scale that up. and see what happens. And yeah, OpenAI were the leaders in that initially and just scaling this thing up by orders of magnitude, having more and more transformers on more and more NVIDIA GPUs. And every time they did that, GPT-2, GPT-3, GPT-4, they were shocked.

The people developing these systems were shocked at the heightened levels of intellectual ability, of generalizability that came out of that scaling. What I find, I think that's incredible. And what I find to me, which is so useful, is how, you know, you can ask the model, like, I don't understand, like, I review a lot of academic papers. and, you know, historically, you know, OLS regression models, you know, that's kind of like the old standard, you know, that goes back many decades now. And then increasingly things get more complicated. The models have gotten more complicated. And, you know, statistical models, this is even really before kind of the current modern era. And one of the things that's always bothered me as an academic is how much jargon there is in the academic world.

So when you read a paper in a peer-reviewed journal, right, there's all these terms that, you know, it's hard to understand what they mean. And interestingly, those jargony terms change from subfield to subfield, right? So if you're like an economist, you may call one thing a certain thing. And if you're an urban planner, an engineer or whatever, you know, whatever discipline you're in, those terms change. And what I find incredibly useful is how you can ask the model, like, I don't understand what that means. Can you explain it to me? And then it'll explain it to you. And you can say, well, just give me an example. Explain it to me like I'm a middle schooler. Right. And that to me is what's like I call finding a magic genie in a bottle. Right. It's just it answers every question you want. Now it'll do pretty much any analysis that you want, but it's just incredible because for many, many decades, academics, you know, the ivory tower academic, you know, protected themselves against, I guess, any threats by surrounding themselves by a bunch of jargon.

And I've never, never liked that of it, of the academic world. And especially when you're trying to explain it to a bunch of undergrads, you have to be able to explain it in a way that they can understand it. Because that's really what learning is all about, is understanding and having a common language. That is the reality that a lot of us are facing now. I mean, for me, as a hands-on data science or AI or machine learning practitioner, you know, you talked about Stephanie Mullen's Panda's book. And it was the case five years ago that if you wanted to be working in pandas, you have to read a book like that. And you'd have this knowledge, you know, reading books like that would give you this moat, this technical moat around what you're doing that would make you difficult to replace. that these are highly desirable skills, relatively complex concepts, that not enough people were learning relative to how much demand there was for people with those kinds of data science skills.

And now, as of, I would say specifically, Claude Opus 4.6, which came out in, I think, February of this year, or around that time, since that model's release, the technical moat around my career is gone. It's threatening, but it's also exciting at the same time. I mean, I think everybody is afraid that AI is going to wipe out every job. And at the same token, you know, we can see, you know, a pessimistic future and a bright future. And I always, you know, my wife comes from a farming family in Nebraska, and I always kind of put myself in the mindset, like, if I lived in, you know, the late 1800s when the majority of Americans were farmers, could I envision what life would be like today? And probably the answer is no. So, you know, how do we envision what life is going to be like in 10, 25, 50, 100 years from now? It can be scary. It can be exciting.

But at the end of the day, I think we just got to make the best of it. And, you know, we can't certainly can't put the genie back in the bottle. That's not going to be possible. No. And it's hard to imagine. and it's hard to imagine it slowing down. People talk about scenarios where we've got to slow AI research down or we've got to have more guardrails, but it's hard to imagine that happening because if you imagine the scenario where, and this is already a politically impossible situation to imagine, but you say okay every Chinese lab every American lab even every Russian lab somehow they all put all their geopolitical issues aside they agree that AI is such a big threat that they need to stop doing research on it, then there's still going to be North Korea or somebody who picks up the baton, smuggles some GPUs into North Korea, and is able to create this next level of frontier intelligence.

Like there's no, like you're saying, there's no way to keep the genie out of the bottle and there's no way to slow down what's happening. At least you can't slow down everywhere. So I'm an optimist. I'm probably an eternal optimist. But I do think that I'm very excited for using these tools to help solve very complex problems such as climate change, which I think is one of the biggest existential threats facing, you know, everyone in the world equally. Right. Whether it's wildfires or flooding or droughts or whatever it might be. You know, a lot of my work that I've done over the years in transportation planning and in, you know, trying to build and envision sustainable cities. All of the forecasts are very, very dire about the direction that we're heading. And having lived both on the Gulf Coast and the Atlantic, you know, I see king tides, you know, impacting, you know, my hometown here in West Palm Beach, Florida.

I split my time between West Palm Beach, Florida and New Orleans, homes in both places, two cities, you know, that are very much threatened by climate change. But it's not just these cities. I mean, look at the floods that hit North Carolina a couple of years ago. All over the world, we're facing these issues, and I'm hopeful that we can figure out a way to reduce carbon emissions. And part of that, I think the market-based society, the market-based economy is stronger ultimately than political forces. Now, political forces are very important, right? But I do think ultimately, at the end of the day, we do live in a market-based, you know, economy, global economy. And when companies can figure out, which I believe is happening, how to create sustainable, renewable energy, how to build cities to be able to be more sustainable and do it in a way that is profitable and hopefully, you know,

improves the well-being for the vast majority of people, then I think we're going in the right direction. And that's kind of where I've been focusing my efforts in my career is how do we use AI to create more sustainable development? 100 percent. And I sit in the optimist camp as well. AI is another tool for people to use. And I think the vast majority of people on the planet have good intentions and want, you know, in their immediate circle, you know, like in their household, in their workplace, in the world in general, I think the vast majority of people, like 95% or more, want things to be peaceful and easy. You know, don't want conflict. And you see that, you know, people, maybe people who have like never been to New York City have this idea of it being so like rough and tumble. But then when you're in Manhattan and you're crowding on the subways or, you know, crowded on the sidewalks, like, you know, everybody's trying to be nice to everybody.

