This is A Very Spatial Podcast, episode 748, October 13th, 2024. Hello, and welcome to A Very Spatial Podcast. I'm Jesse. I'm Sue. I'm Barb. And this is Frank. And this week, we're going to be talking about, well, finally talking about AI, I think. We'll see how it goes. But first, of course, some news.
And the first thing up in the news this week was an announcement at last week's GIS pro conference that The Urban and Regional Information Systems Association, URISA, is rebranding as the Geospatial Professional Network. We have an email out to them. I've heard back from Wendy Nelson, the Executive Director, and we're trying to schedule a conversation to talk a little bit about this in the coming weeks, so keep an eye out for it. You can go to geospatialprofessionalnetwork.
org at this point, but it's still in development. So feel free to go to URISA. org because it's short of type, but since most people that I know no longer actually type in URLs, except for me, and most people search for it anyway, it doesn't really matter how long URLs are anymore, I guess. Well, I mean, you know, when they started with all that. You know question mark random string of letters and numbers. I mean, have you ever tried to type in a URL for like a story map?
Well, I'm just the main domain and then get there from there. But no, I think this is cool I mean they have always had like really interesting events and three sources and stuff for professionals. And so just thinking about Broadening it out From I mean, we knew that it was GIS programming GIS pros, GIS pro that worked in a lot of areas, but sometimes the name would be, it would be kind of a little confusing to people that first were starting out. So I think it's pretty cool.
So I'm excited to hear kind of what their vision is now. And presenting the con, I would say that I think it's weird and I don't know why they're doing it. That's the short of it. I mean, Eurysa is a, is a, is a brand in its own right, you know, and it's just kind of like, yeah, we're going to chuck that. That's that's like, in some ways I feel like the geospatial professional network could have been spun off into its own thing, yet keeping Eurysa around. That's what I think.
Okay. They disagree. But I'm allowed to have my own opinion and I have a podcast and a platform. I can say it. My, my thought along with what Sue said about, you know, this is more understandable to students is for everyone to remember to go back into any documents where you've mentioned URISA as a resource and update the, you know, the link in the name. Okay. With that, then China has completed its national eLoran. Of course, the U S got rid of eLoran.
Then realized oh wait, we shouldn't have done that and is in the process of re Constituating is either and with new updates but china has completed their Implementation. There we go. That's a good word implementation of iliran and so yeah, they of course now have a ground based system to go along with Bido bado Just something to be aware of 60 years to do it. I think we, we forget how long things like this take from the, the, you know, initial implementation and dream to fruition.
And finally, in the news, the European space agency has released a new strategy for observation last year around this time was a little bit earlier. They announced that they would be looking at Developing and strategy and surprisingly, it only took a year for between when they announced what they were going to do and the release of this. And the key here is that the strategy is centered on 23 questions. We're not going to go through the questions.
Feel free to go to their documents and look at that. But the questions are there to kind of guide their next we'll call it 15 years. So through 2040 approach towards Earth observation for European Space Agency. And really understand the year. So there, there are 23 questions, but they group into like six themes. So it is those cycles and systems like the water cycle, the carbon cycle and impacts to that. So, no in hazards and things like that.
So those are, those are the, the broader categories that they're going to focus and the array of ESA sensors are many of them have very tailored missions to help out with this. And so I think that.
That is what we need in our EARTH observation efforts, right, is to have these kinds of strong guiding, well, principle is not the right word, but, but guidance, let's say to help say, look, these are the types of things the sensors have to have, this is what we're going to do with the data, when users, you know, kind of have input in that as well, then it just, it just helps us keep awareness and really look at long term, the benefits of having these systems out there.
And that's it for the news. In the web corner Geo Awesome, the blog has made a call. They have created a You Are a Geo Awesome video series, and they're asking the geospatial community to do short videos on who you are and what you do. As you know, we always talk about, you know, that we're so spread out and broad and embedded within different tasks, industries, goals, and, you know, even the types of work we do.
So it would be a good place to think about doing a short video to, to tell the world this is who we are and what we do. Okay, so we have waited and probably should continue to wait. But I, I decided, okay, let's pull off the bandaid and have a discussion because there were a lot of things this week that kind of Our reasons to talk about AI. There's news items from Google and from near maps and QGIS that all talk about their, their direction in terms of, of AI.
