Map the AI · Daily learning
One small lesson. One step forward.
Learn how AI handles text and generates answers through a short experiment and a question. Open the linked tutorial to explore further.
Your learning preferences
Preferences support future recommendations. This foundation practice currently keeps a fixed sequence.
A learning track for your goal
AI Literacy
Understand what AI is, how it works, and where it fails — without writing a line of code.
The full track uses the existing knowledge graph. Short-lesson mastery currently covers the foundation units below.
Foundation lesson progress
Loading progress…A correct answer is a start. Regular reviews help you remember. Take a break whenever you like.
Let the parameter take a step
Observe how the learning rate changes a gradient update.
The gradient describes the local slope of the loss. Gradient descent subtracts the learning rate times that gradient from the parameter. A suitable small step can reduce loss; a step that is too large can increase it. Real training usually updates many parameters with batches of examples.
One parameter and one training example: input x=1, target 3, prediction w. Loss L=(w−3)², gradient g=2(w−3). Each click computes the new w: w−learning rate×g.
Updates: 0 / 20
w = 0.000 · Loss = 9.000 · Gradient = -6.000
Show each step’s values
- 0: w=0.000, loss=9.000
Compare 0.1, 0.5 and 1.1: larger steps may overshoot and move farther away. These rates apply only to this example; suitable rates depend on the loss and parameter scales. Fitting this one example does not establish generalization.
Loading progress…