Sign in

Chris McMaster

@rheumai.com
462 followers 815 following 158 posts

Rheumatologist, builder of AI things. Carbon over silicon.

PostsRepliesMedia
Chris McMaster @rheumai.com · 15/12/2024
Sora’s idea of a hand exam. This 60 year-old rheumatologist is giving me very 2nd year medical student vibes with this bizarre technique. No synovitis was detected this day. #rheumsky
120
Chris McMaster @rheumai.com · 09/12/2024
Are we the baddies?
000
Chris McMaster @rheumai.com · 03/12/2024
This seems trivial and unhelpful.
440
Chris McMaster @rheumai.com · 02/12/2024
010
Chris McMaster @rheumai.com · 02/12/2024
#ACR24 people would anyone be interested in a Chrome extension that adds the top 5 most similar abstracts at the end of the page.
250
Chris McMaster @rheumai.com · 30/11/2024
Then you agree with my son. Obviously he can’t read, which made him immune to the priming effect of the text. It’s a good example of a more subtle version of the old “ipod” written on an apple adversarial attack.
210
Chris McMaster @rheumai.com · 30/11/2024
@vikhyat.net moondream is siding with my son.
000
Chris McMaster @rheumai.com · 30/11/2024
My 2-year-old and I are debating: brown bear or wombat?
350
Chris McMaster @rheumai.com · 28/11/2024
Rheumatology-adapted version of Llama 3.1 8B, running on my MacBook at 20 tok/s, correctly answering a question that no open weights models other than Llama 405B can answer. No instruction tuning involved, lots of synthetic data.
A variant of Llama 3.1 8B answering a difficult rheumatology question (Answer: test UBA1 gene, because the patient has VEXAS)
280
Chris McMaster @rheumai.com · 25/11/2024
Optimist: The cup is half full. Pessimist: The cup is half empty. AI:
120
Chris McMaster @rheumai.com · 25/11/2024
AI generated gaits are very odd.
010
Chris McMaster @rheumai.com · 23/11/2024
You could simulate the effect of a drug on the body, or you could talk to a chatbot obsessed with the Golden Gate Bridge. One of these applications is just a little bit funnier.
Features in a model mapped to Sparse Autoencoders can be manually “activated” (set to a minimum threshold) to modify the model’s behaviour. 
In cellular models this could involve activating specific pathways to simulate biological processes or observing the effects of a simulated drug. 
In LLMs this might involve constraining the model to generate content exclusively about the Golden Gate Bridge.
000
Chris McMaster @rheumai.com · 22/11/2024
Giving a talk on AI to rheumatologists next week. It’s really just an excuse to talk about Golden Gate Claude.
220
Chris McMaster @rheumai.com · 20/11/2024
Just in case you wanted to know its assumptions.
010
Chris McMaster @rheumai.com · 20/11/2024
Better than o1 preview 😂
200