And I think one of the biggest concerns that I've had about talking about AI is the fact that we've been talking about AI for a long time. We don't always explicitly denote it, but whenever we're talking about Supervised classification or these type of things in remote sensing that, of course, is building on AI. That's been around for a long time.
And so whenever we talk about AI in the geospatial context, traditionally, we've been talking about using expert systems, neural networks to, you know, help decide things. Sometimes using textual content, sometimes using spatial contents in terms of vector, sometimes getting into those rasters that we're looking more and more at with things like machine learning.
So I just wanted to give that caveat before we get into it, because, of course, from here, we're using all of those plus generative AI in this conversation. So. What do you guys think about the use of AI for geospatial technologies nowadays?
So Because I'm a nerd, I opened up chat GPT, and I said, what should I know about GI, GOAI, and it put out, it's kind of funny because it was spitting out a lot of the things that you were just saying, you know, not in different verbiage, obviously, but it was, it was saying a lot of the things that you were just saying. And I was like, oh, okay. It starts with core components of GOAI. It's geospatial data. It's AI machine learning algorithms. It's big data and cloud computing.
And of course, glazing over a lot of the text here, applications of geo AI. It says urban planning, environmental monitoring, monitoring, agriculture, smart cities, public health and safety tools and technology, GIS platforms, AI libraries, remote sensing. And then it goes into challenges and potential impacts for academia and future trends and all that sort of stuff. So the thing I thought was interesting about this is that it's really just.
Spinning back as we know, which had GPT, what it knows out there that it can tap into and say, let me pull in this information. And the one that caught my eye was applications of geo AI, which it lists, like I said, is urban planning, environmental monitoring, agricultural, smart cities, public health and safety. And I was thinking, well, is that We can use, we can use G G I S for a whole lot of other things.
So why aren't, why aren't those things in there with the G O A I bits, you know, because the, the people who wrote the papers that's based on included those, right? Which core to lead you to a circular question of, okay, well, if that's all that G O A I can do, then. Are we limiting ourselves since we knew GIS could do so much more? What is on the frontier of, you know, geospatial and AI? It's kind of an interesting building circular thing.
That's what I thought about whenever I was reading this list from chat GPT. Well, I mean, okay, so I go back and forth on, on various things related to Generative AI and other things, but, but isn't, aren't we talking about different things, right? When, so when we say AI, there's a lot of things going on, and especially in, in your intro, Jesse, and, and what you just, you know we're talking about, Frank, cause, so there's using those types of, of models. Which type of models?
Cause I, I mentioned different ones, so. Yes, okay, so things that could be, like, Lumped under the large umbrella of AI, but are largely models to gain insight from data, right? So neural networks, and in some extent then know to Essentially build decisions based on the initial relationships of data. So there's that right the research part But then there's and you can you could correct because you already give me the big guy and then what you're saying?
Yes, but then like so some of the things about generative ai Are the things where the ai is guiding? How we synthesize and present that knowledge or those insights right that are being gained so You Those are, those aren't the same thing in terms of their goals, if that makes sense. Correct me if I'm wrong here, this is what you're saying is that the way we process data using AI is different than the way we generate knowledge using AI. Does that make sense?
I mean, it's a reduction down, but I mean, like, I, I think that, that, that at least two, if not more kind of ways to see the application of it in the field, but the second one, right, the using it to, to synthesize and, and, Discuss our knowledge right is the thing you're speaking to where it says, well the applications of it are only these as Jesse pointed out, because wherever it was, you know, scraping that from only had those things in it.
So Sue, when you were, were talking about the, the different ways and uses, it would make me think of what, what I was thinking when I was reading through some of the, the show notes which is the use. Of G. O. A. I. For decision making and land use.
And that a lot of times you see it coming from again from a higher level policy in a country and then creating that as as a use case or best practices than being adopted and used in other countries, which to me, mirrors The early days of AI and its major use, which is out of you know governments and out of research institutions.
But at the same time we have the, the use going on In a myriad of ways by by industry and innovators which in some ways is the side that might not being picked up by that generative AI, because that's where we're seeing it pick up the white papers and the research and things and not the side that's you know, not being spoken about, but is being done, which is that industry side.