Yeah. Yeah. For almost 100 percent of the time. And so, yeah, I also am an optimist because it's going to be that vast majority of people who are using AI tools to defend ourselves against the attacks that become easier because of AI. I hope lots of our listeners, as I am regularly encouraging them, are trying to come up with positive ways of making an impact with AI. Let's dig into some of the ones that are specific to you. So you're a world-leading expert on something called transit-oriented development, T-O-D, or TOD, I guess you might say for short, to have some jargon to give you a little bit of a moat in your subfield. We can call it TOD. Tell us about transit-oriented development. Yeah. So first of all, what I'll say is that if you go to other parts of the world, like Europe or Asia, nobody really calls it transit-oriented development because it's just the way that the cities are built, right?

And in the United States even, North America, before cars became to dominate, you know, our cities, like you go to, you know, older cities like New York or Boston, Philadelphia, Toronto. I haven't been in Toronto, but I know you're from Toronto. And these communities were built, well, first of all, if you go back far enough, cities were built around walking cities. So medieval cities were walking cities. And basically, there's been research that shows that the distance across cities is generally about a one hour based on the mode of transportation that you're using. So if you walk from one side of the city to the other, you may have a city that's maybe three to five miles. And the medieval cities are pretty compact, anywhere from two to maybe five or six miles in diameter.

The transit city expanded that. There were train stations put in the middle and rail lines built out into the forests and farmlands at the time. And they would build little walking cities around the train stations. And so now you could move out of the city center, live in a suburb that was itself a little mini walking city. You know, think of Scandinavia. Think of, you know, lots of European cities, Japan, you know, all these different places around New York. and, you know, your distance from one side to the other side would be about an hour of travel time. Well, then, after World War II, we really started building a massive amount of highways and we abandoned walking in transit in favor of building automobile-oriented cities and suburbs and we created Sprawl.

And sprawl has had a lot of unintended consequences, one being that, you know, once you get a lot of people living in the suburbs, your traffic builds up and that creates a lot of congestion that's hard to solve, right? So think of cities like Atlanta, Miami, Houston. You know, once you get so much traffic, it's hard to build your way out of traffic, right? You could build another freeway or you can widen the freeway, but that actually leads to more traffic. Studies show it's a concept called induced demand, right? You build more supply. It actually encourages people to build more low-density suburbs, and that actually creates more traffic. It takes 10 or 15 years, but at the end of that cycle, you actually get more traffic than before you actually widened your road or your highway.

So going back in history, in the late 1800s, there was this concept called Garden Cities. It started out of England, and Ebenezer Howard built, like, you know, these two suburbs, Welwyn and Letchworth, and they connected into London, and they were what we now call trans-oriented developments. That kind of morphed into what was called the Newtown Movement that happened even before World War II and then after World War II. But it was predominantly in Scandinavia, Europe, Japan, other cities like Korea. They kind of adopted some of these Newtown movements. Again, it was focusing on building these transit-oriented communities around train stations and connecting them to the center city. We were very late in the United States to pick up on this concept, but it started like in the 1980s with a movement called the New Urbanist Movement. And that was really put together by a group of like architects, developers and environmentalists that said, you know, we just need to do differently.

We can't build our way out of suburban sprawl. And so the New Urbanist Movement, one of the members of the New Urbanist Movement, was an architect and urban planner out of California named Peter Calthorpe. He wrote a book in the early 90s called The Next American Metropolis. And in that book, he defined this term transit-oriented development. Again, a new term for an old concept. I was a student in the 1990s studying urban planning and was assigned his book in like 1995 or 1996. so it was relatively fresh off the presses and worked on some urban planning studios on this trans-oriented development concept and I kind of fell in love with the idea. It made a lot of sense. I had grown up in a suburb of New York City. I was very familiar with taking Metro North to and from New York City but where I lived, it was so car dependent. As a teenager, it was so isolated. I couldn't get anywhere without being able to be driven by a parent And then the day I turned 16, I signed up for my driving test and license.

And that was like freedom. Right. But in studying it, you know, as a college student, I began to realize how how much sprawl and car dependence. And I got the great benefit of working with two of the leading experts in the world from Australia, Peter Newman and Jeff Kenworthy. They wrote a book in the 1990s called Overcoming Automobile Dependence. And then they launched their second major book in 1999 at the White House with Al Gore called Sustainability in Cities. I was super lucky that they came and were visiting professors when I was a student at the University of Colorado. And they invited me to come to Australia and work with them. And I took them up on their offer and I went there and I worked with them. They were they were beaten up by academics in the 90s. People, I mean, there's published journal articles that basically call them terrible names for introducing these concepts, saying that people don't want to live in compact cities.

