Chuck Kasky, Maryland REALTORS® CEO, is joined by Michael Cunningham, Maryland REALTORS® Director Information Technology, to discuss the power and highly invested push of artificial intelligence.
Listen as Michael and Chuck delve into the evolution of technologies and demystify artificial intelligence, self-driving cars, and augmented reality. Also, learn about how REALTORS® and agents may be able to use these tools to improve their business' efficiency in the near future.
Chuck Kasky, Maryland REALTORS® CEO, is joined by Michael Cunningham, Maryland REALTORS® Director Information Technology, to discuss the power and highly invested push of artificial intelligence.
Listen as Michael and Chuck delve into the evolution of technologies and demystify artificial intelligence, self-driving cars, and augmented reality. Also, learn about how REALTORS® and agents may be able to use these tools to improve their business' efficiency in the near future.
Artificial intelligence or AI is the incredible force behind the greatest scientific advances of our time and our ability to use it in new and creative ways will shape the future of our society for generations to come. That intro was created using chat, g p t. So it's no surprise, it's biased <laugh> . It wants you to think the best of it. So there you go. Ai, of course, is a computer science technology that enables machines to learn from experience, adjust to new inputs, and perform, quote , unquote , human-like tasks. When we think of ai, it conjures up images of everything from automated customer service chat bots to robots taking over the planet . Generative AI is very much in the public domain, kind of in your face like chat, G p T, meaning it generates an output from your inputs. Plenty of analysts believe AI will remove significant numbers of jobs from the marketplace, disrupt how entire industries work, and in the end, rewrite the very fabric of reality. But today we're gonna talk about it in more practical terms. First, starting with definitions, because I think we get confused by what we even mean when we say ai. And eventually toward the end, we're gonna talk about the potential uses and hopefully benefits of using AI in the real estate industry. For example, AI can automate mundane tasks like data entry and lead generation, analyze data within listings and consumer trends to help streamline buying and selling homes. AI can provide virtual reality tours for properties, helping buyers get an idea of what the home looks like without having to physically visit it . Using AI and real estate will allow agents to save time on administrative tasks while focusing on customer service. Additionally, AI driven tools can improve data accuracy and make more informed decisions about pricing market conditions. Agents can actually predict or accurately predict future housing prices based on past market behavior by leveraging predictive analytics tools powered by AI technologies like machine learning or natural language processing, which you'll hear about, or N l P , this could help agents stay ahead of the competition in the real estate market. Hello, I'm Chuck Caskey , Maryland realtor, c e o , and you are listening to Get Real Estate, the Maryland Realtors Podcast. My guest today is Michael Cunningham, Maryland Realtors Director of Information Technology. Michael, welcome to the program.
Speaker 2:Hi, Chuck . Thank you for having me.
Speaker 1:So pick apart what I said <laugh>. We're , you know, people think of this as, you know, Terminator type, I mean the dystopian future or ex Mackinac , if , you know , if you saw that movie when you created these very, very like robots that were became sentient. And that's really something that many people think will never happen , but, you know, we can, we can talk about that as futurists, but when, when we say ai, am I wrong that it's people confuse different parts of it or different segments of ai and let's help to bring some clarity to what we mean when we when we talk about it?
Speaker 2:Yeah, I think there's been a ton of hype in the media over the last six Yeah . Six to 12 months about ai, and that's really not a coincidence. I mean, if you look at New York Times and the Wall Street Journal both published over 200 AI articles in the last, each of them been published over 200 articles in the last month on ai.
Speaker 1:Oh, wow. Um , I wonder it's unavoidable <laugh>.
Speaker 2:It , it is a ton of information coming out and it's really not an accident. You'd also, if you look at, I , I don't know if you've ever heard of Y Combinator?
Speaker 1:No.
Speaker 2:Okay. So Y Combinator is the incubator that got a ton of start startups started, like Airbnb went through them . They've probably had, probably have at least 15 of the Fortune 500 companies in there. They manage over 600 billion worth of investment money.
Speaker 1:Oh, wow. Okay.
Speaker 2:So they're, they're a major player. In their past funding round 35% of the startups that were pitching had AI in their product pitch. Surprisingly because about 18 months previously it was under 5% and 35% had crypto in their
Speaker 1:Subscription. What, what <laugh> What's that? We don't really talk about crypto anymore. <laugh> .
Speaker 2:Yes. Um, so there's a lot of pushing from the marketing side of Silicon Valley, and this is the new darling of the industry right now , which is why you're hitting in so much hype and you're seeing so many products just being pushed out there that slap AI on their feature set because it's the buzzword in the media right now. It's basically the forming of the next new bubble in the tech world. But behind that, there's some real technology and , uh, services like Chat G B T have kind of brought that technology to the forefront and they've got some major backers that are pushing that technology forward. So when most people talk about ai, they're thinking about the Check GPTs of the world, or on the other side of the spectrum, they're thinking about Arnold Schwarzenegger's Terminator.
