The next slide is the strangest idea in this module, so let us set up the one contrast that makes it click. It is a contrast between two things scale does, and they are not the same.

The first you just saw: the smooth, predictable kind of improvement. Make the model bigger and it gets steadily, reliably better at what it already does. That is the scaling-law curve, a clean line you could draw in advance. No surprises. More size, a bit more skill, smoothly.

The next slide is about the other kind, and it is nothing like smooth. Sometimes, past a certain size, a model can suddenly do something it simply could not do at all before. Not "a little better." Off, then on. A brand-new skill, appearing at a size nobody could predict.

So as you read on, keep both in one hand: scale makes old skills smoothly better, and scale also makes entirely new skills suddenly appear. The first was expected. The second was the shock. That gap between the smooth curve and the sudden jump is the whole point of what comes next.