And I don't know where I'm going with this other than that I, you know, when I was thinking of it, I was thinking more of the GEO AI in terms of spatial data is, you know, hugely big data, but then also as. A tool to help to accelerate what we need to do when you're talking about land use, agriculture, sustainability, which also is a little ironic considering the you know, questions and challenges around the AI and it's you and it's use of materials. And so.
What did you tell me about the energy usage? The energy usage, energy usage and also land usage, and then you get into this, you know, big discussion of it, but also with all this stuff we were talking about the ESA and their 23, you know, goals, it's how can we achieve, you know, Those goals, if we don't have some type of geo AI to accelerate what we're trying to do. But then what are the, what are the implications of that geo AI?
Do we have enough people that can make sure that, that what we're doing is appropriate on track, you know, multiscalar. Okay. So this of course is often the issue with. Conversations. There are lots of them and they're all trying to happen at once. And that's, that's, that's kind of what I expected would happen because we're saying the words AI and geo AI and.
These are things that are large umbrella terms, and that goes back to, you know, what I barely touched on at the very beginning of expert systems, neural networks, machine learning, deep learning. These are not all separate. In fact, whenever you look at deep learning, it is using a combination of machine. Well, it is machine learning, first of all, but it's using expert systems and neural networks to do what it does.
So. You know, we've been building over the last, I don't know, half a century out these technologies that get us to where we are today for the raster based analysis and raster image light, our data sets, those type of sets of data that we now have, you know, broad access to. And so, of course, that brings in things that. Chet GPT gave us like our big data access, which made a lot of this possible.
So without easy access to all these images over time, over the entire planet both from above and at level. So, you know, satellites, planes, drones from above.
In car static CCTV, those type of things at level, all these data sets are feeding into machine learning, deep learning, and that doesn't even necessarily get us into the development of the LLMs for the, and I always have to look up what GPT stands for, for the generative pre trained transformers that we're using to help us create Text to help us create audio to help us create video, and I'm assuming at some point in time, we're going to see someone put out a GPT that's
going to be utilizing geospatial data more directly, as opposed to right now. So. I think a tiny conversation within the broader conversation real quick is what are your thoughts on how GPTs, and of course the large language models behind them are being used for the geospatial industry. So how does that intersect with geo ai? What are your thoughts on it?
Well, well, I mean it, the thing that struck me as interesting is the entire time you've been talking a hundred percent appropriately, but it is interesting that if you were paying attention. Out there. You would realize that it's so dominated by raster, like all the stuff we were talking about, the examples, the imagery, all that stuff is all dominated by raster. And of course. The other dominant geospatial model is vector, and it certainly there is a I that operates on vector. That's not true.
That's to say otherwise would be incorrect. But I do believe that the raster piece as much more robust and developed at this point than the vector piece. So, and that's changing and that will get better over time. But it does speak to You know, a large language model doesn't have a raster or vector, really, not really, it's language, it's got the word language in there, and we don't really use language in our geospatial data models.
We stick it in those attributes, we have it, it exists, and we get all the semantic web and all that kind of stuff like that, but we don't really use it in our day to day work. So Really, the interesting piece here is that we've got a technology, a long, a large language model that is designed to use a certain type of data set that we don't actually utilize that much. So part of it is we have to figure out how to, I'm curious if the large language model is going to push the geosemantic web.
More and we're going to kind of get pushed on to us, or we're going to say, oh, we want to play and we'll develop it quicker. That's what I think has to happen before a GPT type of scenario has a significant impact into what we do in the geospatial industry. Yeah, because I have to admit. So again, I don't engage with this. this that much. So, yeah, feel free to correct me on everything.
But my only real engagement with the, the GP style stuff has been in code generation to help me out, like when I have a problem. And so even in the geospatial, that is more an interface to help me if I'm doing another task that may involve other types of, of things like Classification, but so thus creating a language interface kind of thing. So I want, you know, I might generate some Python code or when I'm working on my other work, C sharp code.
And that is the bulk of, of what I do with any GPT style.
So, so I can't really speak to a lot, but, but that's why to me, it's kind of, I see a break there mostly because again, an application, my use of it break between What we had talked about for especially, you know the examples you mentioned related to raster classification and those types of things that we have done with probably AI versus again, the GPT style, which as Frank, I think I explained pretty well that as a language model, right?