Speaker 1:Exactly.
Speaker 2:Um , without much in between and really chat. G B T is only a borderline ai, it's more of a , a language model that happens to have a little bit of smarts behind it. Mm-hmm. <affirmative> . And I think that the technology that's common every come to everybody's attention in the last six months is the ability of the computer to be able to have a working conversation with a human being and for the computer to understand what the human actually is saying and the context of that speech. Not just doing text tope translation, but understand what the human is asking for and what the human is saying. Or take pre-done text and be able to understand what that text says and be able to put that into context.
Speaker 1:So you mean when you go to a website, you know what I was thinking the other day, it says, hi, this is the so-and-so chat bot , you know, ask your question. That's just looking for keywords, right? That's not necessarily really quote unquote intelligent, but you know what I thought about last night that I completely forgot about? Remember Clippy and all those things from Microsoft, the original mm-hmm . <affirmative> , weird little, I got that used to swim at the bottom of your page and it said, Hey, you have a problem, ask me a question. And you clicked on it and you said, where's this? And it told you where to find some feature of Word or Excel or whatever. That's kind of what we're talking about, right. All those, and that was decades ago .
Speaker 2:Right? So what I'd like to say, you know, say about chat G B T is it's kind of the fulfillment of the promise that a clippy or even like a Siri was, right?
Speaker 1:Oh , sure. Of course. Yeah.
Speaker 2:Um, clippy and, and Siri wasn't much of improvement over clippy,
Speaker 1:Right ? <laugh>
Speaker 2:That basically Siri was able to do voice to text translation, right. And it understood a certain number of nouns at a certain number of verbs,
Speaker 1:Right?
Speaker 2:So if I said, call Chuck Caskey right? Says, oh, call beans , telephone call, and Chuck Caskey is somebody that must be in the person's contacts. Exactly. And it can figure that out. But if I said, ring up, Chuck, Chuck, it would have a lot of issues with that. And Siri is sitting here talking to me on my phone right now.
Speaker 1:<laugh>
Speaker 2:The difference between Siri and chat , G b t is chat . G B T understands the breadth of human language and basically has, is a built in the SARS for all language. So it gets the gist of what you're trying to accomplish and can understand that goal that you have. And then how it fulfills it is up to the programmer still.
Speaker 1:Um , you just called me by the way.
Speaker 2:Oh , thank you .
Speaker 1:I'm kidding . That's ,
Speaker 2:Is
Speaker 1:That <laugh> , is that a function of, of the sheer volume of the data that it can manipulate and go through and is that based on the computing power or, you know, what, what are the actual advancements? The the , the algorithms are more complicated. I mean, what are, what are the, okay , so where did the advancements actually happen that brought us to, to this point
Speaker 2:A couple of different things. First of all, the, to develop these models that Chad g PT runs on uses about the electricity of a small size city Yeah . For a couple of hours. Yeah. It uses a , basically a large percentage of the computing power of an entire data center to generate this initial model. Right . Once the model's been created, it doesn't take much. A desktop computer could probably run chat. G P T A powerful desktop computer could probably run chat G P T for you once the model's been created. But the models have basically been trained on the entire internet. And, and what they are looking for is things that go together and finding patterns in how people speak and how ideas interact with each other. And that's its magic sauce is that it has this huge training platform that it's learned upon, and then it's figured out that a sentence will be finished in a certain way. And it seems magical to us as consumers because realistically our lives are fairly predictable if you divide them up into chunks. Mm-hmm . <affirmative> , you know, I had a pretty busy day today. I got up at six o'clock this morning, went paddle boarding, saw some wonderful wildlife, came home, worked for a few hours. You know, I'm , we're gonna have a dinner party tonight, but if you break up all those things, I went paddling. Millions of people go paddling every day . Mm-hmm . <affirmative> , billions of people go to work every day. You know, it's Friday millions if not billions of people are gonna have people over for dinner tonight . Our day is pretty common and so is our speech. Mm-hmm . <affirmative> , we typically will put certain thoughts after certain other thoughts and even the individual work that we do, if you think about the documents you generated today, the emails that you sent. Yeah . There's not that much unique in our daily business dealings from the world as a whole. So it's very easy to predict our behavior and what we're going to communicate next based upon previous experience. And that's all chat G p T is really doing. Yeah . Is making an educated guess based upon what you said in the previous sentence
Speaker 1:Is that machine learning
Speaker 2:The beginnings of it. Right.
Speaker 1:Exactly. What is , what is, yeah .
