The Nature of the World and AI with Rishal Hurbans - ML 177

Hey everybody, and welcome to another episode of Adventures and Machine Learning. I'm your host, Charles max Wood, and we're talking to Rashaul. Now you're in South Africa, which is a place I've always wanted to visit. I think it's the videos of the sharks jumping out of the water and eating the seals or whatever that I've seen from down there.

But you're the author of Rocking. Let me see when I get this right, Grocking artificial intelligence algorithms for manning, which makes you an expert, right because you literally wrote the book on it. But yeah, I'm super excited to talk and just kind of figure out what's in this in these algorithms, how to approach them, how to understand them, and things like that. And yeah, just welcome to the show.

Great, thanks for having me. Chuck one thing around. Expert Actually, personally kind of really dislike that word. Ninja for a ninja.

I dislike all those words. The reason being is I think, you know, it's just a personal thing. I think everyone, including myself, always has something to learn in whatever space you're working in. And when you say expert, people think you have every single onstitute everything.

Yeah, just in the side. Yeah. Yeah, it's funny how people approach that. You know, for some people it's you're an expert because you know a little bit more about a thing than somebody else, right, and then for others it's, yeah, it's this, well, yeah, I can ask you a question you just know off the top of your head.

Boom right anyway, So yeah, in that case, maybe that's a case of human artificial intelligence. Anyway, let's dive in and talk algorithms here for a minute. So we've had a lot of conversations with people over the last few months that we've been doing this, where we've been discussing like specific use cases for artificial intelligence, or maybe we've talked in general about a particular algorithm or approach to solving a particular problem. What's the focus of your book though, because it seems like it's a little bit different from that.

Yeah, so it's quite interesting. I mean, I didn't actually with the idea to write this book. It happened from I've been giving a few talks at some conferences introducing neural networks to people, and it was many developers, and it came from I guess something I saw happening around me. In my work and the community.

I'm kind of part of where people felt quite afraid to experiment with these techniques or technologies because it was really math heavy, or a lot of the literature and material you'd find is really math heavy. So I think after doing a few of these, manning approach me to, you know, see if I'm interested in maybe writing a book with them. And this whole book is essentially a visual tutorial on various artificial intelligence algorithms right and AI, and that term motificial intelligence.

I think everyone has a different definition for it, but basically this book can. Yeah, this book contains not just kind of your typical machine learning and neural networks and reinforcement learning, but it starts at the beginning with kind of how do we solve problems in general? So it starts with introducing the intuition of thinking about data and thinking about algorithms and computing and problem solving, and then it kind of goes into search, which is something we've solved a long time ago, but still something very powerful in a lot of applications these days, and something that I'm really passionate about, which is nature inspired algorithms.

So you know, we've got genetic algorithms, which is based on the theory of evolution. There's a colony optimizing, which is based on how ants actually move around in the colonies ask for help collect food. And then there's also particle swarm optimization, which is based on how birds and bees and those kind of things actually flock. So essentially, in a nutshell, it's a kind of gentle introduction to these different algorithms that with the teaching method is more focused on kind.

Of visual earners. I would say, as well as you know, kind of practical examples that you can work through rather than you know, theoretical kind of math proofs that you that you would study. Right, that makes sense. I kind of want to roll back to the initial approach that you had as far as just talking about how we solve problems and things like that, and how that kind of leads into some of these algorithms.

So what does that look like. We're kind of laying out the fundamentals here. Where do you start with people? Yes, I'd say chapter one is really focused around building the intuition about and the mental model about these different words that you might come across when learning about AI or machine learning or whatever.

The case might be, so kind of creating the taxonomy and this model of how things fit together, that's basically what chapter one is centered around. And that's important to kind of understand the concepts further or wrong in the book. And it's kind of written in a way where each chapter kind of has concepts that build on another chapter. So I might have mentioned that, you know, it's not really math heavy.

