Loading slide
The word "training" gets used a lot in AI. Let's slow down and be specific about what it means, because it's the same thing every time.
Training is just this: make a guess, see how wrong you were, adjust slightly, repeat.
That's the whole thing. There's no understanding happening. No moment of insight. The machine doesn't look at a wrong answer and think "ah, I see what I did." It just measures the size of the mistake and uses that number to nudge every weight slightly, in the direction that would have made the mistake smaller.
You've done this yourself. When you learned to ride a bike, nobody gave you a manual with the exact angles to lean at or the precise forces to apply. You fell. Your brain noticed the mistake. The connections between your neurons shifted, invisibly, toward better balance. You got back on.
You didn't understand balance. You trained it.
The Perceptron did the same, and so does every AI system since. The word "training" always points back to this one humble loop: guess, measure the mistake, adjust, repeat.