How I’d learn Machine Learning & AI in 2024 (if I could start over) -- 7-step Roadmap

Key Takeaways

learning machine learning in AI in 2024 is more important but also easier than ever if someone tells you the exact steps to take I will give you a simple to follow no BS Guide to the seven steps you need to take to be able to write your own AI code in the next 10 minutes hi my name is Tim I've been a python programmer and data scientist for over 10 years after teaching myself with free online resources without having a computer science background making all the mistakes so you don't have to not only do I know all the steps to take to become a data scientist and AI expert but I've since taught all of these topics to hundreds of people in real life programming and machine learning boot camps in different countries over the last few years this means I know exactly what people need to get from zero to getting a data science or AI engineering job and all the pitfalls on the way no matter your background I will tell you exactly where to start if you know nothing at all and what to focus on without any of the unnecessary fluff I will also tell you where to find tutorials for each step of the way and at the end of the video I will also tell you a little trick on how to land your first job all you need is a laptop and an internet con con action let's go python so I would probably start with this one if you have no programming experience python is one of the easiest programming languages to learn and there are incredible free resources out there to get you going I will link some in the description but you can also find some tutorials here on this channel be sure you know all the programming Basics like variable assignment simple math if else statements and for Loops it's also very good to know how to define functions and classes to be able to more quickly grasp the usage of tools that I will explain later on in the video being able to work with numpy arrays and matrices is also extremely helpful for certain machine learning algorithms that make active use of them such as neural networks PCA and other algorithms you should probably also already learn how to use jupyter notebooks which is a very powerful programming environment for python which will make both learning how to program as well as your data science projects later on much simpler math while you can maybe become a decent ml engineer knowing very little math by just using existing packages a good understanding of certain statistics and probability Concepts will make you a great one and help avoid many pitfalls when tuning and interpreting your ml models as well as Elevate you from the crowd I've seen many a data scientist and ml engineer struggle fine-tuning or interpreting their models because they lack the understanding of certain Concepts from statistics in particular which is probably the most important mathematical Branch to know if you want to get into machine learning which is arguably a branch of Applied statistics lastly to be really able to understand what happens under the hood with certain machine learning libraries and maybe even customize them you need to know the basics of calculus and linear algebra however if you have had these in high school or university and remember most of it you should be fine otherwise a few weeks of studying should be enough to get you up to speed for most machine learning algorithms you can simply study more as you go when more knowledge is required for a particular algorithm Concepts that it will be good to brush up on are what derivatives are and how to calculate them what vectors and matrices are and how to work with them for example using the dot product or calculating their Norm certain Concepts from trigonometri such as what a cosine is and how to calculate it will also be helpful if you don't have any math books lying around from school or access to any college classes there are amazing resources on the web such as Khan Academy or even YouTube and many more I will leave links in the description don't get too frustrated with this step just learn the basics and get started with the more applied things that I'm about to show you while looking into more math as you go you can always look up things later on Google is your friend basic data Tools in Python now some more fun stuff if you know the basics of python you can now get to know some of the cool data libraries that other people have coded for us and that are completely free to use hopefully you have already started using Jupiter notebooks but if not now is the time to start doing so it will make your life much easier I also already mentioned numpy earlier which is an amazing library for doing mathy stuff in python as well as working with arrays matrices and higher dimensional objects matrices are two-dimensional arrays that are basically tables of data such as Excel spreadsheets that allow you to work with large data sets very quickly knowing how to work with these large arrays will equip you well for working with images and neural networks later on there is another library that is like numpy and Excel on steroids called pandas which will allow you to do incredibly fast data analysis with tabular data such as spreadsheets and of course if you want to do data analysis you will want to create and look at plots and graphics for which you can use the extremely powerful plotting Library met plot lip there are great and free tutorials for all these all over the Internet just do the basic tutorials but actually code along and don't just read them that will really solidify your skills your python skills will improve along the way at this point you can already call yourself an entry-level data analyst to upgrade to data scientist you will need the next step so here I would suggest studying a bit of theory on the basic algorithms of machine learning before starting to use the amazing libraries that already exist so you have an understanding of how they work and how to tune them as well as when to use which algorithm and why and when to believe your algorithm and when not to machine learning arguably is a field of Applied statistics and now your statistics Knowledge from earlier on as well as some of that calculus and linear algebra will come in very handy machine learning roughly divides into two areas supervised and unsupervised learning supervised learning is when we have a so-called training data set where we know the truth that we are trying to predict such as predicting the price of a house the weight of a person or whether there is a