AI is a monster enormous. I also tried to get into it alone, with YouTube tutorials and technical articles. I drowned in formulas mathematical and code that I didn’t understand. I realized I was trying to build a skyscraper without even knowing how to lay a brick.
What ultimately worked was to stop trying to “learn AI” as a whole. It’s too much. Instead, I focused on a single, stupidly small problem. For me it was: “Can I make a program recognize whether a photo is of a dog or a cat?” Not all animals, just that.
I used a hands-on course, Andrew Ng’s on Coursera, but I didn’t do it passively. As I followed the lessons, I applied every minimal concept to my dog and cat problem. Even though the course used different examples, I forced that knowledge into my project. Theory without immediate application was forgotten in two days.
Another thing: I used high-level tools at first. Keras, Fast.ai. Purist people say you have to start with the basics, with pure Python and numpy. But that kills motivation. If you first get something to work, even with tools that are like a black box, you gain the confidence to later want to know how it works inside. It’s the cycle: working -> curiosity -> fundamentals.
I also joined a small community, not giant forums. A Telegram group of 100 people where everyone was doing projects. You’d ask silly questions and they’d answer without arrogance. The key was to participate, not just read. If someone had an error, I’d try to help them even if I didn’t know how. It forced me to think.
The main mistake was believing I could be self-taught in a vacuum. AI isn’t learned like history by watching documentaries. It’s learned like carpentry: by making sawdust, making mistakes, and having someone tell you “don’t tighten the screw so much.”