Learning AI Poorly: One way to get a foundational stats background for free

(originally posted on linkedIn)

I’ve had a busy week getting the yard ready for the 2023 Jarts Midwest Regional Quarterfinals but it is the kind of work that lets the mind wander. At some point I realized that when I tell people they should read AI and ML papers to stay on top of fun things, I didn’t think that maybe there are a lot of words in those papers that don’t make any sense. That got me thinking, “Is there a way to get up to speed on some foundational math and stats that can make ML and AI easier to understand?” I think there is, and best of all, it is free.

The beginning is a bit of a grind, but worth it. A decent intro to statistics is instrumental. Like, you really need to understand it. Where can you start?

The Khan Academy Statistics Playlist is an excellent intro or refresher for everyone. If you want a lecture based format, MIT Open Courseware has an excellent course called Statistics for Applications on YouTube. Best part is, you can tell people you learned stats at MIT. People love that.

Once you get comfortable with stats, I think you should dive right into a book by Trevor Hastie. He co-authored the book, Elements of Statistical Learning, Data Mining, Inference, and Prediction which is not only free at that link from his website, is also one of the best books on the fundamentals of machine learning. It is a little… well, it kind of has a lot of… uhhh… mathematical rigor and isn’t super accessible for those of us who don’t have that deep of a mathematical background. But, no worries, because he also co-authored An Introduction to Statistical Learning with Applications in Python which I think has the perfect balance of learning how things work without getting lost in the weeds. This is a pretty new edition of the book that is now in Python (the one I read was in R… which was fine because it was way back when I knew R…) There is also a series of YouTube lectures that Trevor Hastie did for the book that looks really useful.

Of course, StatQuest with Josh Starmer should be on your radar. The playlists on Neural Networks / Deep Learning and Machine Learning are fantastic but just the Transformer Neural Networks, ChatGPT’s foundation, Clearly Explained!!! is well worth the 30 minutes even if you don’t know much about stats.

So, yeah, there is a lot to learn but the freely available resources are out there if you know where to look. Take some time, dig in, and have fun!