Unsupervised Brain Models - How Does Deep Learning Inform Neuroscience?

Explore the fascinating intersection of deep learning and neuroscience in this comprehensive interview with Patrick Mineault. Delve into how unsupervised and self-supervised deep neural networks are informing our understanding of brain function, particularly in visual processing. Examine influential papers from the past year that connect machine learning insights to neuroscientific discoveries. Investigate the ventral and dorsal streams of visual processing, concept cells, and representation learning. Challenge existing theories like the manifold hypothesis and discuss current questions in the field. Consider whether neuroscience should inform deep learning development and learn about initiatives like Neuromatch Academy. Gain valuable insights into the ongoing dialogue between artificial intelligence and brain science in this thought-provoking discussion.

- Intro & Overview
- Start of Interview
- Visual processing in the brain
- How does deep learning inform neuroscience?
- Unsupervised training explains the ventral stream
- Predicting own motion parameters explains the dorsal stream
- Why are there two different visual streams?
- Concept cells and representation learning
- Challenging the manifold theory
- What are current questions in the field?
- Should the brain inform deep learning?
- Neuromatch Academy and other endeavours