his is episode six, Certificates and Degrees. So this podcast episode is gonna be a lot fluffier than the last. I realize that the last episode was really heavy, really, really heavy mathematics and algorithm stuff. And I wanna give you a break from that episode before we move on to logistic regression. I may do this from time to time in between very technical episodes, try to punctuate them with fluff so you get a little bit of a breather. This episode, I wanna explore what are the options for you as somebody pursuing machine learning, whether you want to get into the industry professionally or you want to go down the deep rabbit hole as a scientist and crack the mysteries of artificial intelligence and consciousness, or whether you're just doing this as a hobbyist who's curious about the whole thing, working nights and weekends maybe just on side projects. And one reason that I want to address this right now is that in the previous episode in the resources section, I recommended that you all start on the Andrew Ng Coursera course. And while that's not a very difficult course, it is a bit of a time commitment. It's something like a 12 week course. Like I said, self-paced, you could probably finish it in three to four weeks if you dedicate yourself. But some people may be wondering if I may want to go back for a master's degree in machine learning or do one of these Udacity nano degrees, should I take the Andrew Ng Coursera course now, or would it be a waste of time since I'm gonna be learning that stuff from scratch anyway? So let's approach the types of people learning machine learning so we can answer this question. The first type of person would just be doing this for self edification. They just are playing around with machine learning as a hobby for side projects and for fun. I think using my podcast as a curriculum or a syllabus to help guide you in your own self-motivated education would be plenty sufficient for that kind of a person. In the resources section of every episode, I'm going to be providing you sequentially the best resources for learning the details. So early on in the series, like I said, I recommended the Andrew Ng Coursera course. That's one that's gonna keep you busy for the next four to six weeks. Highly, highly recommended resource. And then towards the end of my podcast series, I'm gonna be recommending things like artificial intelligence, a modern approach textbook and the deep learning book. I think for the self learners, coming up with your own curriculum or letting me guide your curriculum would be plenty sufficient. Then there's people who wanna tap into the industry professionally. And what I'm gonna actually speak to right now is a middle ground between those two types of people. On the one hand, we have people who are just doing this for fun. And then in the next rung, we have people who wanted to tap into the industry and they wanna have a very strong resume. Somewhere right in the middle, we enter this world of what's called MOOCs. M-O-O-C-S, it stands for Massive Open Online Courses. And there are some major contenders in this field. One is Coursera, another is edX, Khan Academy and Udacity. So let's talk about these one by one. Khan Academy is actually for high schoolers and kind of AP students or very early college. So the courses that are offered on Khan Academy are things like calculus, statistics and linear algebra would actually be very useful for machine learning education. But I don't think that they're gonna have anything specifically for machine learning. So it's all the prerequisite stuff. In a future episode that I'm gonna have on math, the various branches of mathematics that are recommended prerequisites for learning machine learning, I'm going to make a shout out to Khan Academy. But they're not useful for the purposes of this discussion. So we move on to the next one, edX. Now, I see a lot of discussions on Reddit and Hacker News when people are talking about whether these MOOCs are worth their salt. And unfortunately, edX is never one that kind of makes the cut. It's never really a contender in these conversations. And whether or not this is a good approach, given that data, I'm gonna use that as a telling bit of evidence that maybe edX is not one of the better platforms for learning data science or machine learning. So we're just gonna discard edX. We're just gonna put them out of the equation. The two contenders that I see in MOOCs most commonly are Coursera and Udacity. Coursera tends to be more recommended for one-off courses. They do offer these nanodegree competitors certificates called specializations, Coursera specializations, such as the specialization in data science, in machine learning, et cetera. But I find that people don't gravitate towards those or recommend them a lot on social media. Coursera is fantastic for one-off courses like the Andrew Ng Coursera course. So you'll see Coursera courses recommended from time to time, but they tend to be one-off courses. I think in my own experience looking at Coursera, it's a little bit less professional, has a little bit less industry backing and post-production by comparison to Udacity. So that brings us to Udacity. Udacity has one-off courses as well, free courses that you could take, very good ones, like the deep learning course that was put out by Google. But they also have this thing called nanodegrees. They're certificates that you can earn online. You pay $200 a month. I think they tend to be about one to two year programs, these nanodegrees. And there's some sort of stipulation like if you finish it in due time, then you get half back. These nanodegrees are very professionally put together video series with mentors, assignments, programming assignments and class projects and solid peer-to-peer interaction. I absolutely think that Udacity is the future of online education and education in general. I really think that MOOCs are going to be revolutionary in the education space. I really hope that MOOCs eventually kind of overtake the requirement for accredited university degrees because these are scalable, professionally made nanodegrees and certificates put together by the best of the best in the industry like Andrew Ng and Sebastian Thrun. But however, at present, these nanodegrees are not widely accepted, recognized or respected by employers. Now this is hearsay. This is my own opinion. You may want to get a second opinion about this. And I do plan to revisit this podcast episode in the future. Once I've learned a little bit more about the space, I plan to come back and redo this episode. But from my own findings, from talking to recruiters personally, from reading posts by recruiters and hiring managers on various social media websites, it seems to me that Udacity nanodegrees are not respected in industry yet. Master's degrees are, of course, accredited degrees that are respected by corporations. But Udacity nanodegrees are not. What they do do for the candidate who's applying to a job is they provide you with your portfolio of projects that you have built for that class. And they prove that you have the motivation and self-drive to take an online course and the tenacity to finish an online course. So I have seen hiring managers and recruiters say that they very much respect the initiative shown by people who take these courses. And the side projects that they've built during the course of these nanodegrees. But that the nanodegree itself is not respected. In fact, when it comes to that portfolio that I just mentioned, what I see