It's its main function is to essentially bring out things that are and I don't want to get into like an AI algorithm where you're Our text based well, even if they're, you know, even if you use it for images, things like that, but they're trying to represent the things that we would show to other people to take sort of the other stance that this can be used for is, and I have used it for is sort of a organizing strategy.
I don't know if that's the right word, but certainly strategy for how to approach doing. So, for example, I just wanted to check GPT and I said, how would you evacuate Tampa, Florida? If a category 5 hurricane is approaching? Right. So this is something clearly that, that unfortunately that area of the world just went through ask it how to do that. And obviously to my mind, that has a geospatial component to it, right? Cause we're talking about evacuation routes and transportation.
And of course the, where, where exactly is it going to hit? Is it actually going to be a category five? All those things are clearly we can model with GIS at some level. So I asked her that and it came back with things that were geospatial. For example. Early preparation communications, emergency alerts and announcements. So you should have those this should be communicated through multiple channels, zone based evacuation orders. Tampa has established evacuation zones based on flood risk.
Evacuation will begin with the most vulnerable zones and progress accordingly. Those in low lying areas, mobile homes will be evacuated first. Okay. That's a very reasonable strategy, but it didn't present to me anything that geospatial person could. Just take and use. It just said, look, these are the things that you should go find and then do the analysis to figure out, you know, who do I got to call and who are your contact? I'm going to jump in and ask Frank to type in to the next prompt.
How would I do with this? With GIS? Ah, good question. All right. So let's do the exact same question. No, no, not the same question. Just add that as your next prompt. Okay. So that should just build on this, the conversation with GIS. All right, and and the thing is we're using chat GPD because I actually have access to it. But any of these are going to have the same, you know, All right, so it tells me what GIS is and how it can be helpful in this situation.
It says mapping vulnerable areas and evacuation zones, create evacuation zone layers, flood storm, flood and storm damage data, get NOAA and FEMA's flood hazard stuff into your GIS. Go find some demographic overlay. Design and optimize evacuation routes using traffic simulation, road network analysis. So again, it's telling me what data layers that if I were doing a project, I should go grab and it's telling me the broad approach, the strategy for how to incorporate or analyze this stuff.
But it's not really taking that next step. It's not really providing anything that I don't think a geospatial expert wouldn't immediately go, Oh, I'm just grabbing these things without any trouble whatsoever. Well, okay.
So I think that's, that's where my perspective on this and yours are different is that this isn't for the geospatial expert who knows these things, who's doing this every day, but the new people into the field or the people that are in areas that don't have somebody who has the experience, but has a GIS person who's being thrown into this situation, they can go in and start to. Build through various prompts to where they're going.
I see where you're going, but what, what I was about to say is that the problem I have with this is that it's giving you a strategy that requires geospatial expertise to actually execute. So at some point, you know, it's, it's. As I'm going through these prompts, it's saying, well, you'll get this and you'll get the other thing. I cool. Now, what do I do with it?
And you go, well, you analyze, I don't think this language model is ever going to get an atomic unit where it says, go here, click this, open that, put this in here, run this process. It may, but it's an awful lot of analytical work. To ultimately get there, and I'm not confident it will, but I think in some cases, we're criticizing a system where we know they are making calls to philosophers, linguists, people like that, you know, to put the system together. But yet we're not asking again.
Where are the geographers? Where are the geospatial professionals? So we might be in geo AI. We might be working on it on that end, but on on this end we're not there and we, we need to be there. And again, it gets into, but they might not know to even talk to someone or call someone if they're not being approached to say this is something that's missing. So to, to come back to the question Sue's point of coding, this is a huge thing.
One of the things in the show notes this week is that you know, QGIS is pushing, you know, they give regular updates on new. Extensions that you can use in the software and this week they were highlighting or somebody was highlighting two of the LLM access. So giving you access to you know, use GPTs for discussions about pyjus or, you know, just different ways that you can pull coding capabilities.
And then the other thing that if you go back to Andrew Turner's and I can't remember if it was this year or last year's I think it was actually fed you see not you see but you know, whatever he was talking about the use of generative coming. It wasn't into, you know, doing things. It's basically going to be allowing you to find tools that you think you might need, you know, right now. If you know the name of the tool, you can go into the search and find it.
