What was that trigger point, that moment when you did something with a computer and it triggered something in your mind that made you want to do this as a career option? Maybe you haven't hit that moment or maybe you're already in this career and you haven't really ever had that moment, but a lot of us have had a moment where we realized how cool or how powerful or how exciting this career can be. These tools, these things that we are building with every day, how exciting they can be. In today's episode, I'm hoping to provide a little bit of an in-road to machine learning and to data science in this third part of my interview with Chris Albin. I asked him to talk about the gateway drugs for getting into data science. So I'm really excited about it. I hope some of you become inspired by today's episode. And I really do believe that this field, this particular section of becoming a developer, this particular type of developer and the future is going to be extremely valuable. So definitely pay attention to some of these tools, pick them up, go check them out on Google, whatever it is that you do. My name is Jonathan Cutrell, you're listening to Developer Tea. My job on this show is to coach you through some of the hardest parts of your career and help you level up and become the great developer that you believe that you can be. Hopefully you do believe that. Hopefully you listen to the show because you believe you have strong potentially. You believe that the future of your career is headed on and up into the right kind of trajectory. And not only are you going to be increasing your earning power and increasing your influence, but you're also going to be hopefully becoming more and more satisfied with your day to day work. All of these things are good things. And sometimes one falls a little bit lower than the other, but overall the goal of this show and hopefully your goal align and pushing that graph up into the right. Hopefully you are becoming a little bit happier each and every day that you learn more, right? And as you learn more, you become more valuable to your employer and as you become more valuable to your employer, you end up making more money. Right? There's so many positive options to becoming better in your career. So thank you so much for joining me on this show. I hope that it is inspiring. I hope that you will reach out to me with your questions that you'll actually take to heart some of the things we say on this show, but more than that, I hope you learn to develop your own opinions and test your theories and really get into the day to day getting better as a developer. That's what my hope is for you. Thank you again for listening to today's episode of Develop Tea. I'm going to get out of the way and get straight into the last, the third and last part of my interview with Chris Albin. If you haven't listened to the first or second parts, then you may want to go back and listen to those first. Let's get to the interview with Chris Albin. I think about the race conversation specifically. We have a long history in this country, in the United States of race being incredibly important and it modifying how you would otherwise experience life if you didn't have race as a part of it. So, you know, treading relatively lightly here because I want to be extremely careful in these kinds of conversations. But for example, you know, if you were to look at 1960s America and use only statistics to make decisions, you will make racist decisions. Like it's going, and it's because of some other state. It's not some intrinsic value, but perhaps it's culturally imbued. So, you have to take these values that you have as a group of people, whatever group is going to be participating with these machines and also teach the machine the values. Right? So, for example, if you don't want to be racist, well, now you have to talk about how to teach the machine that racism is wrong. And when you're talking about validating loans, you know, and this is why the conversation about ethics around this is so important. We're talking about making decisions that could perpetuate the same scenarios that it's measuring. Right? So, you look at, is this person more likely to pay their loan off or not? Well, if they're not, then we're not going to give them a loan and we may perpetuate the same problem that caused them to not be able to, that caused the statistics in the first place, right? So, you know, we had to adopt these, the programs, for example, in a group of people we adopt programs to try to reverse statistics or otherwise, you know, affirmatively change them because we have some values that go against what we're measuring. Yeah. No, I mean, it's, you know, it's one of those things where I've noticed this increasingly where there'll be some news article about an algorithm that, you know, like an algorithm that discriminates or an algorithm that is racist or something like that. And like, the key point that like we should all keep in mind in these conversations is it's not the algorithm, it's not the model. Like, the model is like, the model is just a computer program that everyone listening to this podcast runs. Like, it just runs and it just does the thing that we tell it to do. And like, when, you know, the difference is that with machine learning, we're telling it to learn. But like, we're telling it the data