#51 Francois Chollet - Intelligence and Generalisation
Machine Learning Street Talk (MLST)
00:00
The Limits of Neural Networks
This chapter explores the constraints of neural networks in achieving Turing completeness and highlights the challenges in representing complex data structures. It focuses on the use of LSTM models for learning multiplication algorithms while addressing their inefficiencies and biases in expert perceptions of deep learning's capabilities. The discussion reflects on the gap between expectations and reality in AI, particularly regarding the limitations of current model architectures.
Transcript
Play full episode
