Chris J. Maddison (@cjmaddison) / X

What if the next medical breakthrough is hidden in plain text? Causal estimates drives progress but data is limited & RCTs slow. Introducing NATURAL: a pipeline for causal estimation from text data in hours, not years. Paper: tinyurl.com/ppr29 Site: tinyurl.com/web98

A figure comparing the type of data used in RCTs vs. our pipeline, NATURAL. It also include a table that shows that in 4 real-world settings, NATURAL estimates came within 3 percentage points of a real-world RCT.

184K

Chris J. Maddison

@cjmaddison

Aug 4, 2024

Enough already: T cell inflammation and SARS-CoV-2 virus persist in Long Covid. Editor's choice in @ScienceMagazine Immunology from Friday.

science.org

Enough already: T cell inflammation and SARS-CoV-2 virus persist in Long Covid

Virus and T cell inflammation persist in the tissues of patients with Long Covid.

Why long COVID may simply be 'long infection' It is very important to make the message simple and optimistic: it seems that some people have long SARS-CoV-2 infections. We now know what we need to do.

An electron microscope image of a COVID-19 particle

Why long COVID may simply be 'long infection' - ABC listen

🔥 "Skeptics beat their chests about evidence-based medicine and never grapple with what patients should do when medicine doesn't bother to create an evidence base for your condition." --@edyong209 @SFU 🔥

“Finally, maybe this is controversial but ultimately progress in science is bottlenecked by real-world experiments.” If this is controversial in SF, we’re cooked.

Jason Wei

@_jasonwei

Jun 30, 2025

We don’t have AI self-improves yet, and when we do it will be a game-changer. With more wisdom now compared to the GPT-4 days, it's obvious that it will not be a “fast takeoff”, but rather extremely gradual across many years, probably a decade. The first thing to know is that

Today is the first time I've taught since 2021 and the first time teaching in-person. We'll be studying large language models through a close read of the Llama 3 tech report. I'm grateful to be here, and excited for the term. cs.toronto.edu/~cmaddis/cours…

Merry Christmas, friends and colleagues! I've written a short letter reflecting on our progress. Thank you for everything. cs.toronto.edu/~cmaddis/lette…

Letter text can be found here: https://www.cs.toronto.edu/~cmaddis/letters/dear_friends_2024.pdf

What makes a great scientist? Most AI scientist benchmarks miss the key skill: designing and analyzing experiments. 🧪 We're introducing SciGym: the first simulated lab environment to benchmark #LLM on experimental design and analysis capabilities. #AI4SCIENCE #ICML25

I decided to write out a part of the story of my illness for my #LongCovidMoonshot holiday letter, which I am sending tomorrow to @RepRaulGrijalva, @SenMarkKelly, and @SenatorSinema. 🌼

There's a very strong tendency to believe that the medical system has answers for the ailments that you or your loved ones will inevitably develop. Unfortunately, it's not true. The sooner we internalize that, the sooner we fix biotech.

The rebuttal process @NeurIPSConf was too collegial and productive, so a bunch of papers whose process ended on a positive note will be rejected.

When I started in Geoff's lab at UofT in 2011, we were handrolling our neural nets in gnumpy (a simple numpy-like wrapper for CUDA written by UofT grad students). If you've ever had to code an LSTM from scratch without autodiff, you know just how cursed this workflow is.

My experience as well. The reviewer chaos at NeurIPS / ICLR / ICML degrades their prestige for the people in-the-know, who are arguably the people you want to impress.

Ahmad Beirami ✈️ ICML

@abeirami

May 11, 2025

The review quality in TMLR is better because: 1. The authors suggest the AE. This means that the AE is more likely to be the right fit for the paper. 2. The AE selects the best reviewers for the paper who may or may not be in the reviewer pool. ...

Operation Moonshot (Taylor's Theorem). Zeroth-order goal: let congress know we're here 👋 First-order goal: let congress know we're organizing 💪 Full goal: $1B / year in NIH research funding! 🚀 longcovidmoonshot.com

8.8K