But the reality is that at the time, it was not all that popular in the United States. The Great Recession hits 2007, okay? And coming out of the Great Recession, all of a sudden, the real estate industry in the United States, all of a sudden, coming out of that in the early teens and up to the present, decides we're not going to build suburban sprawl the way we used to. We're going to build new urbanism. Why do they do it? They want to build new urbanism, trans-oriented communities. Why do they do it? because it is the most profitable type of real estate development. So it was the profit motivation that has really now propelled TOD to become more the conventional way of real estate development than the exception. It used to be the exception. And the benefits are way better in terms of carbon emissions and economic development.

I mean, there's some major consequences, one of them being gentrification. They're so popular that they get so expensive, and there's a supply and demand mismatch. The demand to live in these places is so high. The supply in the United States is so limited, so the prices go through the roof, right? But that's a problem that we need to solve through creating more supply. And there was a major book called Abundance written by Thompson and Klein a year or two ago, and that really gets to the heart of kind of what I'm talking about. Yeah, that Abundance book is extremely popular, not just in your discipline, but I think in general. Thank you for that introduction to everything you do in transit-oriented development. That was a really, you know, you obviously know this space really well, your professor in it. So for you to talk for five minutes about the area, it's nice to have so many specific names and papers and books to be able to dig into, so I really appreciate that. Now, given that background, what kinds of things does AI, does having these magical abilities allow you to do in transit-oriented development or in urban planning in general?

Yeah. So the leader, the academic leader who was out of Berkeley, Robert Severo, who retired, I don't know, maybe about six or eight years ago now. He had worked on a lot of studies on transorient development and certainly by far, you know, the most popular researcher associated with the term transorient development. I admired him and got to know him, and he was gracious enough to – I was a PhD student at Rutgers at the time in the early 2000s, and he served as an advisor on my PhD committee as an external advisor. And he was doing a study at the time where he identified like about 104 trans-oriented developments around the United States that were as a part of a survey. These were mostly newer developments, not old historic communities. But I took that list and I used that list for my Ph.D. dissertation.

And I started looking at how do these neighborhoods compare to the metropolitan as a whole? Like what was different in terms of travel behavior, car ownership, incomes, race, ethnicity, housing tenure, which means, you know, percentage of people that own versus rent. And so I did this very extensive demographic and economic analysis of how these neighborhoods compared to basically the average. And it was very interesting. And then after that, when I was an assistant professor, I decided, you know, these neighborhoods, these 104 neighborhoods, they're not that unique, right? Because they look very similar to a lot of neighborhoods that I – my parents both grew up in New York City. I mean, they look very similar to a lot of neighborhoods that I saw in Brooklyn and Queens and, of course, most of Manhattan and, of course, a lot of other older cities. And so I said, you know, there's got to be a way to look at all train stations in the United States to define whether or not they have these built environment characteristics.

And so I decided that I would classify every train station area in the United States. And there's about 4,500, give or take, based on walkability. So if it's walkable, then it would meet that classification and whether it has enough density to support transit ridership. It so turns out that you need to have about 4,000 households within a half mile circle, right? That's about 500 acres. That's eight dwelling units per gross acre. So you need about 4,000 housing units to have the minimum density to support transit systems. And then walkability, other researchers more or less said that there's this, there's walk score, right, which has become a fairly popular measure of walkability.

It's a, you know, you can go on to walkscore.com and type in any address in the country, and it'll give you a score from zero to 100. And researchers are saying if you have above 70, it's basically a neighborhood that really supports walking. And transit-oriented development is kind of a misnomer. It really should have been called – well, Peter Calthorpe called it a pedestrian pocket before he called it a transit-oriented development. But that term, pedestrian pocket, just didn't catch on. So he called it a transit-oriented development. That name caught on. But in reality, it's the pedestrian and the walkability, the nature of that neighborhood being walkable, which is actually way more important than transit. And so using that classification of 70 or greater walk score and more than 4,000 housing units, I classified every train station in the United States, and I analyzed them similar to what I did for the 104 stations and was able to see major, major changes.

So places – And to be clear, so doing 100 or so of them, you were able to do that manually. But then once you wanted to do it across the entire United States, that required automation. There's no way for an individual to do that, right? Correct. And at that time, I was using Excel and SPSS to do it. And so what I did at the time when I went from 100 to 4,500 is I was using SPSS and Excel, And I was coming up with lots and lots of data that shows that these, you know, roughly, let's call it a third of the train stations in the United States are TODs. They are way more sustainable than the non-TOD stations and then orders of magnitude more sustainable than the average or other neighborhoods that are mostly auto-dependent. So, for example, in a TOD, it's not uncommon to have 60 to 70 percent of your trips be walking, cycling and transit trips, you know, and maybe 30 to 40 percent of your trips being driving trips.

It also turns out that because a lot of these TODs are focused close to downtown locations, you have a lot of jobs in these locations. So folks like Christopher Lineberger, who is a major researcher and expert in my field, did a lot of research showing that a ginormous disproportionate of GDP is generated and created in what he calls walkable urban places. And walkable urban places is basically like another name for trans-oriented development. So you have just a tremendous amount of economic activity. You have way more sustainable patterns. And if you take that and theoretically imagine building future cities like that, you get to the point where you can make a huge dent in reducing carbon emissions with regards to transportation, which are some of the hardest carbon emission reductions to achieve.