It's not to say that it doesn't contain any math. It does introduce the math to the reader, and you might find that something introduced in chapter two is useful in chapter three and so on and so forth. So that's the kind the kind of makeup of. It, right, So how do you start thinking about different algorithms, you know, because it seems like different ALGs rhythms are kind of so yeah, they're focused on specific types of problems or specific maybe solutions to a wide breadth of problems.

So how do you start thinking about the different algorithms and whether or not you need to know them. Yeah, So there's this concept of old AI and new AI. I'm not too sure if you're you're familiar with that, but old AI is essentially when humans, you know, or programmers would actually figure out what the rule should be for a specific algorithm. Examples of this are things like expert systems, where you have really complex rules, really complex rule set, and the rule set might even be a little bit dynamic, but you always understand what the inputs are going to be and what the possible outputs would be given some input, right, you know, So things like search algorithms that are spoken about in the are really useful there, and that's useful for when the problem is kind of well defined.

So if you're trying. To build a butt that plays chess, I know, chess has been a buzzword these days, with the Queen's Gambit being really popular on Netflix. So if you use that as an example, the rules of chess are well known. The problem that we're solving is looking into the future of all the possibilities of moves, right, Whereas if we're looking at say Spotify recommendations, you know, you're listening to different types of music.

I don't know about you, but I have really diverse kind of taste in music, and I'm pretty sure that algorithm struggles a little bit, but. It's still pretty good, you know. So the inputs there and outputs are I guess kind of well defined, but how you get there is a lot more complex. Does this person like the tempo of the music, do they like a specific artist.

They like the genre? Do you know? There's a lot more unknowns in that case, whereas when you compare it with the chess, the rules of the game are quite well understood. So I think, you know, when you do have a fairly known problem space that you're working in, a classical algorithm like search might be useful, whereas things like neural networks and reinforcement learning, you know, have been really powerful when you have a lot of unstructured data, unknown kind of rule sets, and you kind of want to find patterns in that data to inform future decisions.

Right, So that makes sense. So then what algorithms are available for that kind of a thing where you're looking for those patterns and picking up the parts that you're going to use to make those recommendations. Yeah, I mean that's I guess there's so many options, right, that's I guess part of the point of the book is to kind of inform the reader about what would be useful to use in what scenario. So, yeah, I've just pulled out at the back.

There's a quick summary of it. I don't know if you can read. That, no, but if that's your book, people can go get it and we'll have a link in the show notes. Yeah.

Yeah, So basically it's a bit of a guideline on what to use in what you know scenario. So, for example, I'm pretty sure you're familiar with linear regression. It's a very popular algorithm machine learning, and this is really useful to find correlations in data. So do you want to see if a bigger diamond means that it will sell for more money?

Right? Right? That seems pretty obvious. But what if you take into account the quality of the diamond, the clarity of it, the color of it, where it was sourced from.

You know, the same algorithm but adapted to deal with multiple parameters would be useful in that case. So yeah, I would say it's it's kind of based on how well you know the problem space and if you have data available to train. If you don't have data available, you should have rules. To govern it.

And yeah, the different approaches won't be applicable in those different cases. I gotcha. So yeah, you've got kind of a cheat sheet there as far as you know here the algorithms that you may want to use, you know, under these specific circumstances or with this kind of data or things like that, how do you start to understand how they work? Because I've had it explained to me by some people that some of these algorithms feel a little bit like a black box, right you put the data in, you know how it's set up, but you don't exactly know how it comes to the conclusions it comes to.

Yeah, so a lot of that is explained in these chapters. But one that I'm really passed from about and maybe maybe it'll be really interesting for you, is genetic algorithms. I don't know how much you know about that. But not a lot really.

If you think about the theory of evolution from Charles Darwin, basically what he says is every organism it's about survival of the fittest. So if if you're a fish, for example, if you can get more food, you will survive longer, which means you will reproduce more, which means you would it means your genes would be. Carried on further. And basically, you know, that's a kind of running theory of how we even evolved, and generally what happens is you have two parents.

