How do I actually get started learning machine learning without a PhD?

You don't need a PhD. You need a problem you care about and a willingness to feel confused for a while. The field has a reputation for being walled off behind calculus and linear algebra. Some of that math matters later. But not on day one. Start with the practical stuff. Fast.ai offers a free course designed for coders who want to build working models quickly. It flips the usual approach — you train an image classifier in the first lesson, then gradually peel back the layers to understand what's happening. I've seen complete beginners get a model running in an afternoon this way. Google's Machine Learning Crash Course is another solid entry point. It's more structured, with short videos and in-browser exercises. You'll learn terms like 'loss' and 'gradient descent' in context, not as abstract concepts. Expect to spend 15-20 hours on it. Here's a tip most guides won't give you: pick one tiny, personal project on day one and chase it relentlessly. Not a tutorial project. Something you actually want to exist. Maybe a model that identifies whether your cat is on the couch (so you can automate a treat dispenser). When you're solving your own problem, the motivation to push through confusing moments is completely different. You'll need some Python. Not expert-level — just enough to read code and tweak it. If you've never coded, spend two weeks on a basic Python course first. The biggest barrier isn't the math. It's the feeling that everyone else understands something you don't. They don't. Everyone is confused. The ones who succeed just keep going anyway.