Right. And now that you have AI tools, I understand that you've scaled this up even more, right? Yeah. So I started to collect a lot of data on these 4,500 train stations. I started to see other databases. Like I got a database from the National Park Service has data on all historic properties. I started analyzing it from a historic preservation perspective. I started to get data on property values. So I merged that data into it. I started to get data on a company called Spatial creates really cool data sets that geocode social media data And their main clients are like retail businesses. So if you want to open a restaurant or if you want to open up some sort of a boutique store or something, their data uses social media data to geocode it and to determine like what are the best census blocks,

which is a fairly fine-grained unit of analysis, to be able to locate that store. And I saw tremendous connections between their data and my data. And I got to the point where it was data overload. Like, I couldn't handle that data using Excel and SPSS anymore, and that's where I needed to reach for other tools. And it was, at that point, it was, my choices were R and Python, and I decided to go down the route of Python. Nice. And so to go into like a specific AI project that would also probably have been impossible to do manually or extremely expensive to hire people to kind of annotate in this way, you and a collaborator at the University of Florida recently did a project where you use deep learning on aerial imagery, on Google Street View data, on data of parcels of land, on open street map data, transit schedule data. and you use that to measure the built environment around transit stations in five dimensions, density, diversity, design, destinations, distance to transit.

That's the kind of thing that, you know, we talked about the genie being out of the bottle. That's probably an example of a project that if you wanted to do that without a deep learning model to do the machine vision or to analyze the tabular data that you have for that project, you'd probably be talking about a grant that would be like millions of dollars to hire people to be able to annotate it by hand. And then now you can analyze it probably near instantaneously for almost nothing. Yeah. So actually, I have to say my collaborator at the University of Florida is Professor Shirley Zhao. And her husband, Jacob Yan, are incredible. And Shirley was giving a presentation on some of the research that she was doing. I don't know, maybe six years ago, seven years ago. And watching her present was the spark that I needed to be able to say, I need to learn data science. She at the time was an assistant professor.

She did her Ph.D., I believe, at Johns Hopkins, was doing incredible work at the time. Again, I called it, you know, that time was more data science before we really called it AI. and that she inspired me to really go back and really study computer science because I wanted to do what she was doing. She was an inspiration to me, even though maybe, you know, I was a senior professor and I might have been an inspiration to her, but she was certainly an inspiration to me. And so I started collaborating with her and we started working and I'm continuing to do this work on using Google Street View images and lots of other data. So, for example, just yesterday, I was analyzing for my master's thesis, I was using Tab PFN, which was another model I learned from one of your episodes, and comparing it to XGBoost and comparing it to OLS. And using, I created a way to take a Street View image and to rate it.

And so imagine looking down Main Street and that Main Street's beautiful. Brick sidewalks, you know, flowers hanging from the light poles, beautiful benches, outdoor cafes, very vibrant, you know, versus let's call some arterial where, you know, you don't you barely have sidewalks. You got tons of cars, wide streets. You know, you feel like you're taking your life in your hands just across the street. So I was able, you know, working with Shilay and building on that and using guidance from her team and others, we now can basically use AI to train models and to rate those two images just as we would intuitively say, yes, that image is better than that image and it'll rate those images on lots of different dimensions. And now for my master's thesis, I'm using retail leases to show that retail stores are more valuable in higher rated places than low rated places.

And from an urban planning perspective, this is interesting because imagine you are a local elected official, you're a mayor of a community, and you're working with your urban planners. And they say, well, you know, we want to invest five million dollars to improve Main Street because we are just we're losing ground to our neighboring communities because they have a much better Main Street. And, you know, one of the questions might be, well, what's the cost benefit analysis of that? The reality is that we have a hard time measuring the cost-benefit analysis, but now with these AI tools, we can much better connect the cost-benefit analysis of making that investment to create a much more attractive place. Because previously, we thought these were very subjective things, but in reality, with AI, they're becoming a lot more measurable. Brilliant. And so now to segue from your research to your teaching. So you and I in this current semester at Tulane University, we are co-instructing, although you do the lion's share of the work for sure in this course.

and it was your vision to create a course like this. So I'm honored to be kind of along for the ride and able to contribute where I can. But you devised this brand new type of course. As far as you're aware, I think it's the first AI course in any undergraduate real estate or urban planning program. And you're using Vibe Coding basically. These students, they are not computer science students. They are not math students. They are not data science students. They are urban planning students. And so there's no requirement for them to be, they don't have to know SQL. They don't have to know Python. They don't have to know command line interfaces. But you nevertheless were able to come up with a curriculum that allows these students to be taking the genie out of the bottle and pushing the boundaries of what they thought was possible.

Yeah, correct. So about two years ago, I transitioned from spending 20 years as an urban planning professor to going into a real estate program and running the real estate program. And I've been a real estate developer for over 20 years now. But there are urban planning programs out there that have integrated AI and data science into their curriculum. But real estate has been slower. And the difference is our students in real estate at Tulane, they are often going into making the decisions about what to build. So in the United States, unfortunately, urban planners don't do a lot of planning. They do reacting, right? This is a debate that goes back to a couple of journal articles from the 90s by Reed Ewing, who's another one of my mentors. And he says that in the United States, we do reacting, not planning.