Okay, so you have one individual that might be really good. Let's say let's say this individual is really good at swimming and another individual is really good at climbing. Theoretically, if you mix the two, you get a child that's as swimming and climbing, right, And that's essentially how junic algorithms work. They randomly generate solutions given certain constraints.

So you know, you might have a bit string that represents you know, in the book it's a knapsack problem, but it could be you know, a toy example could be the stock market. Do you buy, hold or sell and you have a bit string saying B for buy, h for hold s for selling. You have a bitstring for a certain timeframe, and you say, cool, this is a solution. Let's apply that to apple stock and see, you know, if it's going to make us money.

And you'll generate thousands of those solutions, and you evaluate them with something called a fitness function. And the case of that stock example I just gave, the fitness function would be, you know, after replaying that sequence of buying, selling, or holding, how much money did you make? Obviously, the more money, the more fit that individual is. Right, then what you do is you'll you'll kind of choose one of those good parents and another parent, you'll mix them and you'll make a child.

That's generally done by kind of taking the first half of the one string and the second half of the other string. And that's the general intuition of how genetic algorithms were. And the reason I wanted to mention that kind of metaphorical example is because that's basically how every chapter starts. It starts with this something you can relate to.

So in this case, there's a story about moths. You know, there's an actual actual proof of that where there was a there was no population of black moths. There was a lot of white moths at a certain point in time and industrialization, you know, things became black because of smoke right in the dustrial Revolution, and the moths actually over generations adapted and they changed color to become black so that they're camouflaged with the environment of bullidings and things. And that's an example of how kind of that concept introduced.

But what I really like to do as well is create flow charts of how the algorithm works. So rather than it being a black box or I'm pretty sure for a lot of people, a lot of the time a mathematical formula can look like a black box. You know, just looking at a complex formula, you know, a lot of the time you get kind of confused. I don't know what's happening.

But what I like to use is flow charts to show you what's happening at each stage. And there's also a pseudocode. So regardless of what kind of the programming you are, maybe whatever JavaScript, Python, C shot, JOVA, doesn't matter, you'd be able to kind of follow that flow chart and follow the sunocode to actually implement the algorithms yourself. Yeah.

Yeah, that makes sense. And I think that's really where things kind of broke down for me. Was just you know, I go copy an algorithm or go copy some library somewhere, right, and I'd pull it over into what I was working on and I'd get the result I wanted. But it was magic in between, right, And I like that idea of yeah, just putting it into a flow chart and just walking through it, I mean even on paper, right, and just kind of going okay, I'm I'm kind of getting the pieces here for what's you know, what's going on, and I can see step by step how it's adapting, right, and so yeah, it's it's got some waiting in there and some other things that it's doing.

But as as it kind of moves through it, it's like, Okay, you know this is selecting for this, that's selecting for that. And at the end of the day, you know, you wind up getting the right answer basically, or you know, you wind up getting a good answer in some cases, right, like the recommendation engine. I don't know if there's a right answer, but there are better answers than others, and so you know, I'm getting a good answer. I'm getting an answer that I'm looking for, and at the end of the day I could see in stages how it got there, right, Yeah.

Yeah, And I'm a very very visual person. I really like kind of representing things visually and related to what you're talking about. For example, in the neural networks chapter, you actually see each node in the neural network, each I guess hypothetical neuron change value, right, and you'll see exactly what numbers kind of got inputed and manipulated to change the value of that neuron, whereas typically when you're looking at a neural network, you give it some input, you give it an output, and the rest is magic.

What I try to do there is, you know, kind of give the reader window into what's happening with each and every element in that network. And let me be honest, most of the time, when we're building solutions, you know, in the real world, we very rarely need to go and understand those calculations. So what's happening at that detailed level. But I think that by understanding it to a certain degree, you have a level of confidence and also ability to kind of debug and investigate and figure out what's happening in there if things do go wrong or if you don't get the results you need.