Speaker 2:People are debating whether that is machine learning, if you have a large enough data set , right . Maybe we really don't know how we as humans learn, you know, how different we're gonna
Speaker 1:Get to that <laugh>,
Speaker 2:You know, how different is it? You know, we all hear people that overuse stock phrases all the time in their conversation and or parrot or paraphrase , um, the news media . You know, how is chat G B T doing anything different than what we do in our daily speech? We're basically mimicking the speech of other people
Speaker 1:When they talk about, and they, whether it's real idea experts or just the, the media, which is where we are all mostly getting this from 'em . And we're not going to the Aspen Institute big Ideas conference, that was just this past week. But, so we rely on the media , um, and maybe some articles and then depending on how far these down, these rabbit holes you wanna go, maybe you can get into some of this very, very complicated stuff. But when people talk about machine learning, what are they really mean and how is that being incorporated into what we could think of as ai?
Speaker 2:I think that there's different schools of thought on that. I think if you're talking in the corporate world , um, they're mostly talking about data analysis. They're looking at how you could take a certain set of facts, correlate it to another set of facts, and have the machine be able to predict a future outcome based upon past data. I think that there's a lot of that in the actual area world. There's a lot of that in the financial investment world. Some of that right now in real estate speculation, especially in commercial acquisitions, that type of thing. Absolutely.
Speaker 1:Yeah.
Speaker 2:So I think on the corporate side, that's what they're looking for is a very narrowly defined AI as a data analysis tool. I think if you're talking about university research, yeah . I think that they are trying to explore general ai, which is a computer that thinks like us and can end up with its own priorities and its own wants and needs and goals, and then branch those needs and goals into other needs and goals to achieve a , an overarching objective, whatever that might be. Think that that's still a ways out
Speaker 1:A a ways Yeah. Far, I think. But yeah. So let's talk for a minute about what machine learning is. Is it predictive analytics or is that different?
Speaker 2:Um, predictive analytics is one of the, the strongest branches of machine learning. I mean, there's, there's a ton of machine learning as far as, you know, we could talk about it in automobiles, machine learning, in product automation, that type of thing. But as far as the real estate market is concerned, it's mostly right now gonna start revolving around analytics. Right. People trying to be able to do more sophisticated searches of existing data. Right. One of the areas that you'll have a little bit of deviation from that branch would be that machine learning allows us to delve into non-standard data sources. So for instance, you can point an AI at a block of text. Let's say that I was doing a house search and I said that I want a house in a community that has a good swim team. There is no way that a Zillow could determine whether a neighborhood has a good swim team. But if you had an AI that went, did a search for all the swim teams in the area, pulled a zip code, pulled their their records from some other data source, and was able to compile that data, you could theoretically have a house search that included a quarry for a swim team or one. If you say that I want a peer in my neighborhood that has good fishing,
Speaker 1:Right, <laugh> that
Speaker 2:Could potentially correlate waterfront neighborhoods and then high end fishing reports from another source, and be able to combine those two sources to give you some, you know, advanced or some expert market knowledge.
Speaker 1:I know for example, n A R was working on some predictive analytics. And one example of this who, if I'm, if I'm farming a neighborhood, just to bring it down to, to some practical, potential practical applications for our members. I live in the Mount Washington neighborhood of Baltimore City, and I am sure there is some kind of a platform that would allow me to say, identify the homes in this zip code 21 2 15 , for example, that are most likely to be for sale in the next three to five years. And assuming that data is out there, which is who am who is currently there, did we have kids? Are they gone? How old am I? How long have I been in the house? What's the average time people in this zip code have lived in their homes? And it's going to potentially spit out a list of PE of homes that are the most likely for me to contact, because potentially those are the ones, at least from the data that I have, that are most likely to be considering selling in the next three to five years. Is that fair?
Speaker 2:Yeah. I think that's doable with existing technology already. Yeah. Yeah. Uh , that's not really ai, that's just data analy . Good data analytics can probably establish that based upon existing databases. Now, if you,
Speaker 1:That's kind of where I was trying to get at the difference between those two because I think yes , in , in my own mind, and I just assuming in other people's minds that we, we don't, we need to be distinguish between that and AI <laugh> . So difference ,
Speaker 2:Well, no. Okay . So, so in your scenario there though, that would be doable. It would be, you would go to your IT person or your sta sta ian probably in AAR's case since they've got the , the staff there, and they would put together that query and put together those different data sources and two weeks later they would hand you back a report mm-hmm. <affirmative> based upon those data sources. What AI is gonna be able to do is move that from moving it over to that I d P department and waiting two weeks to giving you as an agent those tools to make that query in real time . Okay. And ask that question in real time . So really what it's doing is it's lowering the bar.
