Symbolic AI
This is a term that I remember, in undergraduate, to have heard about, but, sempre que considerava que não era aprendizado de máquina, desconsiderava. It seems not the right perspective. But, when symbolic AI is considered in the based-learning and in a sense that neural networks should compress and produce not another neural network, but knowledge with symbols that humans can understand, this is really other level of perspective. It seems the right approach. It is not about understanding the neural networks in terms of internal learning mechanisms. We don't understand the brain and the internal learning mechanisms. We can understand with symbols the knowledge that a neural network possui.
E, um aspecto interessante that I saw professor Atlas mentioned, is about reading the papers. It is there, man. It is about that. And this blog post is acompanhado dessa perspectiva. Da leitura de papers.
So, started reading the paper Why Neural Network Can Discover Symbolic Structures with Gradient-based Training: An Algebraic and Geometric Foundation for Neurosymbolic Reasoning.
Really great that I have been studying Montessori in The Advanced Montessori Methods with the discussion for arithmetic and geometry learning for childs.
Besides that, I have been studying math and logical reasoning.
Great learning.