that we're putting into it. We're telling it like what we, like, I mean, the big thing is like called a cost function, which is like the thing that we're trying to optimize for. Like, for example, a cost function might be the difference between your real salary and my, the salary that I predict for you and want that number to be as small as possible. And that's what we're working towards. And like, well, we shouldn't, you know, like, I think as these conversations go forward, there's really like a conversation has to be had around, okay, well, let's have a cost function that takes into the societal values that we care about. And I don't know exactly how that looks like. And this is why people who are smarter than me are doing this. But there is this kind of point of like, well, no, no, we should, we should include those kind of things like things we want, you know, our company culture to have and things we want our society to have. And like, okay, we should include those into the cost function and it'll learn it. Like, it's, it's the algorithm's not racist. The algorithm does whatever we say. Right. And so we should, you know, in, in view that with what it should care about and what it should learn towards should be the values that we care about. So that's, you know, and again, it's like, we're just getting started in this. Yeah. Yeah. We're talking, we're talking 10 years down the road now and it's stuff that we have to kind of prepare for at least mentally and kind of walk towards. And for people who are going into this field, there are people who listening to this podcast and whether or not you know it, you're probably going to work in this field. There's, I mean, statistically speaking, it's going to become a very large field. Yeah. And I mean, I think like, and any kind of developer shop, I would suspect that there'll be a data scientist or now they're sort of having different titles, but like a machine learning engineer or something who like is there, right? Or a small team of them who works like just like how you have front and devs and back and devs, you'll probably have a machine learning person who has that expertise and, you know, he's, he's using it, applying it to certain parts of the product. Yeah. I totally agree with that. I'm going to, I'm going to shift a little bit here and ask you a question that I intended to ask earlier, but now we've gotten, we're like an hour in. So we are probably going to break this up into three parts. So maybe this is going to be towards the beginning of the third part. But the question in this one, they ask every guest that comes on the shows, what do you wish more people would ask you about? Wow. I guess, I mean, I think the thing that I wish people would more ask you about is like, how can I get into machine learning and artificial intelligence? Because it has a stereotype around it that you have to be a some kind of like math genius. Like I have a PhD and people think, oh, I need to have a PhD in order to do it. And like the truth is that you don't. It's a skill. Like I'm not a math genius. I'm just a person who's done this for a while. And in another, you know, in another world, like I started out doing like web design front and dev. Like I could be a front and dev. I just took a different path and like read different books and took different classes and took different jobs. And now I'm here. And it's a skill and it's a big skill and it's big and it's complicated and it takes like a while to learn. But like it's you, you know, anybody, anybody can learn it. It's just if you go down that path. And like I just, it just seems like a huge disservice to me that the idea that you have to be, you know, a Stanford computer science PhD in order to like have any kind of place in that. And those people have a place, but like they're at the bleeding edge doing the research. But you know, probably like 98% of the job growth in this area and there is and will continue to be a massive growth in this area. It's going to be people applying stuff to, you know, regular everyday concerns, right? It's like, you know, you're not going to be inventing a brand new front end. I mean, maybe you will, but like you're probably not going to be inventing a brand new front end framework. No, no, no, you're going to be like really good at applying some frameworks to problems. And when clients come to you or, you know, customers come to you and they want something, you can apply it and do it well and give it back to them. Like that's what we're going to see. And those people take, you know, you have to study it, you have to learn it and you have to spend time on it. But you're not a, you're not, you don't have to be a math genius. You have to, you have to learn a lot of math. You don't have to be the math wizard. No, you know, like I'm not whatever goodwill hunting. Yeah, exactly. Yeah. You don't need to be the one. So I think the distinction that's so important to make in this, and I agree with you 100 percent. And we're going to actually, I'm going to give you the floor here in a minute to kind of what I said earlier is get excited about it. What I really wanted to say was give us some gateway drugs, some stuff. I really get as interested. And I think a misconception that people have is that you have to be the one that is, you know, engineering, the learning, the way that a