Speaker 1:Um, it's , it's accc , it's it's accessibility and applications and, and the ability of it
Speaker 2:Yes. It's, it's lowering the bar to the level that you can put that into a realtor's toolbox versus having to farm it out to a specialized employee or an external source. Um, and that AI would be smart enough to be able to interpret what you just said there, Chuck, and tie those to the various data points and the various fields in those data points to achieve that result, hopefully. And that's, and that's what they're looking to do. And that's what most of the products in real estate that are they're selling AI or attempting to do right now is take natural language requests and turn those into data.
Speaker 1:How smart are they? If I'm not good with the natural language query, it doesn't know what I meant to ask, it only meant knows what I actually did ask.
Speaker 2:And that is what's being hashed out with Exactly.
Speaker 1:Is ,
Speaker 2:You know, just like Google, you can find anything you want if you do a plus question mark, double quote word minus <laugh> , you know, there's put together all kinds of craziness in your query , right ? You can find you're good at Google, but you have to have the querying skills. This is gonna probably be a lower bar than that. Okay. But it's going to probably be a high functioning
Speaker 1:Mm-hmm . <affirmative>
Speaker 2:Realtor that would be doing this. Not someone that struggles to turn on their computer.
Speaker 1:That's fair at first.
Speaker 2:And I think the bar are lower as the tools get better mm-hmm. <affirmative> , but the initial, the first, the first generation of tools is always kind of hokey and it's always a little rough around the edges. And I'm sure all of these AI products are gonna be rough around the edges as well, and then they'll get more and more refined for the rest of us.
Speaker 1:So I , I don't wanna spend a whole lot of time on it, but, you know, autonomous vehicles, I mean, I had an argument about five years ago with a friend who said, oh, these self-driving cars we're all gonna have 'em in three or five years. And I said, no, we're not. No, because when you think of the sheer amount of data, our human brains sort through and the shortcuts we take through heuristics or biases or whatever, we, we process millions of pieces of information and we make split decisions about, you know , somebody comes in front of us, do I hit and crash my car, or do I run this person over? You know, do, do I avoid a collision and kill a human being? I mean, those are decisions that in I all , leaving aside the insurance costs about who's actually responsible if a self-driving car kills somebody and the courts are handling all that. But it's impossible for me to believe that we will be able to replicate the human mind <laugh> that way with the , the enormous complexity that we don't even understand that somehow we're gonna implant that into a machine. And, and I think this, I mean, I don't know how many more people, you know, self-driving cars have to kill before we come to the realization that they, they're not gonna perform like a human being. I don't know what, what are your thoughts on maybe , and then we'll get down to what actually might up coming in real estate. But this is just a question that fascinates me.
Speaker 2:Well, I think let's, let's have your theoretical automobile here that is better than at driving than a human being. Let's just , let's just call it that. Okay ? Yeah . Think they're better, better than a human being. Mm-hmm . <affirmative> , but there is a incident and the vehicle has to choose between Right. Crashing and killing all of the passengers or by some other action killing a busload of kids.
Speaker 1:Yeah , exactly. Yeah .
Speaker 2:And who is going to program that ethics chip
Speaker 1:Exactly.
Speaker 2:To determine that. And can I get an aftermarket, save my family instead of the bus of kids chip? Right . You know, the selfish, the selfish AI chip <laugh> , you know , will the president of the United States be get the same chip, or will his life be prioritized over someone else's? You know, how, how is that ethics going to work? I also think that right now we accept the fallibility of humans, but not necessarily the fallibility of machines. If you replace that AI with a little old lady that had no business driving on the road, why do we let little old ladies drive on the road? Why do we let teenagers drive on the road? You know? And we attribute that to the fallibility of human beings that no one was at fault, even though clearly was somebody who was in the wrong. Um, there's a lot of ethical issues to work out. I think personally, and this is, I don't think this is the way industry will go. I think personally that you should have all self-driving vehicle areas similar to Lane that I can throw my truck on self drive in the lane . All the other cars in that H I V lane are controlled by the same, by some overseeing computer, and they're all self-driving in that scenario that you're not gonna be facing pedestrians, you're not gonna be facing a little lady that decides to cross lanes doing 15 miles an hour or half plastic bags blowing across the street that confuse the ai. I think it could work. So you , you drive onto the highway, you get in the self drive lane, and you turn off, turn it off, and the vehicle drives for you. When you reach your exit, it goes beep, beep beep, and hands back control over to you and the exit ramp. Could that work? Yeah, I think that could work. Yeah . Do I trust it in downtown Baltimore with all the
Speaker 1:Possible scenarios
Speaker 2:And work zones and everything else? No, not really. Um, unless it gets a lot smarter than it is.
