Peter Dayan: How to study the brain from a computational view | Q-Learning, Memory, Decision Making

In this episode, we have the distinct privilege of speaking with Prof. Peter Dayan, director at the Max Planck Institute for Biological Cybernetics in Germany. Prof. Dayan is renowned for using methods from machine learning to understand brain function, particularly linking neurotransmitter levels to prediction errors and uncertainty. His pioneering work in reinforcement learning and co-development of the Q-learning algorithm, have shaped the field of computational neuroscience in profound ways. He is also the co-author of the highly influential textbook Theoretical Neuroscience and co-founder of the prestigious Gatsby Computational Neuroscience Unit. In 2012, Prof. Dayan received the Rumelhart Prize and in 2017 he was awarded the Brain Prize. On a personal note, it was truly an immense honour to sit down with such a brilliant mind like Prof. Dayan. He was incredibly kind and generous with his time, staying with me until 10:30 PM to answer all of my questions.

Join us as we explore decision making, learning, and memory from a computational perspective.
This podcast is supported by the International Max Planck Research School for Neurosciences, the Cluster of Excellence Multiscale Bioimaging, and the European Neuroscience Institute in Göttingen. If you want to support us, please subscribe to our channel and do not hesitate to drop us a comment under our videos.

Timestamps:
00:00:00 In this episode
00:00:26 Introduction
00:02:21 Topics to be covered during the episode
00:03:39 How do we approach the brain from the theoretical frame?
00:06:03 Experimental setups in theoretical neuroscience
00:07:47 Q-learning paradigm – cornerstone of the brain reinforcement learning
00:18:09 Classical vs. operant learning
00:21:37 The need of using different heuristics
00:22:38 How does one think of decision making in humans and in animals?
00:28:40 Can one relate not having the ability to learn to the Kahneman and Tversky prospect theory?
00:31:09 How does Bayesian inference come into play in terms of decision making?
00:37:38 How does Prof. Dayan see memory?
00:41:20 What happens in the brain when we remember something and when we try to visualize the future?
00:43:46 How does computational modelling address accessing memory?
00:48:08 Semanticization of memory is a limited way of doing memory: the story of the patient Jon in London (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2937213/)
00:51:40 What is the relationship between time and memory?
00:58:12 The role of dopamine in decision making
01:03:37 Dopamine detox trend
01:07:18 To what extent do we need to understand the complexity of the brain in order to understand decision making?
01:12:31 What can the different modalities of biological neuroscience enrich computational modelling?
01:16:11 What will the next couple of years bring to neuroscience and AI?
01:19:35 Predicting the future based on our behaviour

This episode was recorded during the Neurizons conference in Göttingen.

Neuroscience and Beyond team:
Svilen Georgiev
Kristina Jevdokimenko
Ahsen Konaç Sayıcı
Beatriz Apgaua
Mels Akhmetali

#theoretical #neuroscience #bayesian #inference #dopamine #decision #making #reinforcement #learning