How to learn the fundamentals of AI
Caveat Emptor - I am not an AI or Machine Learning expert, and I don't intend to become one. So, take this advice with a pinch of salt.
I am writing this post mainly for my future self because I keep forgetting this every time I start working on a new project. The answer to ‘How do I learn AI?’ is basically the same as the answer to ‘How do I learn X, Y, or Z?’ - just f*cking ship something, anything at all.
After spending more than half a decade in EdTech and experimenting with different learning styles, I've realized that all the time spent on studying the basics of Machine Learning - Linear Algebra, Calculus and god knows what - was a colossal waste of time. Mind you, it was fun. I love mathematics, but it didn't result in any real-world, tangible skills.
If you want a garden in your backyard, you wouldn't start by studying the basics of soil chemistry. You would just plant a sapling, watch it grow and learn from the experience. Then, plant the next one and continue learning from each attempt. And so on...
Action produces information!
Year after year, we lament over our stupid education systems that teach us useless fundamentals, which translate to zero real-world practical skills. However, the system hardwires this dumb learning style, and we forget that it's counterproductive. Never start with the fundamentals of anything. Instead, build your first half-baked prototype using an API, learn from it, iterate, build something better. When you get curious, watch a video or read a book that explains how it works under the hood.
Top-down approach. Start with the big picture and slowly dig your way down to the basics.
I can say this quite confidently - I probably know more about the mathematics used in ML than Pieter Levels (levelsio). But guess which one of us is making bank with an AI product while the other has nothing to show for it?
It's not an easy philosophy to internalize, and I keep forgetting it all the time. I recently made this mistake again - tasked with building an LLM pipeline for our internal teams, I spent days - probably weeks - diving into the documentation of LlamaIndex and reading through the low-level API details, trying to understand how everything works under the hood as the deadline approached. At the last moment, I threw it all away, pulled an all-nighter and shipped an almost fully working product using the high-level API - almost, because I missed building the front-end :(
As I build more products, I'll eventually figure out the fundamentals along the way.
Spending time studying the fundamentals is just procrastination - an excuse to avoid doing the real thing: building something tangible.
I don't want to leave you with just a philosophical tirade, so here's a list of practical resources that'll help you build a solid foundation once you've gotten your hands dirty -
YouTube
Books
You'll continue to come across many more resources. There are plenty out there!
Lest you forget, I'd like to remind you once again - Action produces information.
Until next time, TJ!