machine learns. And actually, that's not true. Like there are tons of algorithms that have been developed already. There's tons of ways of doing this that are still completely untapped. Like they, they aren't done yet. And, you know, very simple things like you're saying. And I say simple, but very simple concepts at least. Like, for example, you know, a neural network, you can implement that stuff with an API like what you're used to in your code for the most part. And you don't have to gain like some massive level of intuition for what it's doing. You can actually, you know, read a couple documents and start playing around with this stuff in a couple hours. Yeah. So with that said, I'm going to turn it over to you to give me just, if I were coming into this and I asked, okay, how do I, you know, how do I, I'm a brand, I'm a developer, I know Ruby and a little bit of Python and some JavaScript. And I want to take this, I'm going to take it to the next level, understand a little bit about machine learning. What do I do next? So I think the big one is if you have any kind of Python experience, just even just a little bit. Like go head over to Scikit learn, which is a library that's used for machine learning up to neural networks. So it doesn't really do neural networks, but it does everything else. So it does like, you know, like random forests and all like, basically the sort of bread and butter of what data scientists do. It does all those things. And the reason that you should totally go there is that their documentation is like a MOOC. Like it is so good. It's so detailed. They've spent a huge amount of time on it. It's both code and theory and math all totally together. And they have all these great examples. And like I'm telling you, just get to the point where you can see the, because what you'll do in machine learning is you sort of, you hide some data, like some observations from the learner. And then you make, you have the learner make predictions, like say Apple or not. And then you compare it to the ones that you've hidden and you can sort of see how accurate it was. So like in reality, this was an Apple and you said it was an Apple. Yay, you like, etc, etc, etc. And as soon as you get to that point, like I feel like the world will open up to you and you can see all this cool stuff you could do. And it's right there with code. It works like any other library, you know, like you choose a function and that kind of stuff totally, totally normal. The other one, the awesome gateway drug is this guy called Andrew Ning, who's a very famous data scientist and he has a Coursera course and I think his videos are on YouTube, but he is Andrew Ning's Coursera course. It starts with basically no math, no matrix algebra, no calculus and he steps you all the way through and like I think it's like 150 videos in each video is like five or six minutes long. Maybe it's 100 videos, but like he steps you through a complete course on machine learning all the way up to more advanced neural networks and he assumes nothing from you. He doesn't assume you can do calculus, he doesn't assume you know what linear algebra is, he explains it to you, right? He works through that and shows you the basics just enough linear algebra to have the rest of it make sense. And it is like the de facto gateway drug for getting into machine learning is this one one Coursera course called Andrew Ning. And it's great and he's an amazing teacher. They're great videos. I think I listened to them like just a few weeks ago, so I'm clearly like going back and doing it, doing a reintroduction again. But there, I mean, I did it because so many people were talking about it and to so many people that, that set of videos was their, their introduction to it. And as soon as they did the introduction, all these like crazy concepts they were hearing about machine learning, they were became concrete. They were like, oh, I know what that is. Of course, cool. That means I could do this. Oh, I wonder if instead of using this example data, I put in my data and like that kind of stuff. Yeah. I like it learn, Andrew Ning's course, if you're in Ruby, I'm sure there's a machine learning Ruby library, but it'll be nothing like scikit learn. Like really the machine learning world is on Python is on Python and there's like scallop you want to do some stuff. And then there's R, which is a statistical programming language, which a lot of people might not have heard. Here, because it's not really like a developer language. But like really Python is like the place to go. It's sort of the Python is like the center of this world. Yeah. Between data science and statistics and building products. So. And I've heard that, I've heard that for more than just you, Chris, I've seen Python kind of be the way of because you're coming over from something like Django where you're, you know, it's NBC and you're building a web app. And then somebody says, hey, did you know you could control a robot with that same language? And it's, it's not super exciting. And I think that there's some, there's certainly some growth in JavaScript. I mean, JavaScript is growing like crazy in pretty much every direction. But JavaScript in machine learning is also growing as well. Because it is a generalized language. And I would imagine, you know, give it a, give it a year or so. And a lot of