061: Interpolation, Extrapolation and Linearisation (Prof. Yann LeCun, Dr. Randall Balestriero)
Machine Learning Street Talk (MLST)
00:00
Unpacking Deep Learning: Interpolation and Extrapolation
This chapter explores the debate on deep learning's capacity for reasoning versus mere data interpolation. It discusses the significance of evolving methodologies, particularly the shift towards self-supervised learning, to improve AI reasoning abilities. The conversation delves into the intricacies of data representation, neural network architecture, and the historical context of different learning techniques.
Transcript
Play full episode