And you know, something that's becoming quite I guess popular or prominent as we move forward is the kind of I guess legal and ethical concerns around machine learning and AI. You know, some I guess regulations or propose regulations want companies to be able to tell the customer or some authority, you know, what what did that algorithm do, what did that calculation do to come up with a decision. You can't just say, well, it's learned from data. They want you to be able to kind of reason about why it's making certain choices.

Example of that is I think there was a story about an organization that seemed to have unfair loan grant. You know, it was granting some people loans and other people it wasn't, but they were in you know, the parameters are quite similar. They couldn't kind of explain why that was happening. And you know that Spock's concerned.

If you don't know why it's happening, how can you make the decision? And I think if you know, a lot of the time is as developers or engineers or kind of people that are implementing tech, we often forget that we are, in some shape or form also responsible even partly for those outcomes. So the better we can prepare ourselves, the better we can understand what we're doing, the better we can make those decisions for whatever the outcome would be. Yeah, that makes sense.

I mean for me, if it is or isn't a good you know, for the loan for example, right, if it's going to be alone, that's going to be paid back on time, blah blah blah. You know, I get that you want some kind of qualification, but yeah, the computer told me no, isn't necessarily a good reason to give somebody. The other thing is is that I think we all have this innate idea around fairness, like what's fair and what's not, and so we always wonder a little bit, oh, well, is it because I'm from this place, or because I look this way, or because I, you know, I want to do something that's unconventional, and that's why it's telling me no, right, And so in those cases it may not be fair, right because somebody else who has slightly different circumstances than me that may or may not actually impact their ability to pay it back any better than I can, may actually get that loan.

And that's what people worry about there too, And so by having it this opaque system, it's hard to know that it's fair. And that's what I think we see a lot people kind of get hung up on, where in reality it probably is measuring ninety five percent of the stuff that's measuring is legitimately plays into whether or not. You know, historically other people have paid back their loans, and so it's this interesting balancing act. I think, just from the standpoint of yeah, you know, is this is this the right thing?

Is it doing the right thing? And these algorithms don't have a sense of morality, they don't have a sense of fairness, they don't have a sense of whatever right they work off of, you know, how you set them up and what you train them with. Yeah, I agree with that one hundred percent. I think, you know, especially with kind of enterprise at the moment, there's a little bit of an obsession with automation, you know, and we look at maybe some of the role models, you know, the giants like Netflix and Spotify and Google and Facebook and all those companies, and you know, they automate a lot of things, and I think there's a place for automation.

But I think, you know, for the example you just mentioned, you know, shouldn't that just come down to human judgment, Like let the computer give you an output, Okay, maybe not as as you know, black and white as yes or no. Maybe you know, give you you know, an indication or yeah, yeah, and let let a person go and look at the details and look at the application, go, you know, do what they need to do to make an informed decision. I mean, when you need a show recommended to you on Netflix.

The consequence of the recommendation being bad is it's very low. You know, maybe you waste fifteen minutes watching something that you didn't want to watch. R You know, there's much more severe consequences in other areas like finance or healthcare, and I think, you know, people got to be careful, especially the decision makers in these organizations. Like there is the obsession to automate things because we're going to streamline things are going to be more efficient, but that doesn't meaning any more effective.

You know, like you said, you can't simulate morality in a machine. Yeah, I think beyond morality, it's you know, I think we're pretty amazing. You know, as humans, we're able to put together these abstract concepts and string together these relationships and we can't explain what's going on in our heads, but we can reason about it that way. Whereas you know, the computer is going to do what you ask the computer to do, and if it has if it sees something new and it doesn't have data.