Speaker 1:So what other things can, can we see switching gears in, in real estate? So , uh, for listeners who, who are agents and brokers and they wanna kind of be prepared for how they can incorporate this stuff into their business models. First of all, you know, one of the things is jobs, and you talk about self-driving trucks and , and so, you know, all these truck drivers, are they gonna be out of business, et cetera , et cetera . And so we're that's a, that's a big part of it. And even in my intro, I, I talked about some efficiencies and, you know, I hate using that word because every time we talk about efficiencies, we usually talk about cutting jobs. So I'm not suggesting that, but what kinds of applications can we see for real estate? You mentioned a couple, but what other possibilities or , or even being used today or being considered or being rolled out?
Speaker 2:Well, I think first of all, I mean the search process, which we've already discussed a little bit, is going to probably , uh, be improved. So your Zillows and your different property search tools are gonna get better and better, which is both going to help the agents and possibly hurt them by, you know, questioning the value proposition with better tools. The consumer's going to be asking the realtor, well, what are you offering me now? So I think on the realtor side of things, listings will become much more important. What you say in your listings will become much more important. The descriptions of your listings are now going to be search criteria instead of just once they've already looked at the pretty picture, they see your listing, now your listing's gonna be picked up by the AI and the AI is gonna be reading it. Or the fact that you had water access and blue ribbon schools and all of those things that aren't really data points aren't fields that you fill out, but those big description tags are now gonna have a lot more relevance when you're talking , when people are searching, all of that stuff's gonna be picked up on more and writing something that really appeals to that customer is gonna get those properties sold faster. I think that there's gonna be some tools for you guys to generate your listings easier and faster. There's gonna be basically intelligent boiler plating for certain neighborhoods. You know, being able to say, this house is in Federal Hill, and having it be able to populate a bunch of information about Federal Hill into those descriptions and maybe put links to external resources that the ais will pick up on for additional information about Federal Hill, like restaurant listings or whatever. It's gonna be a , a lot more data rich environment that they're gonna be able to take. The consumers gonna be able to take a much deeper dive into the data of this property and get a much better sense of what it would be like to live there. Is
Speaker 1:It like this back and forth? I mean, if we know what consumers are searching for , right? So it's gonna know that by, by what are what's in searches and then is it gonna inform the listing so that it's gonna match what we know people are searching for? Is that what you're saying?
Speaker 2:To some degree, you know, it always is gonna follow , you know , that's always gonna follow that. Everybody's going to cater their listing to what's in demand, but I think that it'll be much easier to hit niche markets. Okay . That , that someone, for instance, I live in Anne Arundel County and Waterfront is mm-hmm. <affirmative> is the selling point. Everybody's looking for waterfront water access and water amenities. Right. You know, and being able to search that there are slips available at , you know, both slips available within the community is a big deal. That's not something that's in traditional listing data, but if you're in this market and looking for, for people in this Anne Arundel County market that are looking to be on the water, that's a piece of metadata that you can include and provide in additional information on that will perhaps influence somebody more than a particular neighborhood or a school district. You know, was was , that's traditionally been what people are after is a school district or a specific neighborhood or a specific proximity to aspecific resource. Now you can use all resources, the proximity to any resource, maybe it's even customer defined resources, you know, how close is it to my mother-in-law sell , you know, I'm once saying a property that's within 20 miles of my mother-in-law's house and within 40 miles of my workplace, that would be possible to do with an AI search. And that has, right
Speaker 1:Now, that would be a couple of searches,
Speaker 2:<laugh>. Right? And , and it would be a lot of complicated searching that most customers wouldn't be able to do now it's that you just type in another sentence, another qualifier that you want to find that property. I'm not sure how it'll affect the agent in that regard. I think that an agent could use that very well if prepared for it or it could hurt them if they're not. It's going to be what do the agents' tools of the future look like and how are they gonna leverage this to get those listings? I think that some agents are gonna have to find a larger geographical area to work within, because people are going, without people being quite as constrained to a centralized office now with a lot of the work from home things, people are gonna be wandering further afar and these tools are gonna let them wander a little bit further afar by finding the amenities they want. It'll be interesting to see
Speaker 1:As far as efficiencies, what kind of tasks that, that we do manually, digital assistance, are they gonna get better? Also are those are two separate questions, but yeah . Are you looking to see a real explosion in the capability? I know they're talking about language, you know, that, that I can record a certain number of words and then it will duplicate my voice and it will actually sound like you're talking to me when you're not. I mean, is that coming where it says, I , you know, I had this as Chuck and they say, Chuck, you know, I would interested about this listing in on so and so and I can say, oh yeah, that's a great house. Uh , when do you wanna see it and then go from there? Is that, is that something that
Speaker 2:I think that there's gonna be , uh, I don't know if you've ever heard of the Uncanny Valley effect.
Speaker 1:Say that again, I'm sorry. No ,
Speaker 2:Uh, there's, there's the concept of the Uncanny Valley , um Okay . Which is largely used with movies like Avatar when they first came out, Uhhuh <affirmative> , where things look almost human, but not quite.
