Joining Anthropic

I'm excited to announce that I'm joining Anthropic, starting this week! I'm honoured to be joining such a capable and kind group of people, who I've long admired - Claude was the first LLM that I found myself consistently using. Recently I've been really blown away by Artifacts and Computer Use, and the consistent improvements in Claude's skills.

I've also been incredibly lucky to be part of an amazing journey at Google DeepMind for the past 10 years, where I got to work on more exciting projects than I would have ever dreamed of: the saga from AlphaGo to ...

SSD NAS

Recently I decided to build a NAS to store my RAW photos as well as video content, and to use as a backup target for my MacBook. I wanted something that I could use directly with low latency for interactive editing over 10 GBit ethernet, so spinning disks were out.

However, building an all SSD solution turned out to be surprisingly tricky - the old SATA standard is much too slow for modern SSDs, but getting more than a few PCIe NVMe ports on a consumer mainboard is quite difficult (I wanted to avoid server boards to minimize power ...

Being Productive

Recently I was asked which habits and practices were the secret to my productivity. I figured my answers might be of wider interest. Let me know what works for you!

Why should you listen to my advice? Aside from my papers, in each year of the past decade I've authored ~1k pull requests with ~200k LoC of deltas, with a similar number reviewed:

You can tell when I'm on holiday :)

Learning

My first recommendation is to learn as much as possible, about the entire stack you work with - in AI, ...

Questions about AlphaGo Development

I was asked some questions about the development of AlphaGo and figured my answers might be of wider interest, so here we go:

How low-level does the code need to be in order to get the maximum performance both in training and, possibly, live game?

To maximize playing strength it is necessary to balance maximum utilization of the accelerators (GPUs or TPUs) and focusing on exploring and evaluating the most promising line of play. Unfortunately there are several conflicting tendencies:

• The more board positions we can evaluate the better our estimates of how good each ...

Custom Static Site Generator

Once again I've migrated this blog to a new backend - originally I had been using Octopress, then later I migrated to Pelican after I got fed up with the complicated setup and dependencies. Well, recently I had issues getting Pelican to work after some updates, so I decided to bite the bullet and write my own small static site generator: 開板.

Really it's just an excuse to play around with Rust and make it easier to customize my site, but if it means no more painful upgrades that's a nice side effect :D It also gave me a ...

The Strong Turing Test

In the conventional Turing test (aka imitation game), an investigator tries to distinguish between a human and a computer solely by interacting with them.

This is an interesting setup and has inspired much research, but it doesn't immediately translate into practical usefulness - a computer system may pass as human, but may still not be able to help me accomplish any task.

Instead, I find I'm mostly interested in a stricter variety: in each interaction the investigator chooses a preferred response; the goal of the computer system is to be chosen as the preferred side as many times ...

Saving Aranet4 data to Raspberry Pi

Recently I was curious about the amount of CO2 present while sleeping and working; I ended up buying an Aranet4 monitor. It's great to see the reading live, but I would also like to compare trends over time. The Aranet4 itself can only store a maximum of 7 days of data, so for longer term analysis I need to periodically export the data.

This is where a Raspberry Pi comes in handy! First, install the necessary packages:

sudo apt install bluetooth pi-bluetooth bluez blueman

Use bluetoothctl to pair the Aranet4:

sudo bluetoothctl
> ...

Discovering Matrix Multiplication Algorithms with AlphaTensor

Matrix multiplication is at the foundation of modern machine learning - whether transformers or convolutional networks, diffusion models or GANs, they all boil down to matrix multiplications, executed efficiently on GPUs and TPUs. So far the best known algorithms have been discovered manually by humans, often optimized for specific use cases.

The most famous is probably the Strassen algorithm to multiply two 2x2 matrices using only 7 instead of the naive 8 multiplications:

Through clever addition and subtraction of the individual elements of the a and b matrices this algorithm is ...

Planning in Stochastic Environments with a Learned Model

After extending to arbitrary action spaces and offline RL, we recently published our next step in generalizing MuZero: Stochastic MuZero, which learns and plans with a stochastic model of the environment that can model not only a single deterministic outcome of our actions, but rather a full set of possibilities.

Previous approaches such as AlphaZero and MuZero have achieved very good results in fully-observed deterministic environments, but this type of environment is insufficiently general to model many problems and real world tasks. Beyond the stochasticity inherent in many domains (think roll ...

Our lives are measured in memorable moments

The reason our childhood years seemed to pass so much more slowly than our adulthood is that they were filled with novelty and excitement, memorable first experiences: learning to ride a bike, entering a new school, making new friends.

Events, experiences, things are memorable not by some inherent quality, but rather become memorable by virtue of being different from our everyday life. What matters is the magnitude, not the sign: both the best and the worst times in your life will stick out in your memory.

Memorable experiences come in all sizes: cooking a new dish, meeting up ...