languages are going to have something in this space. But certainly Python being that kind of brought over from the data science world. And I don't know why that is. It may be because the people who chose to do this just really did it very well. And it kind of took off and, you know, did the Pareto thing where it really captured most of the market right away. Yeah, I mean, it's still a weird because like my entry into Python was through scientific programming, which is basically all Python, unless you're dealing with huge, you know, like astronomical data, like literally from astronomy. It's there, there is a really big scientific programming community. And then like, I think that's where the idea of starting it on like you building, things I could learn first happened. But like the reason that I really enjoy psychic learn is because it's mature. Like it's not. Yeah. It is, it's like, you know, it's done a bunch of rounds. People have expanded the docs in a huge way. There's tons of examples. There's tons of stack over for questions. There's tons of tutorials out there. Like you are not on like if you do a lot of deep learning stuff, you're really forced to sort of be on your own and think about the problem for like the first human being to ever have thought about doing this before and trying to like work it out from there. That's not the case for psychic learn. You are well-trot ground. It is, it is literally just like any other library that you've ever used with massive documentation. I think a really cool example of this I found on your site. Something really simple and kind of fun is like a, hey, here's what, here's something you can do. If you're just like kind of mildly interested in this stuff, can you talk about the, the aisle seat article that you wrote? I love this. Yeah. So, in my previous job when I worked for this nonprofit, I was flying all around the world. That was sort of my job was to like fly and do work in different places. And one of the things that I want, I always want, is an aisle seat. I just want an aisle seat. I want to be able to get up and go to the bathroom and then come back and not have to like bother people and that kind of stuff. And a lot of times I would get a plane ticket and I would say the seat number, but like not the plane type and that kind of stuff. It was just like, I was like 12 C and I started to like really become fascinated like just in my head just trying to do it manually. Like what is the chance that this 12 C is an aisle seat? Like what's the probability of it being an aisle seat? Because I was like, should I go up to the front desk and change my ticket or am I just okay and I can, you know, go get a beer and they come back later. And like, so I started out with like building out a set of airplane aisle configurations. So like seat seat aisle seat seat, you know, like that kind of stuff. And I built out a bunch of them and then I calculated the probability of any particular like a like seat, you know, 12 a versus 12 V versus 12 C versus Del D, etc, etc, etc. And the probability of it being an aisle seat or not. And it turns out that C, I think C is the best one. C is like a really high chance of being an aisle seat. You know, it like, whether it's because if it's a two seats, you have two seats, aisle two seats, you're still an aisle seat. If it's three seats, aisle three seats, you're still an aisle seat, that kind of stuff. So it's, it was, you know, it was just a personal thing. But it is like, you know, it's trying to be, it's trying to think through the problem and kind of with some kind of solution that is useful to me, but also somewhat intellectually interesting. Well, it's kind of like, I mean, there are so many examples in your personal life that once you understand how this stuff works, you start seeing it come alive. Like it becomes, it becomes a whole lot more than numbers. And I think that's kind of the magic of it, right? So, you know, stupid, simple stuff like, when should I go to the grocery store? Well, if you know what days, most of the things that you typically buy are on sale, in grocery stores have relatively predictable schedules for what they put on sale and when. So when you know about what your schedule is in terms of how long it takes you to go through something, and then when that thing is on sale, now you can kind of calculate that. Well, you can do it a whole lot easier with a computer, right? Like, it's a little bit harder to do it through intuition and just like trying to remember it. And the crazy thing is, not only can you calculate that with a computer, but then you can automatically add it to your calendar with a computer, you know, it's, and so you're connecting this fundamental thing of math and this data-driven decision making. And then going to something really simple, and now, for example, you probably, if you don't know what kind of plane it is, you pick C because it's the most likely scenario that you're going to get an aisle seat. Yeah, I think, I mean, I think those are the kind of stuff that's really been intriguing to me, like, I keep on going back to this, but like, I'm really interested in how machine learning algorithms work and how to make them better and how to do that kind of stuff. And it is totally not because I'm in love with the, you know, like with the math of machine learning. Like, I'm