How might LLMs store facts | Deep Learning Chapter 7

From the creator

Unpacking the multilayer perceptrons in a transformer, and how they may store facts Instead of sponsored ad reads, these lessons are funded directly by viewers: https://3b1b.co/support An equally valuable form of support is to share the videos. AI Alignment forum post from the Deepmind researchers referenced at the video's start: https://www.alignmentforum.org/posts/iGuwZTHWb6DFY3sKB/fact-finding-attempting-to-reverse-engineer-factual-recall Anthropic posts about superposition referenced near the end: https://transformer-circuits.pub/2022/toy_model/index.html https://transformer-circuits.pub/2023/monosemantic-features Some added resources for those interested in learning more about mechanistic interpretability, offered by Neel Nanda Mechanistic interpretability paper reading list https://www.alignmentforum.org/posts/NfFST5Mio7BCAQHPA/an-extremely-opinionated-annotated-list-of-my-favourite Getting started in mechanistic interpretability https://www.neelnanda.io/mechanistic-interpretability/getting-started An interactive demo of sparse autoencoders (made by Neuronpedia) https://www.neuronpedia.org/gemma-scope#main Coding tutorials for mechanistic interpretability (made by ARENA) https://arena3-chapter1-transformer-interp.streamlit.app/ Звуковая дорожка на русском языке: Влад Бурмистров. Sections: 0:00 - Where facts in LLMs live 2:15 - Quick refresher on transformers 4:39 - Assumptions for our toy example 6:07 - Inside a multilayer perceptron 15:38 - Counting parameters

Published 2024-08-31 · Watch on YouTube ↗
Availability and playback restrictions are controlled by the original channel.