Skip to main content

YouTube channels to learn AI for free, from maths to building agents

By Xplaind editors ·

Cover: "AI channels worth your time." beside a grid of six play buttons, one in brass

YouTube is the largest free AI school in the world, and also the easiest place to waste a month. The trick is to pick channels by what you need to learn next, watch with a notebook or code editor open, and stop when you can build the thing yourself.

We picked these channels because they explain clearly, stay accurate, and teach something you can use. Nobody paid to be on this list. We checked that every channel link below works on 25 October 2026.

To understand the ideas and the maths

  • 3Blue1Brown. Grant Sanderson's animated series on neural networks, linear algebra and calculus, and on how transformers and attention work, are the clearest visual explanations available. Watch them before any deep learning course.
  • StatQuest with Josh Starmer. Statistics and machine learning explained step by step, from p-values and regression to gradient boosting and neural networks. Ideal when a textbook loses you.

To understand language models deeply

  • Andrej Karpathy. A founding member of OpenAI and former director of AI at Tesla. His "Neural Networks: Zero to Hero" series builds a GPT-style model from scratch in Python, and his long general-audience talks on how LLMs work are the best overview for anyone.
  • Umar Jamil. Long, careful walkthroughs that code transformers, attention variants and other model architectures from scratch, with the maths explained.

Structured courses on video

  • DeepLearning.AI. Andrew Ng's organisation. Talks and course excerpts; the full short courses on its website are free and pair well with the videos.
  • freeCodeCamp.org. Full-length courses, often several hours, on Python, machine learning, deep learning and building with language models. Good when you want one long, complete course.

Indian creators

  • CampusX. Nitish Singh's long, structured playlists on machine learning, deep learning and generative AI, taught in Hindi. Among the most complete free curricula in any language.
  • Krish Naik. Practical tutorials and end-to-end projects on machine learning, MLOps and generative AI, plus career advice for the Indian job market.
  • codebasics. Dhaval Patel's channel on Python, data science and AI, with a strong focus on real business problems and projects you can put in a portfolio.

To keep up with research and news

  • Two Minute Papers. Short, enthusiastic summaries of new research, especially in graphics and generative models. Good for curiosity, not for depth.
  • Yannic Kilcher. Long, detailed readings of important machine learning papers. Best once you know the basics.
  • AI Explained. Careful analysis of new model releases and benchmarks, with attention to what the claims really show.

For builders

  • LangChain. Tutorials on building retrieval systems and agents with LangChain and LangGraph, from the team that makes them.
  • Anthropic and Google DeepMind. Talks and explainers from the labs themselves, useful for understanding how the people who build models think about safety, agents and evaluation.
  • sentdex. Long-running, hands-on Python and machine learning tutorials, often building things from scratch.

How to learn from videos without getting stuck

  1. Watch with a goal. Know which lesson or project the video is for. Our paths give you that order.
  2. Code along, then close the video and rebuild it. If you cannot rebuild it without the video, you have not learned it yet.
  3. Use 1.25x speed for familiar material, 1x for new ideas. Pause and write down the one thing you want to remember.
  4. Check the date. AI tools change fast. A two-year-old tutorial on a library may use functions that no longer exist; ideas age much more slowly.
  5. Stop watching and start building as soon as you can. One finished project teaches more than ten hours of video.

Where to go next

For each topic on our paths we pick one best free course, video series or set of docs, and explain why. Browse them all in the course directory, or start with How transformers work to see how a lesson pairs our explanation with the best free video.

Questions people ask

Can you learn AI from YouTube alone?
You can learn most of the ideas from YouTube, but not the skill. Videos explain; you learn by building. Pair each channel with a project and a structured path so you know what to watch next and when to stop watching.
Which YouTube channel is best for learning AI in Hindi?
CampusX teaches machine learning, deep learning and generative AI in Hindi with long, structured playlists. Krish Naik and codebasics also teach in a mix of Hindi and English and focus on practical projects and careers.

Today's AI turns are used

Lessons, roadmaps, videos, practice cards and the code playground stay free with your account, with no limit.