The answer was always there

Every learning system in this journey so far has worked the same way.

A picture arrived with a label already attached. A mushroom arrived with safe or dangerous already written on it. ImageNet arrived with millions of human-supplied answers, one per image.

The machine made a guess. Something compared that guess against the correct answer sitting right there. The difference between the two became the correction, and the correction travelled back through the network and nudged the weights.

That comparison is the engine of everything in the last two modules. No answer, no difference. No difference, no correction.

Which raises a question the journey has not asked yet.

What happens when nobody knows the right answer?