But if you have, I think it was in 3. 2 later versions, and then now in 3. 3, Of pro that you can go and in the search boxes, either at the top of tools or at the top of the interface and type in what you're looking for and what you're trying to do. And it's essentially giving you that. The response that's building on what you think you need to do, not that, you know, necessarily already, those of us who know that can just go and type in things real quick and and get what you need to.
But if you don't know what you need, you can start typing these things in and it's going to find keywords and what you're typing and align those both with tools that are there. But also it's pulling, I think, from and I haven't checked this to make sure, but pulling from You know, the whole help section. So, you know, there's a massive amount of content that's out there online about how to use these tools. And so it's, you know, taking that into account as well.
So I think those are the kind of current face value ways, but the coding side and the kind of how do I do this side, but I think we're going in other directions in the near future. You know, whenever we're talking about these things, one example of going beyond the coding and the, you know, GPT to help us find tools is the question of, you know, there's this thing, this thing that you guys are interested in looking at called deep mapping.
And, you know, whenever we talk about this, we talk about it in the context of the map a lot of times, but really, that map is supposed to be a qualitative thing. So how can we then begin to kind of kind of use this push and pull between the traditional GAO, GAO, GAO AI. There's too many, there's too many consonants. I'm just saying. One consonant, like every vowel in the English language.
So how do we take that in terms of the deep learning and mash it up with what we can do in terms of the generative AI to find information, and of course those that are coming out now to create audio and video as well. So you know, how is that taking us forward, potentially? Potentially. So, or am I, am I breaking, do I need to take this out because this is a plan for a grant sometime? It is not a, it should be a plan for a grant, but it isn't currently.
Okay. So we'll talk around it without talking through it. But I, I just want to applaud Jesse because that was a plot twist. I didn't see coming when this discussion started. And as soon as he said it, my mind just went right there. And so I, you know, I, I think yeah. So I wanted to put in the sarcastic comment whenever you said that many of you are interested in deep mapping. I was like, are we still have it as part of our dissertation?
And that is everyone who's ever done a dissertation will know. We'll just burn you out on that real quick. So I think GeoAI in the large language model sense is going to be critical to doing deep mapping. And I was actually just thinking, I've thought this before, but I was remembering that I've struggled in the past of having a mental model of a deep.
So my mental model of a deep map, and when I visualize it as I actually see a thin map on top with all the stuff that kind of goes subsurface for all the deep bits. But at the end of the day, it's still that point line polygon thing on top for the most part, which is. Reductive by nature. So I, I'm, it, it made me start thinking, Oh, wait a minute. What if we flip that and don't do that reductiveness, which is the intent of a deep map by its default.
Visualization is default start point, and that's one of the semantic geo model really starts to matter. And then large language models can have a huge impact on that, but I don't know how exactly that wasn't for today. Today's the conversation of around these things, not how to do something. Come on, you gotta get money to do that. Yeah, well, I mean, I guess I guess the thing is, is that it feels like a chicken and egg question to some extent, right?
Like, so, Large language models are trained upon things that exist. So the things have to exist before it could be constructed. But if I use to make it, then, you know, it kind of how does this start? I guess is part of my question. Basically, they're almost made for deep diving into things, you know, layer upon layer.
I only Question would would be about the fact that their citations just aren't there yet, the giving credit to who they're they're pulling from, because that would be an essential part of a deep map, but I could really see it as a way to like network out and deep dive through different layers and pull in things that you wouldn't even have considered before, or that wouldn't have been access before, meaning that some of the, the things that you would want to know about
a location that might not normally get the attention it deserves would be able to come to the surface would be really cool. But again, my, my thing would be, you know, to, to know where it's, it's coming from and to have that attribution or that metadata come with it somehow. Well, I think this, this is obviously for another conversation too, I think.
In the deep mapping conceptualization though, a lot of that has to do with, with the question, well can frankly disagree since you're delving into it more than me with the question of narrative, right, and, and direction of story.
So I think there it's interesting because the, the challenge of deep mapping was to take a very reductive visualization, right, of, of features and phenomena on the surface of the earth and try to re inject or inject, alright, all of that richness that is the, the human experience and the natural world that's going on there but not, not everything Without some sort of guiding, guiding hand, as it were, right? So, we'll call it curation.