And what that means is that real estate developers are the ones that are putting the plans together. They're going to a city for approval and the planners react to those plans and they say, yes, it works according to the code or no, it doesn't. And then if it doesn't work according to the code, it's really up to the real estate developers to go back and figure out how to either change the zoning or to amend their plans to make their projects work. And so being in real estate, I really love because, you know, we are training in our class the students that are going to go on to really be the city builders of the future. And so I really wanted to bring my AI expertise to real estate. there were a couple of, you know, obviously some articles, but not really hardly any books on the topic, nothing that I thought was decent. And so I said, gosh, I want to teach one of the first AI

and real estate classes. And you fortunately came and gave the talk at Tulane last spring. And I said, you know, gosh, I would love to bring one of the leading experts in the world on AI to co-teach the class and fortunately you said yes and here we are and I think the class has been so exciting because one or two of the students of like the 15 or 20 students in the class has had any exposure to coding so basically not hardly any of the students have any exposure to coding we have um several Tulane football players um that are and and not to be disparaging against football players Right. They're actually some of the one or two of them are some of the smartest kids in the class. And we I blew myself away because using some of the content from Ed Donner's Udemy class that you recommended, we've had our students creating a video game the very first time they ever opened up a vibe coding,

a vibe coding application, a cursor in that case. And it was incredible. I mean, it blew me away. It blew the students away as to what it could do. And so we've been building upon that, you know, building a Kanban board, building personal websites. And now we're going to be embarking on a final project to create this really cool project with the students to build a tool for the public in New Orleans, hopefully, to better understand what their real estate assessments are. And this has come from my colleague, Cassius Peeler, who ran for assessor of New Orleans unsuccessfully. But he is very interested in getting the public to better understand their assessments. And so we're going to try to create a chat bot that is based in real data. And that is our semester project between now and the end of the semester. And it's a little scary. I think with AI and vibe coding, things are a little scary at first.

You're like, I don't know if I'm going to achieve what I want to achieve, but somehow it works out, and I'm excited for this journey. Yeah, to dig into that project a little bit more because I think it's pretty exciting. So assessments are – who carries out assessments? The government carries out assessments of your property. It gives you a sense of how much tax you have to pay. That's what those are? Yeah, so every property owner gets an assessment of how much their property is worth. And based on that assessment, it determines how much property tax you pay. And so if you have a house and the government says that your house is worth $800,000 and you think your house is only worth $700,000, right? Now, all of a sudden, you're going to be paying higher property taxes. And so there's the ability to challenge that value. These values, depending upon the community, they could change every year or they may change once every four or five years.

And so in New Orleans, it's I believe once every four years. So when these new assessments come out, the individual properties can vary wildly. And so you may be going from paying $3,000 a year in property taxes to all of a sudden now you've got to pay $6,000 a year. Your property tax is double. And so it's like, well, what can you do about that? Well, there's something you could do about it, right? The first thing is get educated about what the process is and how it works. So that's the first thing we're going to have our students do is to build a tool that educates the public as to what is an assessment, how does it work, how is it calculated, who do you go to and ask questions if you're not happy with your assessment. The next question I think that we're going to ask them is, well, how does it relate to my assessment? And the assessment is created by the assessor based on some complicated models that they use. But then you may be going on to Zillow or Realtor when you get your assessment and comparing your property to other similar properties nearby.

and that may or may not actually relate to what the assessor uses to calculate the value. So the third step, if we can get there, is to figure out how can we build in comparable data using databases like Zillow and other major databases like that to be able to ground truth your assessment for the property owner's assessment to see whether or not it makes sense or not. I got you. I got you. And this matters because the value of properties changes pretty rapidly. A stat that I have here from my research is that in Orleans Parish, where New Orleans is, citywide values were assessed to be 23% higher in 2023 than in the previous assessment. So 165,000 owners received new valuations, and the city's saying, hey, your property is worth 23% more than it was last time we looked at it.

Guess what? This year you're going to have to pay 23% more tax, which could be a lot. Well, and that's correct. But one of the things that's interesting about that is it's typically not done on a uniform basis. And so that 23% isn't just applied uniformly to every property. Some properties increase significantly more than that average. And some properties don't increase at all. Right. And so if you're one of these people that all of a sudden you wake up and now your assessment is 50 or 60 percent higher. Right. And you're going to start asking questions. Right. And so that's really what this tool that we're going to try to develop with. And the cool thing about it is that Cassius, who has been, you know, he actually ran the real estate program at Tulane before I came in, and he still is a professor in our program. He, and he had a very, even though he didn't win the assessor's election, he had a very successful campaign.

And I certainly think he'll have a shot at winning in the future if he decides to run again. But he raised a lot of public awareness about these issues. And, for example, like renters. Renters pay property taxes, but they don't realize it because they pay rent. And when that property tax goes up on their rental, it gets passed to the renter in the form of higher rents. So he's been doing a lot of educating on this topic. And he asked us if we could build a tool that could really help raise awareness to these issues so it can be a fair process. No one's saying that, you know, nobody likes property taxes going up. Of course, sometimes when your property value goes up, you know, you do need to pay more property taxes because the city has services and they're faced with inflation just like everybody else. So they need more money to be able to repair the roads and do all the things that they do.