hiring managers and recruiters say time and time again is that portfolio, portfolio, portfolio. Portfolio is the most essential ingredient to getting a job in industry. Having a strong portfolio of projects that you've built nights and weekends, not just little dinky things that can recognize a cat from an image, but something that scales, something that you've deployed to AWS, scaled it horizontally, and shows that you have skin in the game in developing large scale machine learning projects. So that's why I put Udacity kind of in between somebody who's a hobbyist and somebody who wants to get a job. Because it appeals to the hobbyist because you're gonna be learning all the good stuff that you wanna learn from these nanodegrees. And it helps you get a job in the sense that it shows that you're self-motivated and you have a strong portfolio, but the nanodegree itself doesn't help. So what I would recommend for somebody who doesn't have the money or for, and for people actually listening to my podcast, if you're already listening to my podcast, it shows that you have self-motivation enough that you could follow up with the resources that I give at the end of each episode, sort of a homemade curriculum or syllabus for you to go by. You can teach yourself everything you need to know and you can work on your own side projects. So I'm gonna leave it up to you whether or not you want to pursue a nanodegree by way of Udacity, or if you just wanna start working on side projects on your own and learning everything you need to know from these books and videos. Me personally, I don't plan on getting a Udacity nanodegree. Okay, so now we're in that middle category of people who are doing this because they want a job. Like I said, the portfolio is the most important piece for getting a job in machine learning from talking to recruiters and hiring managers. This is sort of the way that web development is in today's generation. In web development, you'll see a bachelor's of computer science listed as a minimum requirement on a job description, but applicants with a strong portfolio can kind of bypass that step. I often kind of view that bachelor's degree requirement on web development job descriptions as a suggestion. I do have a bachelor's in computer science, so I don't know how much that helps, but my education is never talked about in an interview, only my past projects. Similarly, I've heard from people that the master's degree requirement on a job description for a machine learning job can sort of be bypassed if you have a very strong portfolio. So focus on that portfolio, but if you do want that peace of mind and extra edge, then a master's degree is an actually accredited degree by comparison to Udacity Nanodegree, which is simply a certificate, and will get you in many doors that you wouldn't otherwise get into. Now, a master's degree can be very expensive, $40,000 to $60,000 at some universities, but there is one master's degree out there that is online, completely online. It is by Georgia Tech. It is called OMSCS, Online Master's of Science in Computer Science, and it is $7,000. So it is substantially cheaper than your typical master's degree. The way I understand it works, I'm not entirely certain. I believe they use Udacity courses in lieu of certain classes. It may just be even a Udacity Nanodegree disguised as an accredited master's degree. I'm not sure, but I do know that they are partnered with Udacity, and they're using at least some Udacity courses in their program. So I think they're able to drive down the cost of their degree because they're benefiting from the infrastructure that Udacity provides. Me personally, what I think I'm going to do is start working on some personal projects nights and weekends to build a very strong portfolio so that I can put in applications to companies. If I find that that's not going very well, then I plan to apply to this OMSCS program. Now, a dirty little secret about me, I have worked professionally in machine learning for a time, but I was sort of grandfathered into the position. So I didn't kind of earn my way there. That's why I say this episode is a little bit opinion-based based on things that I've read from conversations around the web and conversations that I've had with recruiters. But you may want to get a second opinion on some of this stuff. Okay, PhD. Why would you want a PhD? Now, a PhD is not necessary for getting a job in machine learning. It is not necessary. A master's is plenty sufficient. And in some cases, you can bypass that master's requirement by way of a very strong portfolio. So why on earth would you want a PhD? Now, a PhD will obviously get you into more doors, like a machine learning engineer role at Google or Facebook, some of the more hardcore basement artificial intelligence development roles. But I find that most people who do go for their PhD, that's not why they're doing it. They're not doing it to crack in the industry. They're doing it for the sake of the PhD itself. The thing about industry is that your job is going to be prescribing to you the types of projects that you work on and the nature of work in industry tends to be a little bit boring. I'm talking about recommender systems like Amazon's recommender system or anomaly detection programs or some very simple charts and graphs sort of data analysis kind of gigs. You're working in machine learning, yes, but you don't really have the freedom to explore all the greatest nitty gritty rabbit hole science that's going on in the realm of artificial intelligence and the discussions that are happening around consciousness and the futurology of modern research in AI. Going for your PhD in machine learning or artificial intelligence affords you the time to dive deep into the stuff that makes AI fun. If you see yourself as a mad scientist who wants to solve these riddles, then a PhD is for you. Now, a PhD program pays you, I don't know if you know this, you get your master's degree, you pay for your master's degree, $7,000 for the OMSCS program. But once you get into a PhD program, they pay you, now they pay you about $30,000. That's by comparison to $160,000 that you might be making in industry in machine learning. So it's kind of a compromise. You're gonna choose on one hand, do I get to work in all of the really cutting edge research and all this fun stuff and make $30,000 and maybe be working 80 hours a week? Or do I work in industry where I have a decent work-life balance and a very high pay scale, but the kinds of projects I'm working on are less likely to be very entertaining on average? That's that compromise that you're gonna have to come to terms with. Me personally, I'm going to try to crack into industry first by way of a strong portfolio. If that doesn't work, I'm going to get my master's degree and I'm going to reevaluate maybe three years down the line whether I wanna go for my PhD. I definitely see value in the PhD program. Really those hard questions being tackled by the scientists is very appealing to me. So I'm gonna leave the decision up to you. If you have any experience or any questions around this topic, please comment in the show notes. Maybe we can get some discussion going back and forth, maybe some more concrete evidence one way or another. I'm going to post some of the conversations that I've seen from around the web in the resources section and let you come to your own conclusions. That's it for this episode. And in the next episode, we'll get back to the technical details with logistic regression.