What Is AI?
How Machines Learn
8 min+70 XPStep 1/12
Learning as tuning dials
Every ML model is a function with millions of adjustable dials called parameters (or weights). Training means adjusting the dials so the function's output matches the examples.
Start with random dials → predict garbage → measure how wrong → nudge the dials slightly toward 'less wrong' → repeat millions of times. That loop is learning.
analogy
Like adjusting the shower temperature: too hot, turn down a bit; too cold, turn up a bit — until it's just right. Training does this with millions of dials at once.