Let's say we're doing which you know, deep learning or something, it's sees something new, but it doesn't have a storic data that can inform what that means it's not going to know what to do, whereas you know generally we do yep. So I really like your approach to kind of breaking down these algorithms and then thinking about how they come into play, and you know, some of the ethics and morality around this stuff. But I want to kind of switch gears back to the algorithms themselves.

So we've talked about evolutionary algorithms right where it's okay, we take you know, kind of half of this approach and half of this other approach, or half of this information, half of this other information, and you know, kind of play it through the system and see how it works. And I'm pretty sure we've talked a lot about neural networks. That's something that it seems like a lot of people are working on right now. But you've got this swarm intelligence, You've got ants and particles and you mentioned at the beginning of the show, and they're in your book.

So how do those work? Like, how are those different from the evolutionary or neural networks? Cool? Yeah, let's talk about then and colony optimization, And what's really cool about that one is it's almost identically mapped to how ads operate in the wild.

So if you've ever seen a. Trail of ants walking in your kind of garden or on your patio or something like that, you might notice that they always keep this kind of single file trail that they're following, right, And why they do that is something called pheromones. So ants drop pheromones, and they have a variety of different pheromones. And pheromones are just you can see as perfumes.

They each have a different kind of scent, and an ant will drop a pheromone and another ant will will pick up on that pheromone. So the reason you see this trail of ants not to say that no ant walked any anywhere else. The reason they keep on that trail is because the pheromone intensity is the strongest on that trail, right. And what we've found through studies on ants is that ants actually find the shortest path between between a food source and you know, their their home or the colony.

So and they use this concept of pheromones. So imagine a few ants take two different routes to get you a food source and back. The ants on the shortest path actually end up leaving behind stronger pheromones and obviously the more ants on that trail, the more intense that that trail of pheromones becomes, and that becomes the shortest path between the food source and colony. Yeah, and essentially, the ant colony optimization algorithm is really great at solving shortest path problems.

I'm pretty sure you've you've you might have heard of the traveling salesperson problem where you know, a salesperson has to travel to one hundred different cities, but they have to minimize their their kind of total distance. Yeah, this algorithm is perfect for solving solving those kind of problems. It's also been used to solve telecom network routing problems, believe it or not. And basically, the algorithm works in a way where you simulate these ants you basically deploy and starting at different or going to different locations, and then they move to another location based on an element of randomness and the pheromone intensity on that path and the emergent behavior is that the shortest path between all of those different notes will become apparent after a certain number of iterations of the ants moving around.

So that's quite a fascinating algorithm that solves, you know, important problems in computer science and real world application. But it's really cool that it's inspired almost directly from watching real life ants work. Yep, that's fascinating. They're kind of I have Google Maps in my head from that, right, But yeah, yeah, there are there are other uses besides.

I guess you talked about the network, right, and that matters because you can get data back faster, right if you go the most direct route. Other places where this algorithms used, Yeah. I think it's generally most useful when you're trying to find the shortest path, whether that be you know, networks or other. Other problems.

Maybe you're a logistics company that needs to root your trucks, you know, to optimize fuel optimized deliveries. You know, the algorithm can be useful there. It's also been used in strange ways, for for example, edge detection of images. You know, you can manipulate an image based on the pixel intensity, and the pheromones along the edges of a picture might get I guess highlighted, And people have used that algorithm to yeah, do edge detection and images, So yeah, I get, but that's I guess that's a bit of an outlier.

There's much better ways to do edge detection or let's say more efficient place. Makes sense, what's the particle's algorithm? Yeah, a particle swarm optimization is based on you could say, the flocking and behavior of birds or bees. So essentially, when you have a very large solution space, right, So imagine that.

Well, the example I use there is you've got a drone. You're trying to make a drone, and you're trying to find the correct ratio of plastic and aluminium, okay, to use such that the drone can take off as quickly as possible, but it can also be resistant to the wind. Right, That means. It can't be too light, but it also can't be too heavy.

