By Wojciech Gryc on November 20, 2023
There are lots of blog posts, resources, and articles on how to learn AI/ML. Why am I adding one more? I find most of the current lists are self-promotional or focus too much on other blog posts rather than developing a deep and fundamental understanding of AI/ML. This post outlines some (free!) resources that I personally believe will help you build an incredibly deep foundation in AI/ML.
The courses or lecture series are presented in descending order of priority or preference. I explain what I like about each one in their respective section.
Fast AI is a great balance between theory, coding, and actual working examples. Their entire course (and textbook!) is free and open, and everything is available to run via notebooks. If you have some Python experience and keen to learn from the ground up, this is the resource you should start with.
Andrej Karpathy is a founding member of OpenAI and led Tesla's computer vision team. His Youtube lectures delve into first principles AI/ML, where you start from a single line of code and build GPTs, deep neural networks, and more.
Andrej's lectures are a fantastic place to start if you want a digestible complement to other courses. Be aware that breadth and digestibility also mean you might gloss over many topics, and Andrej also doesn't cover mathematical foundations in depth.
The "Learning From Data" course is a series of lectures recorded by Caltech's Professor Yaser Abu-Mostafa. These are comprehensive lectures that delve into the theory of machine learning. They are a great complement to Andrej Karpathy's videos, in particular.
... and of course, it's not all about courses and lectures! The following textbooks are fantastic resources for learning AI/ML.
This book, coauthored by Yoshua Bengio, is a fantastic and deep introduction into deep learning. If you are specifically keen to learn about neural networks, this is the place to go. It is also freely available online.
For those interested in a deep and theoretical understanding of statistical learning and why it works, this textbook is for you. It almost exclusively focuses on the mathematical foundations of learning algorithms. It is great for those interested in joining cutting edge research labs and companies, but won't cover more modern programming topics, frameworks, or software engineering tools.
This 1,000-page AI/ML textbook starts at the very basic and works its way through to statistical learning, machine learning, deep learning, decision theory, and more. It is incredibly comprehensive and used by numerous universities as a standard introdictory AI/ML textbook.
Note that this textbook is not free but well worth reviewing. If you can't get access to it, checking the table of contents on its site is enough to guide you in the right direction with the other resources shared here.
There is no one-size-fits-all approach to learning AI/ML. The most important part of the journey is to get started and make progress regularly, however incremental.
With the above in mind, we recommend you review all the resources above. Use them all to inform your knowledge and learning, diving into the deeper theory of some of the textbooks if it makes sense to do so, or keep things focused on engineering via Fast.ai and Andrej Karpathy's videos.
As you make progress, you'll be able to address countless real-world problems. Good luck!