Speaker 1:Oh, oh yeah. Okay, okay.
Speaker 2:Okay. And that's going to be be the case I think with AI a bit. Yeah, yeah.
Speaker 1:That
Speaker 2:You're going , they're gonna be close, but not quite
Speaker 1:<laugh>. Yeah, I've heard some, I've heard some of those vo vocal synthesis and you're right, they're close.
Speaker 2:So as far as the automation tools that you were talking about, Chuck, right? Mm-hmm. <affirmative>, I think that forms is gonna get largely automated.
Speaker 1:Yeah. Right.
Speaker 2:You'll have the ability to fill out the basics of forms and have them be able to self correct based upon , um, your feedback so that you'll be able to say, Hey, change that form to $650,000 instead of $649,000 and make those type of changes, which will help realtors in a lot of ways dealing with a mobile environment that you can change forms via your smartphone or something like that. A lot of boilerplate stuff will be able to be auto created , like listing information, listing details, all of that could be easily automated and it can look at your existing samples. So it's not writing stuff from scratch, it's writing stuff from your previous examples mm-hmm. <affirmative>. So you can say, write this in my style and it will look at your other listings and write in a similar style. So it's not just Oh , right. Generated. It's generated as if Chuck were writing it.
Speaker 1:That's important point. Yeah.
Speaker 2:So that it really does have that, that feel of your stuff. I think there'll be some automated marketing and outreach tools using ai. Um, they'll probably be bad at first <laugh> , but <laugh> , um, they'll improve really. It's early days still, I think.
Speaker 1:Well , yeah. And that's I guess the point. Yeah. What about virtual reality? Add that , talk about that a little bit after you're finished that thought .
Speaker 2:I think , yeah. I think augmented reality is more , um, yeah . What you're talking about. Uh ,
Speaker 1:Yeah, I am actually, yeah.
Speaker 2:And that is coming along that will Yeah . That will be here within the next three, three years. It's already a little bit there. You can already snap together a bunch of pictures mm-hmm. <affirmative> and fill in the blanks and make a little bit of a tour. Those will get better. Yeah. I think that's gonna be a big hit for remote sales. Sure. Sure . People coming in from out of area and even in person maybe at first viewing, you know, if you could, if you can get enough market penetration into the market that there's enough listings that actually have them. I mean, we've had virtual tours for years.
Speaker 1:Sure. Exactly. Yeah. Um ,
Speaker 2:25 years ago when I had my own company, I was sending , taking those 360 bubble cameras out and filming , filming properties and doing tours of , um, new construction. They've improved a lot, but it's just a re-imagining of the old in that case. But the tools are getting there, you know, all of the furniture stores now have the imagine this couch in your space stuff.
Speaker 1:Exactly what I was gonna say. So take the furniture store stuff, and then you plug it into the home tour and you put a VR headset on and you say, drop the couch that I already have, that I'm gonna move and put that in this new living room and then see it in 3d. How cool would that be? Yeah .
Speaker 2:I , I think it , I think that's doable with today's technology. It's just a matter of getting a platform that people are gonna agree on. Right . Um , that's always the issue is getting enough market penetration that the product that you own is the product. You're not picking up a beta over , uh, vhs.
Speaker 1:Well, you, that's a perfect segue to my next question. And do you see this, which is something that I saw come out of the Aspen , uh, IDEA conference that this , and it already is too late, so we know for a fact, so I shouldn't even ask it as a question, but we, this, this follows the same pattern that we had from the auto automobile to every other major technology, including as you mentioned. Well, that was a market driven thing, but, but I'm specifically talking about the advancement of technology that is always leading the way, and then when, and if government regulation has to come into play, it is just 100% of the time going to be playing catch up with the technology. And I think you see that in social media and other areas more, more recently, but this is already following that pattern, isn't it?
Speaker 2:I think that it is. I'm not sure how you control it though.
Speaker 1:<laugh> .
Speaker 2:The bottom line is that they've now released the equivalent of the previous version of chat, G P t,
Speaker 1:Uhhuh
Speaker 2:<affirmative> that will run on a laptop.
Speaker 1:Oh, wow. I didn't know that.
Speaker 2:All of chat G P T can be stored on a standard hard drive of your lap , that laptop that you've got right there. Right. All the knowledge that's contained within chat g PT could be contained on that laptop. It would be slow.
Speaker 1:Yeah .
Speaker 2:But it would run.
Speaker 1:Hmm .
Speaker 2:So you need a massive video card , actually, I should say a desktop computer, not your laptop. You need a desktop computer. Massive , a massive video card with lots of memory, but it would run.
Speaker 1:Wow.