not a mathematician. I was never in love with math. I'm totally in love with what you can do with it. Like, that's sort of the thing. Like, the idea that I could make these, these predictions and they could be accurate, and I could say something with confidence that I couldn't say before is like, it's such an amazing feeling. Like it feels like you're launching your app every single time you do it. Like, it's just so nice to see that. And that's what made me addicted to it. Like, it's not that I love linear algebra. It's that I love what I can do with it. Yeah. And the idea of pure mathematics, like, is so dull to me. Like I'm really glad other people care about it. I really do because they can do all the hard stuff. But like, I don't care about pure math. But I love making predictions about things. And so everything I do in my life can, you know, eventually get smarter and I need to care less about these things. And like, it's not about having R2D2, but it is about having like my fridge know when I'm out of milk. And it doesn't need to tell me that I'm out of milk. Like, I don't want to notification on my phone that I'm out of milk. I just want more milk to arrive. Like, and all that kind of stuff. And that's what I'm, you know, like that's what I think a lot of us are working towards. One of the tools we talk about using in today's episode is Python. We talk about Psychit and we talk about Python. And what platform better to run Python on than Linux, right? Linux and Python go very well together. Well, if you want to run Python in the cloud, then you can't just hook up your Linux box in your house, right? You're going to want a server. The solution to your Linux server problems is linode. The code provides you with a Linux server for as little as $5 a month. That will give you a gigabyte of RAM. Of course, you can scale up from there. They have high RAM plans. They announced this high RAM plan recently this year. In fact, you can get 16 gigs of RAM on a server on a Linux server in the cloud for $60 a month. That's a fantastic deal, especially if you're doing a lot of this data crunching. Of course, all of their plans are hourly. They have a monthly cap on all their plans. You can get a server running in under a minute. And the reason why we can talk about linode contextually on almost every single episode, by the way, is because Linux is such a powerful and all-purpose kind of tool. And having a server in the cloud, there's so many use cases for it. And it's applicable to today's episode. And it's going to be applicable to so many other episodes of this show. And when we aren't talking about specific things, like, for example, Python, it is applicable because there's so many things you can do that you can spin up your own personal site. You can create a data science learning environment that you can access from anywhere. You can have full control over this server. By the way, it has native SSD storage. It's built on an internal network that is 40 gigabits, right? Very fast, Intel E5 processors, really fantastic product that Leno provides highly available servers much better than the Linux box that you have at home to crunch your numbers on. So go and check it out, spec.fm slash linode. Now on top of all of this, if you've been a listener of Developer Tea, you know this. Now, you get a $20 credit just for being a Developer Tea listener, all right? So use the code Developer Tea 2017, that's Developer Tea 2017, spec.fm slash linode. Thank you again to linode for sponsoring today's episode of Developer Tea. If you think about it and we've talked about, we've said this word a few times and it may be, I want to clarify the word and explain why it's so important in these discussions. We said the word prediction over and over. And I think a lot of people, when they hear prediction, they immediately think, oh, it's you know, like they're going to try to predict the stock market or, you know, they're predicting the future, right? Really what prediction is about is about assistance. So I'm going to walk this out for a second. When you predict, for example, what seats are most likely to be aisle seats, at least with the airplanes that are on the market today, right? When you predict with a reasonable level of accuracy, all you're saying is you're describing the data, right? You're giving some assistance to the most likely outcome. So for a person who is making that decision, you are predicting the outcome of the decision. So they can be more informed about what decision to make and that's assistance. So you're, if you are in this space and you hear the word prediction, immediately think assistance, right? Especially when it comes to this optimistic view, at least, of hopefully we're being assisted, right? But even if it's the advertiser that you are assisting, you're assisting them with showing the most likely to convert ad, right? This concept of prediction, I think, turns people off because they don't understand the whole context of what prediction means. And it's not just about predicting, you know, in the traditional sense of prediction. Yeah, I think, I mean, you know, like the thing about advertising is that it like advertisers want to show you as that you consider content. Like, that's the real dream goal that like they would show you a bunch of ads and you would think that they were interesting