Essays
Democracy as an Alignment Problem
Asimov imagined Machines that run the world well and answer to no one. What if we built the opposite: an AI-mediated democracy that listens to everyone, and treats keeping it honest as an alignment problem?
On building diffusion models
With this article, I wanted to share to share some of the experiences we had building RFD3, and how this project changed the way I thought about building generative models to solve real world problems.
From Matrices to Transformers
Here, we take a brief look at some of the interpretability work for the transformer architecture, mainly based on anthropic's thread on intepreting information circuits within transformers. We build up an intuition for how information is propagated through transformer models through matrix multiplication and further ground intuition by introducing the Kronecker product notation.
Introduction
This blog consists of a small collection of projects of mine that I've turned into readable articles. Here, you'll find everything you need to know about the blog!