If you search “how to learn AI” you will find dozens of lists recommending the same five or six platforms. That part is not the hard problem. The hard problem is sequencing: what to study first, how much math you actually need before you touch a model, and when to stop reading and start building. This guide walks through that sequence in order, using free resources at every stage.

Start with what AI and machine learning actually mean

Before opening a single course, spend an afternoon understanding the shape of the field. Artificial intelligence is the broad goal of getting computers to perform tasks that normally require human judgment. Machine learning is the dominant technique used to get there today: instead of programming explicit rules, you show a system many examples and let it find the pattern itself. Deep learning is a specific approach to machine learning built on layered neural networks, and it is the technique behind most of what people now call generative AI, including large language models like the ones behind ChatGPT.

Elements of AI, a free course originally built by the University of Helsinki, is a genuinely good starting point for this stage because it explains these ideas without requiring any code. Google’s free Machine Learning Crash Course covers similar ground with slightly more technical depth. Either one is enough to get the vocabulary straight before you move on.

The five stages, in the order that actually works

Skipping ahead to generative AI or large language models before you understand basic machine learning is the single most common reason people bounce off self-study. The stages below build on each other.

Stage 2: Python, but only the part you need

You do not need to become a software engineer before you touch AI. You need enough Python to load a dataset, clean it up, and call a library function. freeCodeCamp’s Python curriculum and Kaggle’s free Python and Pandas micro-courses both cover exactly this slice, and both are free with no time limit.

Stage 3: Core machine learning concepts

This is where a structured free course pays off more than random YouTube videos, because the ideas build in a specific order: what a training set is, why you hold out a test set, what overfitting means, and the difference between regression and classification. Kaggle Learn’s Intro to Machine Learning and Google’s Machine Learning Crash Course both teach this directly, with runnable code instead of only slides.

Free courses worth starting with

You do not need a long list. You need one resource per stage that you will actually finish. These are free, well established, and cover the stages above without requiring a paid upgrade to see the core material.

• Elements of AI: a no-code introduction to what AI is and where it shows up in everyday life.
• Google’s Machine Learning Crash Course: a free, hands-on introduction to machine learning concepts with exercises.
• Kaggle Learn: short, free, code-first micro-courses on Python, pandas, and intro to machine learning, each just a few hours long.
• freeCodeCamp: free, project-based Python and machine learning curriculum with a certificate of completion.
• Hugging Face’s free courses: hands-on introductions to natural language processing and transformer models once you’re past the basics.
• Udemy’s free course listings: rotating free and preview courses across AI topics, worth checking once you know what you’re looking for.

For a full breakdown of what each platform offers, see our guide to free AI courses across Udemy, Coursera, Google and other providers, or browse Udemy’s current free listings directly on our free courses page.

Build while you learn, not after

The people who actually stick with AI are the ones who start building something small in week two or three, not after they feel “ready.” Readiness does not arrive on its own. A workable pattern is to pick one small, boring project, something like predicting housing prices from a public dataset or classifying whether an email is spam, and rebuild it as your skills improve. Kaggle hosts thousands of free datasets specifically for this kind of practice, and posting your work on GitHub gives you something concrete to point to later, whether that’s for a job application or just your own record of progress.

• Pick one dataset and one problem, not five.
• Expect your first version to be rough. That’s normal, not a sign you’re behind.
• Rebuild the same project with a slightly better technique once you learn one.
• Write two or three sentences about what you tried and what you’d change. Future you will thank present you.

How long this actually takes

Plenty of marketing copy promises AI fluency in a weekend. In practice, most people who go from zero to comfortable with the fundamentals spend somewhere between one and three months of steady, part time study, meaning a handful of hours a week rather than an intensive full time push. Basic literacy, being able to explain what a model does and follow along with AI news without getting lost, comes much faster, often within the first couple of weeks. Genuine hands-on competence, being able to train and evaluate a simple model yourself, is what takes the longer stretch.

The biggest variable is not intelligence or background. It’s consistency. Thirty focused minutes a day for two months will get you further than a single weekend crammed with videos.

Common mistakes that slow people down

• Jumping straight to large language models: without understanding basic machine learning first, the underlying concepts (training, tokens, weights) won’t make sense.
• Collecting courses instead of finishing one: bookmarking ten free courses and finishing zero is the most common failure pattern.
• Treating math as a prerequisite instead of a companion: you can start with applied tools and pick up the statistics behind them as questions come up, rather than front-loading months of theory.
• Never sharing your work: a project sitting only on your own laptop doesn’t build the feedback loop that public practice does.

Frequently asked questions