and useful and like as good as a Buzzfeed article. And clearly they fall short because we've all seen ads. But like, that's the goal. And then, you know, when an airline, for example, predicts that their plane is going to be late. And so they email you and give you a voice message hours before the plane is actually late because they predict it's going to be late and they don't want you to come to the airport. They say, hey, hold off for four hours, which they've been started to do recently. They're trying to make that experience better for you. Now, of course, they're doing it because they want their product to be better. So you stay with them. But the idea is that you're like, if you like their product better as in flying with them because you're pleased that you weren't sitting in an airport for three hours, you just left your house three hours later, you're going to enjoy your time with them and you're going to stay with them. And like, those are those, you know, that they're predicting stuff. They're trying to predict what the plane is late. They're trying to predict what it's going to rain. They're trying to predict which products you want. And, you know, it's the whole idea behind this is just to make everything nicer for individuals. And in the case of business, they're customers. But in the case of, you know, like social good, which is what I used to work in before, like it's about predicting, you know, like refugee flows or anything like that. And there's just lots of areas in there. But it is, it's trying to sort of take, you know, some set of data and just adding a little extra piece to it, right? Like that, that one more step, which is like here it is and saying, and we have another observation that we haven't looked at before. Is it red or is it blue? And like, we look at all the previous data and we figure it out and then we say, okay, it's probably blue. And like that kind of stuff is, you know, sort of the bread and butter of what we do. But it, the whole idea is to make people's experience better because if people's experience isn't better, you know, like, well, that's, we're probably doing it wrong. Yeah, I totally agree with that. And ultimately, you know, you look at it from the perspective of the advertiser, but eventually, down the road, we're talking about assisting with life, right? Assisting you with your milk problem is a perfect example of this, right? The simple things are actually really important in this space. So definitely agree with that. So are there any other gateway drugs, any particular things that you've found to be kind of like eye-opening or that first experience was so memorable for you? Yeah, I mean, I think, you know, the big one, I, you know, is going out and seeing a prediction happen for yourself, right? So like seeing a model that's trained and that you give it something and then it reports back what it thinks it is. And there's actually like a lot of sort of web apps out there that do that for you, just like as a demonstration of different machine learning algorithms and stuff like that. But like, that's, to me, like, I, I, that's sort of, that was totally my gateway drug of just, you know, training a model and then adding some observation that I'd never seen before. Like, you know, he's a man. He's 45. He's driven for two years. And then it's like prediction, you know, like says something about him and it's like, wow, like I, I, like, I don't, I'd never knew you could do that before. And in machine learning, what's so interesting is that, you know, because you have, because you hide data from the learner to evaluate how well it does, you can actually see the truth and compare it with what your prediction was. And like, when you see how close they are, it's like, it feels like magic. Like, yeah, you predicted my salary was $46,000 a year. And really, it's like, you know, like 48, like, whoa, like how, how did you do that? That is, that, that to me is the best part. It's like those, those ads that you see at the bottom of, you know, bad news sites. And they're, they're, you clickbait ads and they're like, enter your name and wait 30 seconds and be blown away. This is like that except for real. It's kind of a crazy thing. Now, it's, it's, you know, it's being able to see stuff that we weren't able to see before. Yeah. It's, it's a great, like, it's, it's awesome. And like, you know, as, as someone who's used technology their whole life, it's a, it's something that I like never really thought about growing up. Like, I always just assumed that we would hard code everything. Yeah. And the idea that we don't have to do that and like, it can actually work at scale and just, you know, crazy stuff like that is, is awesome. Yeah. Absolutely. Chris, this has been fantastic. A really enjoyable discussion. Let's, let's take a minute and talk about partially derivative. If people are listening to the show and they enjoyed some of the topics we're talking about here, would you say that partially derivative is more or less like this, this episode? So partially derivative is, you know, how we describe it is, it is the bar after a data science or machine learning conference. So you go, and like, you know, there's all these great learning resources out there, like Andrew Nings thing and you watch the lecture and then afterwards, you're at a bar with all your friends