Speaker 1: In this video, I'm going to put DeepSeek R1 model into test for a simple mapping task to see how it holds up. Now in case if you're not aware, DeepSeek R1 is a large language model that's currently taking the internet by storm, especially because of the fact that now it's being seen as a direct rival to OpenAI's ChatGPT. Alright guys, let's fire up DeepSeek, which you can do simply by heading over to chat.deepseek.com and I'm going to specify a very simple task like I want to plot the 20 most visited cities in the world on an interactive map. So let's go ahead and provide the instructions to it right over here. And additionally, I'm also going to specify that if I were to go ahead and click on one of those points, I should be seeing the name of the city and probably the rank as well as the number of visitors. So let's just go ahead and specify that as well. And for this task, I intend to use Python in a Google collab environment. So I'm just going to specify that as well. Now, one of the cool things that you get to have with DeepSeek R1 is first of all, just like in the paid versions of ChatGPT, you can actually turn on the web browsing option simply by clicking on this button. So whenever it's necessary, it will basically refer to the web depending on the task that we're getting at. And even if you look at the very thing that I'm asking, obtaining the 20 most visited cities in the world, to get the most up to date information, I think referring to articles from the web would be a wise thing to do. So I think in this particular case, I'm just going to turn this option on. And with DeepSeek, you can actually get this feature completely for free, at least in this current version. And right next to that, you can see an option to turn on the DeepSeek R1 reasoning model. And when we do that, we get an outlook into how his internal brain is sort of making relevant decisions by applying reasoning. So I'm just going to go ahead and hit enter. And let's see how things are going to work out. And guys, this is exactly how this reasoning model works. So you can see what's going on. It's basically finding conflicting information. And based on reasoning, it's deciding to make the ultimate decision because you can actually find conflicting information when you do things like web searches. That's how that reasoning model actually comes into play, especially in a task like this. And of course, followed by that, we get the complete output. And we will see whether we will be able to get this to work directly or whether any alterations are required. So right over here, you can see that based on the information it managed to find on the web, it basically listed out the 20 most visited cities, along with the rank and the corresponding number of visitors. And of course, to plot this on a map, you're going to need latitude and longitude. So you can see everything was completely automated, you didn't really have to go and do any searches by yourself, it basically grabbed all the information, including the lat long information by itself, and put that into a pandas data frame. And of course, as you can see right over here, we're making use of the volume library to display this on an interactive map. And it's pretty cool that they actually provide some key insights from the data that it managed to find as well. For example, Bangkok happens to be the most visited city in 2024 with 32.4 million visitors and a couple of other remarks as well. Alright guys, I'm just really eager to see how this is going to work out when I copy this into our coding environment and see whether it actually works or not. So for this, as I told you guys, I'm going to use Google collab. And in case if you're not really aware what Google collab is, it's nothing but a simple coding environment, I'm just going to leave it at that, which you can instantly access if you happen to have a Google account. So you don't really have to bother with installing any development environments or anything like that, especially if you're not really into programming. In order to access collab, all I'm doing is I'm logging into my Google Drive, head over to the corresponding working folder, right click, go to more and click on Google collab option right over here. And that should open up this sort of a development environment. And all we have to do is just head over to DeepSeek and we're just going to copy this one by one. So first of all, what it's doing is it's basically installing the corresponding libraries, just going to run this by clicking right over here. And that should actually install and prepare my development environments by installing the corresponding libraries that DeepSeek thought is necessary in order to carry out this task. And you can see that it basically ran without any issues. And after that, I'm just going to add a new coding cell. Let's go ahead and copy the second cell, copy the code from the second cell into this cell right over here. After that, you can just simply go ahead and run. So you can see it's basically importing those libraries and specifying the corresponding data to create a pandas data frame. And it did that instantly. So we'll go ahead and create another empty cell. And we'll go ahead and copy this into this cell. And we'll go ahead and run this. You can see that instantly, it managed to plot the top 20 most visited cities in the world on an interactive map like this. And one of the cool things is that I actually instructed the model to make sure that when I click on a point, it should be displaying the name of the city. Its rank and the number of visitors as well. So let's see whether it managed to do that or not. Let's go ahead and click on this city right over here. And of course, it's showing me the name of the city, the rank and the total number of visitors along with the corresponding year, which is something that I didn't even ask it to do. But it's really nice that it managed to understand that when we state the number of visitors, it's generally wise to state to which year that particular figure belongs. And for example, let's say if I were to click on this button right over here, you can see that Seoul is the 11th most visited city in the world, according to the data from 2024. And similarly, if I head over to this side of the world, let's see, well, you can see that London ranks as third most visited city in year 2024, according to the data that DeepSeek managed to find. Pretty cool, isn't it? All right, guys. Now, when we look at this map in this manner, it's kind of hard to get a quick idea about the most visited cities, even out of these 20 cities that we have picked. So I'm just going to propose something like this. Let's go ahead and color these markers in a way that it shows the first five most visited cities in red, maybe the next five in blue, the following five in green, and the last five cities in yellow, something like that. So we can head back, and I'm just going to say All right, guys, I'm just going to go ahead and copy this code from this cell. And let's head back to our Google collab environment. And we're going to have to create a new coding cell. I'm just going to paste this and you can see how it basically defined the colors according to my request. And you might notice right over here that we probably would have to actually get the corresponding data set from the previous coding cell because it didn't really take the time to print that out, which is totally fine. I'm just going to grab this set of data right from here from its previous response and just going to paste that right over here. And after that, we can just go ahead and simply run this. And guys with that, did you see what happened? Well, first of all, thanks to this legend right over here, I can immediately get a sense as to which cities have been visited the most. So you can see that the top five are right over here, Bangkok happens to be number one, Singapore happens to be number five, we have Dubai, and London and Paris. And let's see if we were to look at the last five cities from this compilation of this 20 most visited cities in the world. Let's try to see what we get if we were to click on this marker right over here. So that happens to be Hong Kong, ranking at 20, Bali, 19, Milan, Mallorca, and Barcelona. Pretty cool, isn't it guys? Well, before these large language models, we actually had to manually go ahead and find out this information by ourselves, whether it be the rank of each city, the corresponding figures, and quite importantly, the corresponding latitude and longitude information. But you can see that all of that is basically enabled just with a single click thanks to this kind of large language models. And this task, if you think about it, it's actually not that complex. So if you were to actually use chat GPT, you probably might be able to see which model is going to outshine in which particular aspects. But as the tasks become hard and complex, you'll probably start seeing which model outperforms which depending on the outputs that you're going to get and based on the level that it's going to make your life easy when you compare the user friendliness and the performance of each model. So guys, I'm just going to leave it at that. You guys can go ahead and experiment with these models. And even if you want to do a one to one comparison, deep seek versus open AIs chat GPT, well, that would be an interesting comparison. So if you do have any questions, you can add a comment down below. I'll see you guys again with another video soon.