LLMs have surprised the world with their abilities. Sometimes, though, these models fail spectacularly in reasoning tasks. So there is active research to improve these models both at the architectural level and at the level of techniques to improve reasoning skills.
On the other hand, there are still obscure points about how these skills emerge. Similarly, what does it mean to learn? What does it mean for a neural network to learn? Can an understanding of human learning drive a new way of learning for the AIs of the future?
Here, this article is about that. A journey into the learning of artificial and natural brains.
During this article, we will discuss:
How current LLMs are learning, the advantages and the limitation
What does learning actually mean and how is different from memorization
What factors are affecting learning and how efficient learning is emerging
A discussion on how we can think of a new learning paradigm
