Elements of AI
A friendly online course from the University of Helsinki and MinnaLearn on what AI can and can't do, with no maths or programming needed.
Pairs with lesson 1, What is AI, really?
Hand-picked courses, videos, papers and tools worth your time, each with a note on why and which lesson it pairs with.
16 resources
A friendly online course from the University of Helsinki and MinnaLearn on what AI can and can't do, with no maths or programming needed.
Pairs with lesson 1, What is AI, really?
Andrej Karpathy's one-hour talk on what an LLM is, how it's trained and where it's heading. The best single video to watch before or after meeting the LLM.
Pairs with lesson 3, Meet the LLM
Grant Sanderson's animated series on how neural networks see and learn, with pictures doing most of the work. Start with chapter 1 and carry on from there.
Pairs with lesson 9, Neural networks without the scary maths
Paste any text and watch it split into tokens and token IDs, right in the browser. The quickest way to try what the tokens lesson describes.
Pairs with lesson 5, Tokens: the LLM's alphabet
Run Python notebooks in the browser, with some free GPU time. Use it for any code in our lessons without installing anything on your laptop.
Google's hands-on introduction to machine learning, with videos, visualisations and coding exercises. Good once you're comfortable with a little Python.
Pairs with lesson 11, How models learn
Microsoft's open course for developers who want to build with LLMs: prompts, chat apps, search apps and more, each lesson with code you can run.
Pairs with lesson 3, Meet the LLM
A short DeepLearning.AI course with Cohere on embeddings, dense retrieval and reranking, with Python notebooks. It builds directly on the embeddings lesson.
Pairs with lesson 7, Embeddings: meaning as numbers
Learn to use, fine-tune and share transformer models with the Hugging Face libraries. A practical next step after the Transformer lesson.
Pairs with lesson 10, The Transformer, step by step
Download and run open models such as Llama, Gemma and Mistral on your own computer with one command, and call them from your code.
The 2017 paper that introduced the Transformer, the design behind today's LLMs. Read the introduction and look at the figures first.
Pairs with lesson 10, The Transformer, step by step
The paper that named retrieval-augmented generation. Read it after the RAG lesson to see where the idea came from.
Pairs with lesson 8, RAG: giving the model an open book
Showed that asking a model to write out its steps improves its answers on maths and logic problems. A key paper behind today's reasoning models.
Pairs with lesson 13, Can AI reason?
An interactive book that teaches deep learning with the maths, the explanation and runnable code side by side.
Pairs with lesson 11, How models learn
Andrej Karpathy builds neural networks and a small GPT from an empty file, line by line, in Python. Best saved for after the capstone lesson.
The classic textbook by Goodfellow, Bengio and Courville, covering the maths and theory of neural networks in depth.
Pairs with lesson 9, Neural networks without the scary maths
Know something better? Tell us about it.