Speaker 2:So how do you contain that, that ability with the bar being set so low mm-hmm. <affirmative> ? Um, I don't think that you can, I think that there are also a lot of market forces aligned Yeah . To making this happen. Right now you have your open AI for sure , back for sure backed by Microsoft. You've got , um, Bard backed by Google, and you've got, I forget what Facebook's is called, but they've got one too now. Yeah . And they're major players that are behind this. And how do you tell,
Speaker 1:Right?
Speaker 2:I don't know that you can tell computer generated or you can right now a little bit, but another generation will you be able to tell that something's been computer generated or not, you know? Right. I don't know. I don't know what the, the future will hold and I don't know how they would regulate it. They can declare it. You can't do it <laugh> , but, you know, I'm not sure that they have the ability to enforce th those decrees. Yeah.
Speaker 1:They're still struggling with crypto, speaking of crypto. We still don't have, they don't still can't get their arms around that.
Speaker 2:Yeah. So I mean, yes, they're going to try to regulate it in some, some way. I don't think that these language models are really a threat to anybody other than writers. I mean, <laugh> , <laugh> , you know, they're not, they're not going to threaten anyone's sacred cows. Right, right . You know it enough to get the legislation going. They're gonna get everybody's interest up. Now, if you start getting closer to general ai, I think you're gonna get a lot of people involved. If you start to get into automobiles, I think you're gonna get a lot more people involved. I think what this will do though is start to make people silo data. Oh , I think that chat g BT was trading on the internet, you know, some of these image creation ais Right . Track on commercial photo libraries that , that all have to them . I
Speaker 1:Read that. Yes.
Speaker 2:And who owns that data?
Speaker 1:You know, so you , you do have some copyright issues there, especially on that photography stuff that I did read an article where there was something a that generated that was potentially violating somebody else's copyright. So there is some room there. Yeah, I see.
Speaker 2:Yeah. So I think, well, plus you're going to have all of these databases that are out there, you know, you know, are you gonna let , uh, chat G b t read LexiNexis?
Speaker 1:You mean it hasn't already?
Speaker 2:I don't know. I don't know if it's in there or not, but , but , but I'm just saying, you know, will you Yep . We'll , Westlaw open up their database to, to them and allow them access to that, or all of these different companies that hold all of this demographic data, gonna let them
Speaker 1:Demo . Yeah .
Speaker 2:You know, because it's gonna demonetize a lot of these valuable properties that they have.
Speaker 1:I mean, state laws are you talking about LexisNexis, what they do is just repackage a lot of stuff that's in the public domain anyway. I mean, they don't, they don't own Maryland law. They publish it or at , or Yeah, actually it's them. Um, and whatever states, other states or what other publishers, but that, you know, once it's passed by the legislature or, or congress for that matter or the regulations and, and you know , those stuff or , or court cases, those are in, those are public. They make money off of repackage them , printing them, repackaging them, formatting them, putting all of the other, you know, annotations and all, but the wrong information is public. So Sure. I'm assuming that the internet has read every case by every court, every law passed by every legislature and can find that somewhere.
Speaker 2:Yeah. So, but the annotations are valuable to, to a service like absolutely gt because that's how it gets the context of it and what
Speaker 1:The Yeah . That's point
Speaker 2:Action should be to those laws and the interpretations of them. Um , but yeah, so things like that, I mean, but every webpage could be scanned and the data off pulled off of it to be compiled.
Speaker 1:Yeah.
Speaker 2:You can ask TP to compile a list of the boards of the top homeowners associations in Maryland mm-hmm . <affirmative> , and it'll go to each of those homeowners associations websites and, and strip the data straight off their websites. You know, what if it starts feeding off Facebook data, you know, and knowing everybody's the relationships that all of us have with each other, you know, that these are Chuck's friends, and then I tell chat , g b t give me a list of Chuck Cassie's closest friends and pulling his Facebook friends and then what their net worth is and it goes, or what , what their addresses is. It goes up and pulls their SDA data mm-hmm . <affirmative> for all those people because it's being able to combine all these different pieces of data on all of us. So that's where some of G D P T stuff, they from Europe and stuff like that come in of privacy laws . Yes . And how, how we start protecting ourselves from these search engines that can find anything. It's a big issue.
Speaker 1:And Yeah. Yeah. My last co my last question is sort of related to that, but we, we, we talk about that this information and where it gets it from and it , and it knows the internet. I mean, it read the internet for Christ's sake. We used to think of that as a joke, right? <laugh> , we used to say, oh, I read the internet, I know everything about it. Well, this actually did, but here's my question. I it's about bias. Because when you look at the subset of people who quote unquote wrote the internet, it's a specific type of person and, and even a certain very specific for the most part demographic. And isn't it a problem, I know it is with, for example, facial recognition software notoriously bad at recognizing faces and analyzing faces of people of color because none of it was written by people of color. And so is bias by virtue of the source material, I , is that a constraint here that we should be concerned with?