cat or a dog in the picture a human has previously done the task manually and we can show the data to the computer to learn from it and then predict unknown values for new data it has never seen most of the algorithms that do this sort of predictive modeling are based in one way or another on linear regression even neural networks one of my favorite resources on the topic that basically got me started in machine learning is the book an introd ction to statistical learning which the authors offer completely for free on their website it will take you from the very basics of linear regression and classification to the more complex topics such as tree based algorithms support Vector machines or unsupervised learning more than memorizing algorithms it's important that you truly develop a deep intuition of the simple algorithms like linear regression and classification because most if not all of the more complex algorithms are just fancy versions of these even neuron networks are technically just layers of linear regressions so make sure you truly understand these more basic algorithms this will make you a 10 times better data scientist than simply knowing how to use the big AI libraries because you will truly understand when to use what algorithm and how to tune it and when to believe what it tells you the authors of isrl have also made a series of really awesome videos covering the topics in the book that are also completely free and can be found here on YouTube I will leave links in the description another great resource that is also completely free is Andrew Ang's machine learning specialization and related courses on corsera Andrew Ang is a Stanford professor and world-renowned Ai and machine learning expert as well as an amazing educator and his courses are truly amazing this is an introductory course but I would not skip it as I mentioned already being strong in the basics not only will make you be able to use and understand the more complex algorithms more easily but their understanding is also very often quizzed in interviews for data science and machine learning positions at big companies such as Google or Facebook or wherever you may choose to apply if you are looking for a job this step was probably the most timeconsuming one but you would now be able to pass most theoretical data science interview questions but now to the Practical tools lastly either alongside learning the theory or afterwards you can work with some real data using existing libraries such as scikit learn scikit learn is arguably the most important and popular machine learning library in the world and makes machine learning easy as eating a cake once you truly understand which problem requires the use of which algorithm psyit learn also has great tutorials and comes with toy data sets to practice on psych learn has you covered for the most common basic machine learning algorithms out of the box and is built in such a way that once you know how to use one algorithm you know how to use another one since the syntax is almost always exactly the same if you've already mastered all the steps before this one I think you can get a good grasp of psyit learn in one or two weeks you can now call yourself a data scientist let's move on to deep learning and neural networks these are arguably the work courses of much state-of-the-art AI while it's technically just one of many ml algorithms out there it's hard to get around them these days if you liked Andrew Ang's machine learning course you can now take his next course on corsera on neural networks and deep learning they're also great resources on YouTube such as three blue one brown who makes amazing math videos with Incredible visualizations and also covers neural networks Andre Kathy is another amazing YouTuber who in his neural networks Zero to Hero series takes you through the building of a neural network algorithm from scratch I will again leave links in the description as for deep learning libraries there are several Frameworks with some popularity among them pytorch Caris and tensorflow each has their pros and cons and at this point you might also already know enough to know which one best suits your needs if you don't have any idea what you want to start with I'd probably recommend a Caris tutorial first since Caris is maybe the simplest of them all and can be used as a userfriendly API to sit on different backends including pie torch and tensorflow since now you have worked on some tutorials with toy data sets it's time to get your hands dirty and work on some real world data sets so you can start working on real data projects at any point during your journey and maybe anytime you learn a new concept or algorithm you can already try your luck with some real data or you wait until you have done tutorials for all the important algorithms to start for real data maybe you have some data you want to work with maybe some big Excel spreadsheet from your old job or some class you took maybe you want to export data from your favorite Health tracking device or any other smart device that you own if not you can simply search on the internet for some data sets a good resource is kaggle which is a data science competition platform and online community kaggle allows you to publish or find data sets to work on and there are even competitions many of them with real price money make sure you start with easy projects though and don't expect to win any money especially early on so you don't get demotivated many of the participants are professional data scientists with many years of experience plus they probably have better computers than you do which will get more important the more complex the algorithm and the bigger the data set which leads me to another Point don't start with huge data sets work your way up and get a feeling for how much your computer can handle once a certain problem is too much for your computer you might want to learn how to spin up instances on cloud computing platforms such as a WS Google cloud or Microsoft Azure any or all of the projects you work on be it real or toy data sets you can publish on your GitHub or personal website GitHub is a platform for developers to store manage and share their code so while learning you will actually already build a portfolio for your job applications which can be sent to potential employers and might also help employers find you if you found this video helpful share it with someone who you think might also like it and get started on one of the tutorials in the description or on this very Channel thanks for watching