Machine Learning Street Talk (MLST)13 Des 2020

#033 Prof. Karl Friston - The Free Energy Principle

This week Dr. Tim Scarfe, Dr. Keith Duggar and Connor Leahy chat with Prof. Karl Friston. Professor Friston is a British neuroscientist at University College London and an authority on brain imaging. In 2016 he was ranked the most influential neuroscientist on Semantic Scholar. His main contribution to theoretical neurobiology is the variational Free energy principle, also known as active inference in the Bayesian brain. The FEP is a formal statement that the existential imperative for any system which survives in the changing world can be cast as an inference problem. Bayesian Brain Hypothesis states that the brain is confronted with ambiguous sensory evidence, which it interprets by making inferences about the hidden states which caused the sensory data. So is the brain an inference engine? The key concept separating Friston's idea from traditional stochastic reinforcement learning methods and even Bayesian reinforcement learning is moving away from goal-directed optimisation.

Remember to subscribe! Enjoy the show!

00:00:00 Show teaser intro

00:16:24 Main formalism for FEP

00:28:29 Path Integral

00:30:52 How did we feel talking to friston?

00:34:06 Skit - on cultures (checked, but maybe make shorter)

00:36:02 Friston joins

00:36:33 Main show introduction

00:40:51 Is prediction all it takes for intelligence?

00:48:21 balancing accuracy with flexibility

00:57:36 belief-free vs belief-based; beliefs are crucial

01:04:53 Fuzzy Markov Blankets and Wandering Sets

01:12:37 The Free Energy Principle conforms to itself

01:14:50 useful false beliefs

01:19:14 complexity minimization is the heart of free energy [01:19:14 ]Keith:

01:23:25 An Alpha to tip the scales? Absoute not! Absolutely yes!

01:28:47 FEP applied to brain anatomy

01:36:28 Are there multiple non-FEP forms in the brain?

01:43:11 a positive conneciton to backpropagation

01:47:12 The FEP does not explain the origin of FEP systems

01:49:32 Post-show banter

https://www.fil.ion.ucl.ac.uk/~karl/

#machinelearning

Denne episoden er hentet fra en åpen RSS-feed og er ikke publisert av Podme. Den kan derfor inneholde annonser.

Episoder(271)

What Most People Get Wrong About Evolution | Akarsh Kumar

A lot of people in AI treat evolution as a dumb fallback, basically random search for when you can't take a gradient. Akarsh Kumar thinks that is wrong. Selection hangs on to partial solutions, so mut...

10 Okt 47min

How a Voice Agent Learns the Rhythm of Conversation — Shawn Wen

Tsung-Hsien (Shawn) Wen, CTO of PolyAI, tells Tim Scarfe why voice agents are harder than text agents. Voice adds time, and a good conversation depends on adapting to the person on the line, not just ...

1 Okt 1h 10min

Who Checks a Proof No Human Can Read? — Leo de Moura

Leonardo de Moura created Lean and co-created Z3. ---This episode is sponsored by Parallel.Parallel, where agents find answers: web search, extraction and deep research APIs built for AI agents.Start ...

30 Sep 1h 14min

When AI Research Starts Moving Faster Than Human Research - Zhengyao Jiang

Weco let an AI coding agent rewrite the harness around another agent for eight days: its code, prompts and tools, while the underlying language model stayed fixed. Tim Scarfe asks Weco co-founder Zhen...

26 Sep 43min

How Deep Learning Finally Cracked Messy Tables - Frank Hutter

Frank Hutter, co-founder of Prior Labs, talks about TabPFN, a tabular foundation model that makes predictions in a single forward pass, and the research behind it.TabPFN is pre-trained on synthetic da...

23 Sep 1h 53min

Why Scaling Prediction Cannot Create Intelligence - Alexander Mattick

Alexander Mattick is a researcher at Fraunhofer IIS and a PhD researcher at the University of Technology Nuremberg (UTN), and a regular on Yannic Kilcher's Discord. He first came on MLST in 2022, afte...

21 Sep 2h 14min

How Physical AI Learns Across Language, Video and Action — Ming-Yu Liu

The car making a left turn at the start of this episode was never filmed. Cosmos 3 generated it. Ming-Yu Liu, who leads the Cosmos research at NVIDIA, explains how one model can describe a video, gene...

15 Sep 25min

Speech Recognition Is Not a Solved Problem — Pavan Kumar Reddy

Pavan Kumar Reddy leads audio research at Mistral AI. He joins Tim Scarfe for a deep technical tour of Voxtral — and explains why the frontier of deployed voice is still a cascade of specialised model...

14 Sep 1h 42min