So there's this kind of interesting balance in the materials that you're using. And obviously, in this very simple example, we're discounting, you know, the aerodynamics of the design and set. So if you have that problem and you want to find these ratio, I guess the simplest way to do it is the brute force. Right, you try every possible ratio between the two.

But when that is a real number and you're dealing with you know, zero point one plastic and zero point three five aluminium, right, you know those levels. You're going to have to brute force that, and that's going to probably take a computer months or years, depending on the constraints you have. Right, So what a particle sworm optimization algorithm is. It puts these different particles at different outputs.

Okay, so at different points in that solution space. So one particle might be at zero point one and zero point five zero point one plastic zero point five aluminium. Another one might be at one point two plastic and two point five eliminium. Right, they get randomized in this space.

And basically how it works is similar to how birds actually flock. Is each particle tries to move itself to the center of the swarm. So if you imagine this three D space with these particles everywhere, right, and it would be a three D space with these two variables and whatever the output would be that each particle tries to get to the center of the swarm, but it also determines if its fitness is better than the swarm's fitness. Right, So every particle takes to consideration the result from the swarm.

So if I have a friend all the way down there, and you know he has a way better fitness to me, I'll move towards that particle. At the same time, I want to move towards the center of the entire swarm. And just based on those principles, the emergent behavior is that after many iterations, you find that the swarm finds a global optima, right, I mean, it does have this this risk of falling into a local optima. So we talk a little about local optima and global optima.

You know, there might be a good solution at one part of the search space, but a much better solution elsewhere, And that's where you kind of tweak these characters over. You know, how much should I favor my own fitness, how much should I favor the fitness of the swarm? And you know, how much should I favor kind of trying to get to the center of the swarm. But yeah, that one's also quite quite a fascinating algorithm.

It's used for optimization problems. So like, if you're I don't know, a factory that's trying to produce a new kind of biodegradable bodle, you know, and you want to experiment with different materials, that algorithm would be really useful. The critical thing about that algorithm is you need similar to the genetic algorithm, you need a good fitness function. So if you could simulate the chemistry of different kind of compounds, and there are programs that can do that.

You can use something like a PSO to design the perfect kind of biodegradable bottle that will also lost somebody, you know a decent amount of time, just as a hypothet right, Yeah, that's that's actually how birds when you see those those patterns of bird flocks, that's actually how they they work as well in principle. Interesting, that's really interesting. Well, I think we're kind of getting toward the end of our time. If people want to learn more, I guess they can go pick up your book from Manning and we'll put a link in the show notes.

I also have a link for Manning for I think thirty five percent off or coupon. Sorry, so we'll put that in the show notes as well. And then I think they actually sent me some codes I can give away. So if you go to dev chat, nott TV, slash Rock Cammel, then I will or grock AI because it's AI in the title Grock AI.

You can actually enter your email address for the giveaway and we'll give away five copies of the book. Rishall, are there other places that you'd like to send people though, to kind of level up on this stuff. Yeah, I'd say not necessarily places. But what I'd say is, if you're interested in getting into, you know, the kind of topic.

I know it's hot and trending at the moment. There's a lot of kind of courses and degrees and all sorts of things out there. But if you feel like it's kind of out of reach or it's not for you, I'd urge you to take a take a shot, Like doesn't have to be through my book, just go, you know, try something simple, go on to Cagle. Caggle has a lot of core resources to get you kind of up to speed at least with the basic concepts through some sample data sets.

But I guess the core message here is if you think you can't do it, maybe challenge yourself and try at least, Because something I've learned is that you know, when you let this kind of fictional story you make up in your head that something's way too difficult for you, then it's definitely not going to happen. But if you at least just try, you know, you can make it happen. Yep, absolutely, all right. Well, the last segment of the show is picks.

Picks are things that we just shout out about on the show. So, for example, a lot of people will pick like books or TV shows, they'll pick a particular thing that they learned. So it generally is things that are making your life better. Right, So I'll go ahead and throw out some picks just to kind of give you an idea of what we're talking about here.

