All your base are belong to J. K. Rowling. 1,000 REVIEWS IN 26 DAYS WOOHOO AWESOME POWA! 30 DAYS 1,189 REVIEWS COMBO IS CONTINUING! YEAH! YOU PEOPLE ARE THE BEST! THIS IS SPARTAAAAA! Ahem. The third-g
Can You Trust An AI Press Release? Lawrence Chan Of course not. Here’s how leading AI labs mislead consumers, journalists, and each other. Every month or two, an AI lab releases a new model that they
Epistemic Status: Soldier mindset. These are not (necessarily) our actual positions, these are positions we were randomly assigned by a coin toss, and for which we searched for the strongest arguments
Is AI takeover like a nuclear meltdown? A coup? A plane crash? My day job is thinking about safety measures that aim to reduce catastrophic risks from AI (especially risks from egregious misalignment)
Epistemic status Still trying to work out my thoughts on this. Things change pretty regularly. My current thinking on technical AI safety questions and threat models likely diverges by now reasonably
Aug 2023 Comments at substack. Here’s a rant from my evil twin Tyromight. Say that when people apply for their first driver’s license, 1% get Executive Platinum licenses. For life, they get free use o
Oct 2023 This is the first part of an article that just appeared in Asterisk magazine. You can read the whole thing here. One of the first things internet writing teaches you is that you don’t get to
Sep 2023 I don’t sense that I’m viewed as particularly skilled at human interaction. Still, some poor fools sometimes ask me for advice, and I find myself repeating the same little speech. For context
After spending a lot of time with language models, I have come to the conclusion that tokenization in general is insane and it is a miracle that language models learn anything at all. To drill down in
This is a guest post by Max Buckley, a software engineer at Google and fellow AI researcher. By some twist of fate, this blog has become the chronicle of the evolution of integer tokenization. In an e
Obligatory disclaimer: This post is meant to argue against overuse of infohazard norms in the AI safety community and demonstrate failure modes that I have personally observed. It is not an argument f
NB. I am on the Google Deepmind language model interpretability team. But the arguments/views in this post are my own, and shouldn't be read as a team position. “It would be very convenient if the ind
Whether you’re about to start house-hunting, or planning your savings for the next 5-10+ years, it’s helpful to understand how mortgages work and how much you’re likely to be able to borrow, as this w
What do you worry about more: Getting exercise, eating vegetables, or the air you breathe? While most things that clearly improve health are well known, one is insanely underrated: Fixing your air. I
Say you want to plot some data. You could just plot it by itself: Or you could put lines on the left and bottom: Or you could put lines everywhere: Or you could be weird: Which is right? Many people t
Jun 2024 Comments at substack. Mindset matters more than where you go. Who you go with matters more than where you go. After seeing each other for a few months, many new couples take a short trip, whi
A friend has spent the last three years hounding me about seed oils. Every time I thought I was safe, he’d wait a couple months and renew his attack: “When are you going to write about seed oils?” “Di
Updated Oct 2022 1. You’re in the mood for destruction. One day, you hear about this phenomenon of “radiation” where matter gives off energy. You think—perhaps you can harness this property of nature
bilal