So, so that, that'll be an interesting challenge, then, if you can bring in something like the, the large language model. They can, as, as Barb says, pull in from everywhere. And, you know, you mentioned citation as a, as a, a thing to be concerned about, which it is. But also, too inherently, The use of these subsystems don't necessarily apply any human framework of hierarchy or organization or direction. And so I think that's interesting in the implication, right?
But that's just me thinking about it because the narrative guidance of navigating through space is something that I have a lot of interest in. So to me, that would be an interesting challenge. And then you always get the, again, Sue, with what you're talking about, you also get in the inherent bias of things in that digital divide you know, into these, these deeper questions that come up with deep maps and anything that might be when you're working with spatial narratives.
It's just a really interesting topic to start thinking about. So not to get too much into deep maps. Sue, I think your interpretation is the, I hate to use this word, but classical definition of a deep map. I actually have a different opinion about what a deep map is and could be and should be. But it's not radically different. And I think that, that it doesn't preclude matter of fact, it, My definition in conceptualization would allow those narratives to take forth to take place.
So you would, it's not like you would get rid of them and there shouldn't be there or something like that. They would definitely be there. I had some point to that as opposed. Oh, right. So I did into chat GPT is how would I make a deep map? I asked it that. And it gave me back two, a long bunch of stuff. And about one third of it is. I believe I can, I can read all of our dissertation advisors. I can read it in his voice.
Because clearly as one of the people who are, are very much in deep mapping from a geography standpoint and have been developing a lot of the concepts, some of it is his verbiage. And then oddly enough, a lot of it is Esri's verbiage is they actually specifically talking about as we story maps and ArcGIS online and all that sort of thing so. That was kind of interesting that we're all struggling a bit with, with what is a deep map and how it works and how it focuses.
And let's try my next prompt is how do I use, how do I use geo, geo AI in a deep map? Because I'm not sure even the AI knows exactly how to come up with something. I'm sure to come up with something, but I think it's like this, it's conceptual still like we don't really know exactly how to do all that. And really, to be perfectly honest, it is giving me information that's not radically different than when I ask, what is GeoAI?
So it's just saying, well, this is GeoAI and this is kind of what deep maps are. So I don't know. Squish, I guess that's kind of what it's saying. I have to just say that this overall discussion of AI is bringing back a feeling I had early on when I was just getting started with geospatial, which is you would go out and you would look for what you need. Or what you wanted to do and the answers you would find would be very generic and you know, you'd want specifics.
You want, how do I do this, but you'd get back and that, you know, here's what we think we could do. I don't know, go do it or wouldn't that be great. And, and it's, it is that stage of, but I know some of the things now that I wanted to do Give me the tools, you know, give me the, the steps, give me more direction from here. Well, I think that it kind of follows the arc of conceptualization technique and then tool. Right. And so we're still in that. Haze of conceptualization and technique much.
We're not at tools. And so many people are just looking for, for tools, because I just, let me give me some buttons to hit and things to do, and we're not quite there yet. But I think that to go back to the intent of this, this conversation of AI, I think we're similarly in AI when it comes to geo AI, and some cases we have tools. We have actual real good tools with raster that do great stuff. In some cases, when it comes to vector, we have.
Some tools and some workflows that use tools and some thoughts about how tools could tools, you know, we're somewhere in that zone, but we're talking about using this in a more language sense in the case of, for example, deep maps. There's other ways we could use language models and mapping, I think, but deep mapping is a great example. We're still in the conceptualization you know, technique side of that spectrum. And so that. I think, I don't know.
Is there anything else anybody wants to say this time about AI? Cause again, it's, it's one, the goal today was to say AI is a monolith, but that monolith has many pieces within it. And we actually, as geospatial professionals have been using it in many ways before the generative pre trained transformers and LLMs. But now we are beginning. To have access to those as it becomes more readily available. So is there beyond that or as part of that anything else?
For me, I think the, the interesting perspective that I get from this is that we're in the stage and, and of our adoption, potentially, of large language models and things is that it's very much driven by what you bring to it already, right? So if you approach it as an expert who can already evaluate the results it gives you in a certain way, and then, and I think both Frank's and Barb's comments, right? So then, then you want more, right? You're like, no, I know, I already know this.
Pretty much you just reorganized it for me or synthesize some thoughts that I already could have had myself. And so I want I'm ready to do more, but someone who approaches it with no experience, right? We'll just go. Oh, all right.
