Explainability, Reasoning, Priors and GPT-3
Machine Learning Street Talk (MLST)
00:00
Challenges in CNNs and Knowledge Representation
This chapter analyzes the obstacles convolutional neural networks face in recognizing rotated images and emphasizes the need for innovative research in their architecture. It discusses the intricacies of integrating prior knowledge into AI, the significance of logical reasoning within NLP systems, and the importance of knowledge graphs in enhancing machine learning. Additionally, it reflects on the historical interplay between mathematics, human cognition, and artificial intelligence, advocating for a deeper understanding of core knowledge to advance AI capabilities.
Transcript
Play full episode
