AI Alignment Forum

A community blog devoted to technical AI alignment research

Popular posts: 4
Total points: 864
HN comments: 397
Rank: #1,529 company

Visit alignmentforum.org → RSS feedOn Hacker News Dec 2019 – Nov 2023

Math & Algorithms 4 AI & Machine Learning 3 Life, Learning & Writing 1

Most points Most discussed Most cited Newest

A Mechanistic Interpretability Analysis of Grokking

2023202 points54 commentscited by 5AI & Machine Learning details

Neel Nanda reverse engineers neural networks that have "grokked" modular addition, showing that they operate using Discrete Fourier Transforms and tr…

When do "brains beat brawn" in chess? An experiment

2023124 points79 comments AI & Machine Learning details

Understanding “Deep Double Descent”

2019108 points16 comments AI & Machine Learning details

Double descent is a puzzling phenomenon in machine learning where increasing model size/training time/data can initially hurt performance, but then i…