But nobody likes, I don't think I've met anybody or very few people that love it when their property taxes go up. Right, exactly. So that's a relatively narrow example. And I would say all of the examples so far that we've discussed in this episode, The past 20 minutes or so, we've been talking about ways that AI could be used in real estate, in urban planning, and they've been relatively narrow in terms of, okay, I can assess the value of my property more accurately and maybe save some money on my taxes. John, what excites you really big picture in this space? When you talk about things like being able to tackle climate change and AI playing a role in that. What are some of the big ideas that excite you about how AI could accelerate human development and improve the world? Yeah, so I'll give you two examples, very different examples. One is this global TOD database that I've been working on.

And the other one is kind of an existential crisis facing education right now, which is students using AI to write their papers and do all their work and figuring out a way to be able to improve the educational system through expanding the use of oral exams. So to get into the first one, I did a sabbatical three years ago at Oxford University where I know you had been. And I love Oxford. I did my first sabbatical there in 2013 and my second one in 2023. And when I was there, I had talked about my American database of 4,500 train stations. I started working with some colleagues in France and England. And I previously had did a postdoc in Australia. And so I've gotten to know folks from all different parts of the world, you know, in my in my field.

And I thought to myself, you know, what if I start going to my colleagues in different parts of the world and say, hey, what if you kind of do the similar analysis that I've done in the United States with this TOD database in the United States that I have? And what if we do it in France and England and in Australia and other countries? And so I thought to myself, I could kind of create a network, explain to colleagues about how I did the work that I was doing and get them to replicate that in their own countries. And I even started having conversations with folks who thought it was a great idea and they were interested. But coordinating that kind of research across the whole world is very challenging. So at best, I thought, well, maybe if I could just get five or six countries, that would be a huge improvement over just, you know, having data from one country. So that was three years ago.

Then all of a sudden, about a year ago, just over a year ago, this was even before I started using VibeCoding. I started using coding, using ChatGPT at the time to help me write code that I would use. And I said, gosh, can I extract data from a global database to be able to create, you know, so for example, there are global databases out there with lots of geospatial information. So one of them is OpenStreetMath. That's the biggest one. And that's the one I've been using mostly. But there are other ones out there that I will increasingly integrate into my new global TOD database. I was able to extract information on over 110,000 train stations across the entire world. And now I have this global TOD database.

I've been working on it with my colleague, Billy Fields. We're writing a book on it called Green TOD. Should be coming out hopefully next year with Cambridge University Press. and it is mind-blowing the kind of work that we are doing and what we're able to achieve that would have never been fathomable, like I said just a few years ago Yeah you doing I think you expressed to me that you can be doing work for that book like international TOD analyses that would have taken years and like postdocs, hundreds of thousands of dollars to be able to do, and you get it done in hours. Billy and I were working a few weeks ago, and he said to me, and I fully agree with him, he said, You know, the work that we just did in the last two hours probably would have taken a $500,000 grant and at least two or three years.

And we would have had to collaborate with probably half a dozen people, you know, experts in different fields, postdocs, PhD students. and literally the work that we did just in a couple of hours was equivalent to what it would have taken about a half a million dollar grant to do just a couple of years ago. It's pretty wild. I wonder if that will shift researchers from kind of mentoring their postdocs, their PhD students, to just sitting alone at a Claude Code session. Well, you know, I do have a fear about that And I am, you know, looking into setting up a new doctoral program in the Tulane School of Architecture and Built Environment where I run the real estate program. We also have a sustainable urbanism program and architecture and historic preservation. So there's a lot of interdisciplinary thinking, and we're excited to think about creating a new Ph.D. program. I think on the one hand, you can absolutely make the argument that I do not need to work with other folks and I can sit and chat with a bot.

However, I don't think that that will happen because. I do think that I went into this field as a professor because I literally enjoy working with students, mentoring students. Right. We all want to leave a legacy. We all want to help folks that are earlier in their careers or, you know, get them going. Now, you know, do I need to go get, you know, a half a million dollar grant to do this task? Well, probably not. But if I were able to get a half a million dollar grant, think about what I can actually achieve now. Right. It's like building a, you know, a three eighty six computer, you know, back in the 1990s, you know, versus now, you know, having supercomputing capabilities. And so I have so many ideas about where we can take this to. And I did just apply for a major grant with colleagues at TU Delft, Oxford University and Cranfield University in the UK.

to do this. And that study would basically take my global TOD database and do a lot of energy modeling around transportation, land use, and building systems to figure out how we can create carbon-free or significantly reduced carbon communities to achieve a future goal to where we can bring down global warming. And so I think we can, I don't think we're at the point where we could just say, yeah, we could just sit around and do it ourselves because, you know, the challenges that we're facing externally out there in the economy and Mother Nature necessitates us to have the supercomputing because we don't have a lot of time if we want to, you know, figure out how to bring down global warming. We've already surpassed a lot of the temperature targets that we said we shouldn't surpass. Right. Yes. So AI helping with climate change in the ways you've detailed in this episode, better planning.

