AXRP - the AI X-risk Research Podcast

38.2 - Jesse Hoogland on Singular Learning Theory

Nov 27, 2024

Jesse Hoogland, executive director of Timaeus and researcher in singular learning theory (SLT), shares fascinating insights on AI alignment. He dives into the concept of the refined local learning coefficient (LLC) and its role in uncovering new circuits in language models. The conversation also touches on the challenges of interpretability and model complexity. Hoogland emphasizes the importance of outreach efforts in disseminating research and fostering interdisciplinary collaboration to enhance understanding of AI safety.

18:18

Creator website

Episode guests

Jesse Hoogland

AI Summary

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

Singular Learning Theory (SLT) enhances our understanding of neural networks' generalization capabilities and informs AI safety practices.

The refined local learning coefficient offers a novel approach to investigate neural network behavior, improving interpretability and identifying risk factors.

Deep dives

Applications of Singular Learning Theory

Singular Learning Theory (SLT) provides a framework for understanding why neural networks generalize effectively, which is crucial in evaluating their predictive capabilities in real-world applications. This theory helps improve evaluation benchmarks to ensure they are predictive of actual behaviors once models are deployed. Furthermore, SLT addresses interpretability issues, such as identifying when a model may execute risky decisions, like a treacherous turn during operation. The advancements in SLT are intended to create tools that can probe and enhance the safety of AI systems.

Intro

2min

Exploring Singular Learning Theory and Outreach Efforts

2min

Exploring the Outreach and Concepts of Singular Learning Theory

4min

Understanding Model Complexity in SLT

6min

Exploring Multigram Circuits and Layer Coordination in Neural Models

4min

You may have heard of singular learning theory, and its "local learning coefficient", or LLC - but have you heard of the refined LLC? In this episode, I chat with Jesse Hoogland about his work on SLT, and using the refined LLC to find a new circuit in language models.

Patreon: https://www.patreon.com/axrpodcast

Ko-fi: https://ko-fi.com/axrpodcast

The transcript: https://axrp.net/episode/2024/11/27/38_2-jesse-hoogland-singular-learning-theory.html

FAR.AI: https://far.ai/

FAR.AI on X (aka Twitter): https://x.com/farairesearch

FAR.AI on YouTube: https://www.youtube.com/@FARAIResearch

The Alignment Workshop: https://www.alignment-workshop.com/

Topics we discuss, and timestamps:

00:34 - About Jesse

01:49 - The Alignment Workshop

02:31 - About Timaeus

05:25 - SLT that isn't developmental interpretability