Yuandong Tian: Inside-out interpretability: training dynamics in multi-layer transformer

Yuandong Tian: Inside-out interpretability: training dynamics in multi-layer transformer
Share:

Download MP3


Similar Tracks

Nicholas Carlini: The security of LLMs UC Berkeley EECS
John Martinis: Advanced Fabrication of Superconducting Qubits for a Quantum Computer UC Berkeley EECS
Hough Transform | Boundary Detection First Principles of Computer Vision
Trump Thanks Qatar for Their Generous Jet Bribe & Accidentally Does a Socialism | The Daily Show The Daily Show
Transformers (how LLMs work) explained visually | DL5 3Blue1Brown
Jacob Steinhardt: Using AI to understand AI UC Berkeley EECS
Lars Rasmussen: From Theory PhD to Startup Land UC Berkeley EECS
Schiff Takes To Senate Floor to Lay Out Trump's 10 Most Corrupt Acts So Far Sen. Adam Schiff
Martin Wattenberg: Models within models - how do LLMs represent the world? UC Berkeley EECS
The Factory of Ideas: Working at Bell Labs - Computerphile Computerphile
David Bau: Interpretability and model editing UC Berkeley EECS
Bo Li: Benchmarks and evals, safety vs. capabilities, machine ethics UC Berkeley EECS
Michael I. Jordan: A Collectivist Vision for AI UC Berkeley EECS
2015 10 30 Claude Shannon MIT Video Productions External
Stuart Russell - AI: What if we succeed? NORA – Norwegian AI Research Consortium
Richard Murray - One in a Billion: How to Make Sure Autonomous Systems Are Safe (Enough) UC Berkeley EECS
11. Introduction to Machine Learning MIT OpenCourseWare
Vision Transformer Quick Guide - Theory and Code in (almost) 15 min DeepFindr