johnwkrakauer

Policy-Gradient Reinforcement Learning as a General Theory of Practice-Based Motor Skill Learning Mastering any new skill requires extensive practice, but the computational principles underlying this learning are not clearly understood. Existing theories of motor learning can explain short-term ad...

New Pre-Print:
www.biorxiv.org/cgi/content/...

We’re all familiar with having to practice a new skill to get better at it, but what really happens during practice? The answer, I propose, is reinforcement learning - specifically policy-gradient reinforcement learning.

Overview 🧵 below...

New preprint written with the wonderful philosopher William Ramsey: Mental Representation Without Neural Representation: Understanding The Evidence osf.io/preprints/ps...

First shot across the bow from ongoing project with Jake.

Of course. It’s extraordinarily that we even need to have this discussion - again - I despair.

‪@benhayden.bsky.social‬
@tyrellturing.bsky.social‬
@jmgrohneuro.bsky.social‬
@pessoabrain.bsky.social
I see a lot of talk on here about how we should avoid
"x does y" talk because the brain is "a dynamic, reverberant, reciprocally interconnected system".
But this does not follow.
A thread...

Excited to share this new work

www.cambridge.org/core/journal...

Agree, it is not a useful term at the level of neuroscience. It is an undeniable capacity at the psychological level. What the neuroscience of this capacity will look like is anyone’s guess.

What do neuroscientists mean by the term representation? A group of neuroscientists and philosophers discuss the use and misuse of the term “representation” across the cognitive sciences.

Terrific podcast relevant to our debates here about “What is an emotion?” But in the case of emotion, it’s turned up to 11 b/c (unlike “representation”), everyone alive has intuition and interest about the answers (including the public).

www.thetransmitter.org/brain-inspir...

Audience Faves: Hasok Chang on 'Epistemic Iteration' The HPS Podcast - Conversations from History, Philosophy and Social Studies of Science · Episode

Great interview with Hasok Chang on 'Epistemic Iteration':

The idea that we don't often start scientific inquiries from a solid foundation. We knowingly start from an imperfect position, and use the outcomes to refine and correct the original starting point.

open.spotify.com/episode/6tbT...

...it basically confirmed what is already well-established: LLMs (& LRMs & "LLM agents") have trouble w/ problems that require many steps of reasoning/planning.

See, e.g., lots of recent papers by Subbarao Kambhampati's group at ASU. (2/2)

It was fun working on this with David and Melanie.