And the second area that you were talking about is education. And this is an exciting one, a very interesting one. You know, people talk a lot. In fact, the day of us recording, I published a podcast episode, a Friday episode. I'm just quickly looking at the number here. Yes, the day that you and I are recording this episode, we published episode 1036 of this podcast. And the whole episode is just me, after having done research and scripting an episode, about how AI is impacting how kids learn. Now, most of the focus for me was around kind of elementary and secondary school education. But the same kinds of problems faced tertiary education in universities where historically you used to be able to assign a research paper. Somebody would write a paper and I guess they paid another human to do the work for them. They would have to do it themselves.

And now, you know, and now, you know, one of the things that I said in that, Episode 1036 was that teachers today report getting like the same paper from many of their students back because that's like the most generic answer the chat UPC would turn out on that particular topic. And so you have partly leveraging AI yourself, co-founded a business called oral exam dot AI that is an AI driven solution to this AI created problem. Yeah, so let me just kind of go back to how this all started. So about a year and a half ago, we have what's called student learning outcomes and assessments, okay? And so we have to, and as a program director, I have to look at all of our courses and our different degree programs. You know, we have a bachelor's program, a minor. We have a master's program. We have a grad certificate. So for each one of those, you have to, in order to be accredited university, right, you have to understand, like, what do you want your students to learn and how do you measure if they are achieving those goals?

And every university has to do this, right? And so I've always struggled with this because I was always thinking to myself, gosh, we have what's called learning management software. We use one called Canvas. And so, you know, all of your Google Classroom is one that's typically used at, you know, the K through 12 that my kids use in their schools. And so, you know, the teachers increasingly are posting all their notes, posting, you know, their exams, posting their lectures in these LMS systems. And so I, you know, rather than trying to figure out what the student learning outcomes were based on just looking at the grades, because you're not really supposed to use grades, right? Like you're supposed to really have a question on these learning outcomes. That's not just like, oh, yeah, the kid got an A on the final exam.

That doesn't really speak to what they've learned itself. So I wrote a program using Python, and I extracted the entire course output from Canvas into a text file. and I uploaded that text file into ChatGPT at the time. And I said, why don't you tell me what the student learning outcomes are? I mean, this has all the stuff in the course, including all the grades and everything. And I did this for every class in my program. And I emailed it to all the professors who teach in my program, thinking that they were going to say to me, John, this is not good. AI is crap, and you cannot use it to understand student learning outcomes. Well, that did not happen, right? I got back a bunch of, yeah, it looks pretty good to me.

Yeah, I don't see any issues. It was one or two, you know, comments, but they weren't all that significant. And the data that I got back from that was incredible. And so I started using AI to help me with assessing our student learning outcomes. Well, at the same time, my colleague, Daryl Glade, who teaches in my program, he's a professor of practice. He and I are working and, you know, we're talking about, of course, every professor is talking about what do we do about, you know, chat GPT and students using it to write their papers, you know. And of course, at that point, it's like, well, we ban, you know, we ban the use of it. We have them write in blue books. And so if you go back, you know, hundreds of years, the educational system was based on the Socratic method. And to this day, if you get a Ph.D. or if you go through like medical boards, right, you have to do an oral exam.

And the oral exam is a complement to the written exam, or in some cases, a multiple choice exam. But it is really to make sure that you know your stuff. Well, the reason that writing became a popular form of assessment and evaluation at universities is because you can't really conduct oral exams for more than a handful of students in a class. So writing became a way to have a class with 30, 40, 50 students in that class and still be able to teach what you want to teach and test the students knowledge. Well, now, all of a sudden, we can't trust writing anymore. And so we Daryl and I, we saw that there were some folks that were talking about scaling exams, oral exams using AI. And we thought that was a brilliant idea. And so we really jumped on it very quickly. And with connecting the oral exams with oral exam.ai to be able to connect that to student learning outcomes and assessment really, I think, allows me as a program director to have much greater confidence that we have a pathway into the future.

that can allow students to preserve the integrity of the learning process without outsourcing their learning to an AI model. Right. So the idea is students can submit an essay, as they normally would, so you can now assign an essay and feel confident that the student knows what they're talking about because when they submit it in real time, they get asked questions by an AI model, which is something that is very easy to imagine an LLM being great at. And then they have to answer on video what those questions are. And then you provide back to the instructor, the professor who signed the paper, some assessment of whether it seems like the student actually understood what was in the essay or not. Yeah, exactly. So what happens is that you can connect it to your LMS or you can basically just have the system send out an email to a student So they could upload their paper or the paper can be pre-uploaded that they wrote.

And then the student gives them a short, typically five to seven minute interview about what they wrote. So it's asking them about, you know, about the details of what they wrote. And it's using AI to help determine whether they authored it and whether they understand it. So it's some ways it's using AI to combat AI. but it's previously or I don't know if previously is still the current model I believe is that a lot of professors are using AI detectors and AI detectors are notoriously bad and AI is getting so much better at being able to evade the detectors and what happens is you have students who are accused of of cheating by using AI to write their papers and in one or two cases there's been lawsuits where, you know, universities have been sued because the students, you know, they say, I did not use AI to write this and I got a zero on my paper. And so that breaks a fundamental trust between a student and, you know, not only their professor, but the entire university. And it's a

terrible situation. And some research shows that these false positives are very significant. You You know, it might be up to four to five percent of, you know, of detectors are creating false positive. You multiply that times all the papers submitted across, you know, a major university with, you know, 30,000 students. Now, all of a sudden, you're creating a nightmare to figure out, you know, you're talking about hundreds and hundreds, if not maybe thousands of papers a year that you've got to adjudicate whether or not the student actually wrote it or not. It's a horrible situation that we call like an arms race. Yeah, I think your solution is more elegant than those kinds of detectors, which I don't think the detector system is the way to go. So, yeah, so oralexam.ai, great solution, John. And thank you for giving us so many interesting ideas for AI in education, urban development, climate change, real estate. Really interesting episode. Do you have a book recommendation for us?