Speaker 2:It all depends on how you're using the tool. Yeah. Um, if you're asking the tool to do creative writing, for sure.
Speaker 1:Yeah.
Speaker 2:You know, if you're asking it to ad hoc, write a , write an article from scratch, based upon these couple criteria, it's going to find the most likely set of words that go together, like I said earlier. And if that is a set of words that are most likely has been because it's been written by a certain demographic of people, then yes, it's going to view those viewpoints as prominent and push those first. So look large acts as an echo chamber. Um, it's not necessarily biased, it's simply a reflection of what it's been trained on, which may be biased.
Speaker 1:Exactly. And that's why, for example, to bring back to real estate in a search, what are the crime statistics for this specific, whether it's a zip code or neighborhood or whatever. And we know that in, in some circles, that is code for the demographics and that it might be a, you know, majority minority area. And that's supposedly in my mind, reflected in the crime statistics. And so we're not, not an explicit , uh, bias, but it certainly is something that I'm gonna carry with me.
Speaker 2:Yeah. I mean, I think it's part of a larger issue. I mean, it's not
Speaker 1:Restricting Oh , no question.
Speaker 2:I mean, anytime someone is targeting a specific demographic, as you and I both know, you know, you buy a , a database of data on people and it's basically largely upon census data and then making extrapolations off of census data and then feeding it from there. Exactly. And if you have somebody that doesn't fit that census profile for that particular sub zip code, then it's gonna make mistakes. I think that the bigger thing is not expecting it to do the work for you, that the internet is full of bias. I mean, if you follow Yes . You know, if you go to any chat chat channel or whatever, you know, you'll have people arguing both sides of the coin fighting over it with, you know, militant fury over, you know what color Eminem is better <laugh> ,
Speaker 1:<laugh>,
Speaker 2:You know, we all know
Speaker 1:It's green , so stop that. Yeah,
Speaker 2:Right. Exactly. <laugh> much less a major topic of the day. Um ,
Speaker 1:Exactly . <laugh> .
Speaker 2:So no , I would not trust it not to be biased because it's trained on a biased source. Right . With that said, that's my
Speaker 1:Point.
Speaker 2:Um, you know, like open AI is trying to de-bias their ai. Yeah . Um , which I think is a mistake as well. Um, they're basically putting their fingers on the weights to try to get it to be unbiased, but I think that that starts getting, starts making everyone question the AI in general as far as it's only as good as it's programmers if you start waiting your biases.
Speaker 1:Exactly.
Speaker 2:Um, and I think that that simply creates an echo chamber of a new set of biases that the programmers are putting into it. So I think it's a dangerous path to walk down to try to unbiased data , um, maybe selectively choose where you get your data from. Maybe, but I don't know. I don't think there are any good answers here, Chuck.
Speaker 1:Yeah , no, I , I know I wasn't the final thought here for me, and then, and then I'll let you have the actual final thought is that these are tools, like all tools, they aren't the end of the conversation. And it requires us to use these in a way that we consciously know have limits and that we explore, but we know of the limits and we know what they are, and we double and triple source things that we don't just take the easy way out because this thing wrote my term paper for me, and I'm just gonna hand it in even though it's riddled with errors because all of this stuff is, can be very flawed. Um, and that we just can't take its word for it. That that's pretty much where I am. What about you?
Speaker 2:Yeah, I , I'm pretty much, you stole my thunder on it being a tool <laugh>. I use it all the time at work.
Speaker 1:Yeah, I know . Yeah, we talk about it. Yeah.
Speaker 2:I can write an outline of a article that I wanna rewrite . Yep . I give it all the facts and it simply builds an initial structure for me. I use it as a draft and then I rewrite what it wrote. Is that cheating? No, it's using it as a tool, but the ideas are mine. When you start depending upon a machine to provide ideas for you, then you're simply parenting the internet, which is never a your choice.
Speaker 1:That's the point. Yeah. Uh , well I think that's a great final thought, Michael, thank you so much. That was , um, I enjoyed that a lot. Thanks for, thanks for coming. Thanks
Speaker 2:For having me. Chuck .
Speaker 1:Uh, to our listeners, thank you for the privilege of your time. This is Get Real Estate, the Maryland Realtors podcast. I'm Chuck Caskey , Maryland Realtors ceo . Thanks as always. To our esteemed producer, Joshua Woodson , please subscribe wherever you get your podcasts. Like us, share us, give us five stars if we've earned them, and more importantly, give us feedback, including guests you'd like us to invite or topics to explore. So be kind, stay safe. E o Wilson said, the real problem of humanity is we have paleolithic emotions, medieval institutions, and God-like technology.
Speaker 3:Hi ,