AI from zero
Start here if AI feels like magic. You will teach a computer by example, watch it get less wrong, build a single artificial neuron, see how a chatbot writes one word at a time, decide what an AI agent is allowed to do, work out where a model can run, see how AI products try to stay safe, compare open and closed models, and try real models that name photos, write down speech, and draw from noise. No math or code required.
- Level
- Novice
- Length
- 13 lessons, 3 hr 21 min
- Assumes
- Nothing. Curiosity is enough.
- 1What is AI, really?Most of today's AI is software that learns its behavior from examples instead of following rules a person wrote. Teach a computer by example and see what that means.12 min
- 2Turning things into numbersModels only see numbers. Learn how pictures and objects become numbers, and why choosing the right measurements decides what a model can learn.12 min
- 3Getting less wrongLearning starts with measuring how wrong you are. Fit a line by hand, score it with one number, then watch a computer improve it step by step.14 min
- 4The artificial neuronThe building block of every neural network is a tiny calculation. Set its weights by hand, see it draw a line through the data, then let it learn the weights itself.15 min
- 5Networks of neuronsOne neuron can only draw a straight line. Connect a few and they can carve out curves. Train a tiny neural network and watch it happen.18 min
- 6How a chatbot writesChatbots write one word at a time, each time choosing from a list of likely next words. Build the simplest version of that idea and write with it.15 min
- 7What AI gets wrongModels learn only from their data, and can be confidently wrong outside it. See a model fail far from its training data, and learn why fluent chatbots can state falsehoods.14 min
- 8AI that takes actionsA chat assistant writes replies. An agent also acts, searching, editing files, sending email. Play the program around one and see who really acts, why it should ask you first, and what goes wrong.12 min
- 9Where your AI runsSome AI runs on your phone or laptop, some in a data center far away. See what leaves your device in each case, and work out where a model of any size can run.15 min
- 10Safety in AI productsWhat happens between your prompt and a product's answer, from hidden instructions and refusals to jailbreaks and flattery, and how a watermark can show where a text came from.20 min
- 11Open and closed modelsWhat open weights, open source, and closed models mean, what real model licenses allow, and the trade-offs in capability, cost, privacy, and safety. Compare licenses clause by clause.18 min
- 12How AI sees, hears, and makes picturesPoint a real image model at photos or your camera, see your voice as a picture before a model turns it into words, and watch a model draw a digit out of noise, all on your own device.22 min
- 13When someone attacks an AI appSomeone can hide instructions in an email or web page that an AI assistant reads. See it happen with a small real model, which protections hold when the AI is fooled, and what you can do.14 min
Labs in this track
- Teach by exampleAdd labeled points and watch a k-nearest-neighbors model shade the whole plane with its guesses.
- Which measurements matterPlot sixty fruit using any two of four measurements and see which ones let a model tell apples from oranges.
- Fit a lineDrag a line through ice cream sales data, score it by its squared misses, then watch gradient descent find the best line.
- One neuronSet the weights and bias of a single artificial neuron, watch it draw a straight decision line, then let it learn from examples.
- A tiny neural networkTrain a network with one hidden layer on curved data and watch several straight lines combine into a curved boundary.
- Next-word predictorA tiny language model trained on Aesop's fables. See its next-word probabilities, pick words yourself, and change the temperature.
- Confident, and wrongA model trained on one region of data is tested somewhere else. Its accuracy collapses while its confidence stays high.
- Approve or denyPlay the program around an AI assistant booking dinner. It proposes actions; you decide which run, and see what a payment request hidden on a web page does.
- Will it fit?Turn a parameter count and bits per number into a download, the memory needed, and a speed ceiling, and see which models fit a phone, a laptop, or need a data center.
- Find the watermarkA word model writes with a hidden statistical watermark. Count green words with the key, see how length and strength matter, swap words to wash it out, and test your own writing.
- Read the licenseChoose what you want to do with a model and see what MIT, Apache 2.0, OpenRAIL-M, Llama 4, Gemma, CC BY-NC, and a closed API's terms allow, each answer quoted from the text.
- Drawing a digit from noiseAsk a small trained network for a handwritten digit and watch it turn random noise into one, step by step, in your browser.