But. That also means that your evaluation of your next steps after utilizing these tools will be very different, and so I think that therein is the, the experiential crux of the problem, right, is that, that it's hard to imagine those of us who are experts, it's hard to imagine going back to when we didn't know And thinking about how we might have used something like, like this, if we're specifically talking about the LLM kind of thing and GPT and stuff, it's
just hard to, to dial that back, but I, I have questions, ultimately, I have questions about those who don't have that framework already in their use, but. I made a comment based upon Sue's comment is it's a little bit like a recipe versus a chef, right? You call a recipe and that's important. But as a chef, you may go, well, I need to know this technique and then, you know, this ability I can mix in a little bit of that. It's sort of that difference between the expert and the.
You know, non expert in a lot of, a lot of sense. I think that analogy works for me, at least. But the point I was going to say is that AI is a great sounding thing, but really it's not. That's the thing that a lot of people are missing. It's not really intelligent. It's just doing some astounding processing on the back end. So it looks a little bit like it's seeing some insight, but it isn't. There's another way you can do AI that we started, we started doing this approach of machine learning.
Back in the 60s, essentially, arguably even before then and we started another way of doing that at the same time and as technology has progressed, we've turned out that the way that we are doing right now has worked quite well. The equations, the algorithms we have have been quite good, and now we have the, the power and the data to actually make this into something. However, it's.
Remarkably stupid if you get down to actually thinking about intelligence, the other way is effectively the kind of the way we think about it in science fiction, you know, where you're essentially training somebody from birth until an adulthood. That's the traditional way intelligence has always worked in pretty much all living creatures that have intelligence. The problem is, is we haven't gotten very far on that at all.
And so when we think about AI, we tend to think about it in the science fiction sense. But in reality, we're in a different sense. And I think that part of the challenge here is AI sounds cool, but we really do need to be thinking in terms of large language models and machine learning and that sort of stuff to indicate the philosophical limitations.
And so forget the word or the letters AI, the discussion of artificial intelligence, focus on what you're using, because as you saw very quickly at the beginning of this episode or conversation it got very confusing because we were all talking about different parts of artificial intelligence and just calling it AI. And even whenever we tried to say, okay, well.
You know, focus on which one it is, and it's still hard to do that because we're just so used to talking about these things as general geospatial technologies that we don't always use the explicit terms. And can we at least all go on record as saying shoving all the vowels into one word with only one consonant is. Awful. Do I. It's just a terrible name. I get it. I get how we got there, but it's terrible. Well, you would love Japanese, right?
It's consonant vowel, consonant vowel, consonant vowel. Well, again, at least another consonant or somewhere, you know. So instead of G O A I, it's G O R I G I don't I don't know how to fix it. All I know is that G. E. G. E. O. A. I. It's all of them. Just stick a Y in their G. O. I. E. And then just to make it all round about everything. Because why is kind of both sticking to my statement of just ignore a I and talk about what it is, which which type of analysis or model it's using.
So with that, then We will, I'm sure, come back again to AI because we didn't necessarily do a great episode this, this time, because there's just so much to talk about that. It's hard to deep dive on any one section whenever you're first starting. So on defense corner, as always, we encourage you to go to events. GeoWeek is going to be taking place the 10th through the 12th of February in Denver.
And for those students who are interested in going to GeoWeek they have announced just this week and it's only open until the end of October. So be sure to check on it quickly. The GeoEmpower event scholarship, and this is going to award three, I believe, students who apply basically a full ride to GeoWeek. So check that out if you are interested. The 16th International Conference on Environmental and Rural Development is taking place March 13th through 16th in Tokyo, Japan.
Abstracts are due by November 15th. Of course, if you want us to add your event to the podcast, send us an email to podcast at veryspatial. com. If you'd like to reach us individually, I can be reached at sue at veryspatial. com. I can be reached at Barb at veryspatial. com. You can reach me atFrank@veryspatial.com, and I'm available at kindaspatial, and of course, if you'd like to find any of our contact information, head over to very spatial.com/contact.
As always, we're the folks from very spatial. Thanks for listening. I we'll see you in a couple weeks. Be to me, to me. Ah, it's clear to me. It's clear. I can see it. It's not too clear. Be to me, to me. Yeah.