So the first pick I'm going to throw out is a book. It's a nonfiction book. Most of the books that I read anymore are, but this is one that I've been listening to on audible. It's called The Hero with a Thousand Faces by Joseph Campbell, and it is a seventy year old book.

But what he does is he walks through basically kind of the epic journeys of the heroes in religious, mythological and cultural iconology, and he kind of gives you the formula for creating those same kinds of stories. And what's funny is is like the current state of the art sort of for movies and TV shows, they all follow this same story arc and it's funny because I've started, like, even in a TV episode, I can almost pick out, Okay, that guy's the culprit, right, you know, if it's a mystery, right, that's the murderer that And it's just kind of funny, you know, this is this is the breakthrough that the protagonist is going to have, you know, just listening to this book, I've kind of picked up some of that, and it's been kind of funny because, Yeah, a lot of these movies and TV shows are somewhat formulaic that way, right, you can kind of pick out the major elements.

You may not be able to call like exactly how the hero is going to break through, but you can usually pick out they're going to break through in somewhat this way, involving this person or that person, blah blah blah. Right, so you kind of get the idea. And some of these movies too, like they try and turn it on their heads on. One example is the Tenet movie that came out in September.

I want to say, we went and saw it, right, and so you anyway, they turned some of the tropes around, but at the end of the day, it's still followed the formula. So anyway, fascinating, fascinating books. I'm gonna pick that. And then the other thing I'm gonna pick is I have a smoker.

It looks like a mini fridge sitting out on my porch, but I have a smoker, and I just I love putting meat in it, and just, oh man, so good. I'm looking forward to smoking a couple of pork butts this weekend, and I just I freaking love the thing. So I'm going to pick a smoker. I'll put my particular model in the or something that looks very much like it if I can't find the exact one on Amazon, because I think we got at Walmart, but I'll put a link to it in the show notes as well, just so that if you're thinking, oh, well, that'd be nice.

I think the one I got was one hundred dollars on a Black Friday deal and they're usually like two or three hundred dollars. So I'll see if I can find a similar deal for you folks and let you know where you get it. But yeah, I just you know, put a rub on the meat, throw it in there, and after you know, I think the ones that I have the pork butts that I'm going to be running through like eight pounds, so I have to run them for like ten hours and then yeah, that's so good anyway, So that's my pick, Rachelle, what are your picks?

Yes, while you're talking your storytelling a book reminds me of another book with a very different purpose. It's more towards kind of marketing, and I guess communicating your brand or product or service. It's called building a story. Brand and oh so good, Donald Miller.

It's actually exactly what you describe, but using those elements to you know, engage with an audience, engage with people. So I guess that would be one pick related to your one so so good. I think, yeah. Other thing that's made my life better recently is I challenged myself to kind of do something in October.

You might know Inctober. It's a art kind of challenge that gives you prompts every day to draw something around those prompts. So they might say hope, and you think of something to draw about hope. And what I did is I read a different topic about the world, not necessarily tech related, but maybe co two emissions.

Maybe how's the fishing industry kind of evolved? Over time and for thirty one days, I basically just researched a completely different topic that I wouldn't normally go and learn about, and I kind of just wrote some doodle. About it. So you can actually check it out on my website urbans dot com.

But what I would say is, you know, if you could learn something new every day, it's quite an ask, or maybe even just practice something new every day or something different every day. Yeah, that's added a lot of value to me personally. So I don't know if that qualifies as a pick, but that would be that would be one of mine as well. Awesome.

All right, Well, if people want to connect with you online, where do they find you? Yeah? I think it's probably the best place. Is my website, just ourhurbans dot com or Twitter or LinkedIn.

I'm pretty sure if you type my name in you'll find me. I haven't. I haven't found another person with my name and surname, so. Should be pretty easy.

Nice. All right, Well, let's go ahead and wrap this up.