Neural Networks and Its Discontents Part 2: Forward Forward Algorithm, reflections on a recent interview with Geoffrey Hinton

Ashkan Beheshti
11 min readMay 30, 2023

šŸ¦Š Welcome to Part 2 of our exploration into the fascinating world of backpropagation and its limitations in the realm of artificial intelligence, and forward forward algorithm as an alternative. In this two-part series, we breifly delve into the linchpin of AI learning, its significance, and the emergence of a promising alternative: the forward-forward algorithm.

In our exploration of backpropagation in part 1, its significance in AI, and its history, we have laid the groundwork for a deeper understanding of the learning mechanisms behind neural networks. Now, we have the opportunity to delve even further into this topic through an enlightening interview with one of the most influential figures in the field of deep learning: Geoffrey Hinton.

In EPISODE #112 of Eye-on.AIā€™s podcast, Hinton provides a comprehensive explanation of his new learning algorithm, the forward-forward algorithm, which offers a fresh perspective on how the cerebral cortex might learn. By examining this alternative to backpropagation, we gain valuable insights into the future of AI and the potential for more realistic models of learning.

Throughout the interview, Hinton touches on various aspects, from the philosophical foundations of perception and experience to the practical methodologies he employs in his day-to-day work. His insights into the limitations of backpropagationā€¦

--

--

Ashkan Beheshti
Ashkan Beheshti

Written by Ashkan Beheshti

Psychologist-Data Scientist, exploring the interplay between human learning & machine learning

No responses yet