Most of our learning happens via unsupervised learning, but how is that problem considered or not considered in deep learning?

My understanding of the question is that we, as humans, seem to be able to work with not much feedback or in an unsupervised way. But why is it that deep learning is requiring supervision? As I said, there is a big gap between how machines, how deep learning is working and how human brains work. I think the correspondence between the two is not very tight. So people do all kinds of analogies, but actually, they're two very different systems right. And there is no intention also to make deep learning like the brain, right. So we should not think of it like that. But it is true that one of the fundamental problems to solve with deep learning in the journey towards AGI is how can we learn with lesser supervision? How can it learn to generalize better, right? It cannot be unsupervised, mostly, but at least with less supervision, how do you do it? I think it's a fundamental difference in how the models work. The deep neural networks that we have today have many, many parameters, millions and even now billions of parameters. And to train such large networks with gradient descent, you need large input data, there's no way of overcoming that we would need fundamentally different ways of modeling to overcome this problem.