and people are talking about the companies in the space and the topics in the space and their own, you know, like issues around imposter syndrome and their own pathways for learning and it's casual and there's only drinking on the podcast and, you know, there's just like a sort of the, the sort of guards are down and so like on our podcast, we don't have people who pitch products, we, you know, like we had Microsoft on, we had some guys from Microsoft and like I got them to like explain their product is like, you know, swear word, awesome data lakes, like it's not, you know, like everyone here is smart world developers, we don't know what's happening and, you know, it's not about pitching products, it's just about hanging out with friends and talking about data and being super casual about it and our goal is not that you learn anything from it, our goal is that you just enjoy yourself and laugh like if I could measure every single time someone laughed out loud, that would be my metric, but I can't so that's fine. But like we're trying to make it accessible to people that we are talking about things like deep learning and neural networks, but we're not describing the details of it. There's no math involved, you know, it's, it's just us like having examples of doing it and humorous examples or thing we really rely on to just, you know, show how these things can work and not work and all that kind of stuff. That's fantastic. I've enjoyed every minute of it that I've listened to you. So I would definitely recommend it. And it's subscribing just like anything else, any other podcast I suppose in a, you found you on overcast pretty easily. Are you anywhere else that people can find you easily? I think, yeah, as far as I know, we are everywhere because we did that a long time ago. So like iTunes and overcast and their sound cloud and stitcher and all those and I think we're just, I think we're everywhere. It's the way to put it. Cool. Great. Yeah, highly recommend checking that out. Partially derivative is the name of the podcast. Partially derivative. The math joke. And then of course, Chris Albon is Albon, am I saying that correctly? You know, I don't know what it is. Albon Albon, whatever. It's fine. It's whatever somebody says and you burst out. Chris Albon.com. If you want to, there's 400. I saw it at the bottom in the footer. 495 pages on your website, by the way. So it's really quite a lot of content there. There is. Yeah, that's, I was a few years of me riding tutorials. Yeah, that's great. Things like, for example, the airplane seat article, you can find at Chris Albon.com. Chris, have one more question for you. If you could give 30 seconds of advice to developers, what would you tell them? Are developers wanting to do machine learning or just developers in general? You can do both. We'll get a whole minute of advice out of you. All right. Well, I'll do one that'll be combined for both. So if I was going to give 30 seconds of advice, it would be, don't think about machine learning as some advanced, you know, algorithm or set of algorithms that you need a PhD for it or you need to hire some with a PhD for there are areas of machine learning that are like that, but the bread and butter of machine learning is things like scikit learn and very understandable algorithms that you can literally use like any other library and you can go there and you could be using it and coding it in, you know, 10 minutes from now, make your first set of predictions and two days from now, you could have it in your product. That's fantastic. Two days, can you imagine implementing something that is smarter than you could and always working to get smarter? That's such a cool thing. Go and check out scikit learn for sure and thank you, Chris, for all of your kind of lowering the bar of expectation for some people and really without devaluing the practice, I think that's so important is it's not as hard to get into as you think it is, but it's more valuable also than you think it is. Absolutely. I think helping the cause, the greater cause of machine learning and AI. Chris, thank you again for coming on the show. Wow, no, it was absolutely a pleasure. Thanks so much for listening to today's episode of Developer Tea. Make sure you go and subscribe to partially derivative a fantastic show about all things data science and it's incredibly approachable. If you are just getting into this stuff, you can listen to this without feeling like you're in over your head. Go and check it out again, partially derivative and whatever podcasting app you use and while you're in there, go ahead and subscribe to Developer Teaas well if you haven't already done that. That'll make sure you don't miss out on future episodes of this show. Thank you again to Linode for sponsoring today's episode of Developer Tea. You can run your scikit, experiments and put them up on a Django app all day long on Linode. Go and check it out. Spec that of them. Slash Linode. Don't forget the code Developer Tea2017 for $20 worth of credit and check out. Thank you again to Linode for sponsoring. And now to the most important part of this show, you, thank you so much for listening to today's episode of Developer Tea. This show wouldn't exist without your listenership. So thank you so much for listening. My name is Jonathan Cutrell and until next time, enjoy your tea.