AI School › What AI is · Lesson 3 of 130

🧠 How a machine learns

Many examples, many attempts, many corrected mistakes: it is called training.

How did you learn to ride a bike? You try, you fall, you adjust, you try again. A machine learns the same way, just much faster: it makes millions of attempts, and with each mistake it slightly adjusts its "weights", which are like internal knobs. This process is called training.

At the beginning the AI gets almost everything wrong: it says "cat" when it sees a dog. But every time it makes a mistake, it receives a correction and turns its knobs a little. After millions of corrections, it guesses right almost every time. The secret is practice, just like for you: except it can do a million exercises in an hour.

🎒 Try it in class (or at home)

Experiment: blindfold a classmate who must throw a ball into a basket. After each throw say just "more right" or "more left". How many throws until they learn? That is training.

← Lesson 2: Where you meet AI every day · Lesson 4: What data is →

📱 Download the Android app (beta) iOS coming soon

Say what you mean. Get what you need.

Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.