As a regular listener, you probably expected I was going to ask you for one. I do have a book recommendation, although I would like to put in one plug. So I myself have launched a new podcast called Building Places, and it's on all the major platforms. And our website is buildingplaces.io. And I hope to have you on as a guest on my show, which is focusing on real estate. So it's basically focusing on a lot of Tulane alumni and students and faculty are mostly my guests. But it's been a lot of fun being a podcast host for the first time. Been enjoying it. But, yes, so my book recommendation is a book that I picked up just last week at the airport by Adam Grant called Hidden Potential, The Science of Achieving Greater Things. I think it's the perfect book in this era because what it basically says is you don't need to be naturally gifted at things in order to be great at them.

You just need to unlock the hidden potential in what you're good at. I have the book right here. I read the endorsements by Serena Williams, Mark Cuban, Malcolm Gladwell, and Yo-Yo Ma. And I said, okay, if these people say this is a good book, and I've been reading it as not disappointed. It reads very much like a Malcolm Gladwell type of prose. Excellent. Sounds like entertaining and educational at the same time. Now, John, thank you for plugging the Building Places podcast. But I always end, my final question is always, how should people be following you after this podcast? So, yeah, so the Building Places podcast is one thing. You can look up for me on an upcoming episode of yours as well. And I know I'm going to see you in class every Monday, every Wednesday for the rest of the term and hopefully be on this term as well. But for our listeners who enjoyed hearing from you today, what are the best ways for them to follow you? Probably at this point, the best and only way is on LinkedIn. So John Rennie, and I just have gotten so busy and have not really been able to keep up with or just really keep up much with the other social media platforms.

And so I pretty much predominantly stick to LinkedIn. Yeah, LinkedIn is where most of our guests are posting these days. John, thank you so much for taking the time with all the things you do, being a professor, co-creating an AI startup, up doing all of the real estate investment. We didn't even get into that, but you run a firm called Priority Funds that does real estate development and investing. That's a big part of your life. So amazing that you could take the time out and do this episode with us. Yeah. Also, again, grateful to be working with you. One more thing. One of the best ways, if you're interested in real estate, which I know is probably a very small percentage of your listeners, is we have a phenomenal grad program that is now been transitioned to a hybrid, mostly online program. So you can be a student anywhere in the United States, but we do meet in person several times a year. And so you get to meet myself, you get to meet my fellow faculty members and your fellow

students. so if you're interested in real estate development check us out Tulane Master of Sustainable Real Estate Development program and you could live anywhere really anywhere in the world and be a student in our program but our courses are synchronous so you pretty much do have to live in the United States unless you want to stay up really late at night and take our classes. Nice all right great tip there for folks who are interested in that graduate program. All right. Thanks again, John, for taking the time and I'll catch you in class. All right. Thanks, John. Have a great weekend. 10,000 train stations where two hours of work now matches what once took a half million dollar

grant in several years. He talked about how the undergraduate AI in real estate course we co-teach at Tulane together had students with no coding experience vibe coding a video game in their very first session and how his startup oral exam.ai gives students a five to seven minute AI led oral exam on the essay they just submitted a fairer alternative to AI detectors whose false positives can reach 4 or 5%. All right, I hope you enjoyed today's episode. As always, you can get all the show notes, including the transcript for the episode, the video recording, any materials mentioned on the show, the URLs for John's social media profiles, as well as my own at superdatascience.com slash 1039. That is today's episode number 1039. And if you'd like to engage with me in person as opposed to just through social media, I would love to meet you in real life this week at the Open Data Science Conference, ODSC AI West, running from October 27th to 29th in San Francisco.

I'll be delivering a keynote on how to thrive in the agentic AI era, amongst other fun duties like hosting fireside chats and emceeing. I hope to see you there. Thanks to everyone on the Super Data Science Podcast team, our podcast manager, Sonja Breivich, media editor, Mario Pombo, partnerships manager, Natalie Zajski, researcher, Serge Massis, and founder, Kirill Aramengo. Thanks to all of them for producing another super episode for us today for enabling that super team to create this free podcast for you. We are deeply grateful to our sponsors. You, yes, you can support this show by checking out our sponsors links, which you can find in the show notes. And if you'd ever like to sponsor an episode yourself, you can get the details on how by making your way to johnkrone.com slash podcast. Otherwise, please share this episode with other people that would be interested in AI applied to urban planning and real estate. Review this episode on your favorite podcasting platform. Subscribe if you're not already a subscriber. But most importantly, I hope you'll just keep on tuning in. I'm so grateful to have you listening and I hope I can continue to make

episodes you love for years and years to come. Until next time, keep on rocking it out there. I'm looking forward to another round of the Super Data Science Podcast with you very soon. Thank you.