chief ai research @doximity. ICU doc, prof @chumontreal. prev co founder @pathwaymedical
Montréal, Québec
Joined November 2011
- Tweets4.6K
- Following3K
- Followers2.9K
- Likes6.6K
Louis Mullie, MD retweeted
We ran 8,000+ physician dosing questions through 9 popular LLMs
Over 2% of the dosing claims were wrong and almost half of those errors were serious or potentially lethal
@Doximity Ask now checks every dose before you see it
Louis Mullie, MD retweeted
𝗧𝗼𝗱𝗮𝘆 𝘄𝗲 𝗹𝗮𝘂𝗻𝗰𝗵𝗲𝗱 𝗗𝗼𝘀𝗲𝗖𝗵𝗲𝗰𝗸™ 𝗶𝗻 𝗗𝗼𝘅𝗶𝗺𝗶𝘁𝘆 𝗔𝘀𝗸, 𝗮 𝘃𝗲𝗿𝗶𝗳𝗶𝗰𝗮𝘁𝗶𝗼𝗻 𝗹𝗮𝘆𝗲𝗿 𝗳𝗼𝗿 𝗱𝗿𝘂𝗴 𝗱𝗼𝘀𝗶𝗻𝗴.
A probabilistic system shouldn’t be the final word on a deterministic question. Dosing is one of those questions. For a given drug, route, and patient population, there’s a labeled answer, and even the best LLMs get at least 2.3% of those claims wrong when answering from memory.
𝗗𝗼𝘀𝗲𝗖𝗵𝗲𝗰𝗸𝗧𝗠 parses each dosing claim into drug, dose, route, and frequency, then compares it against our drug database using explicit rules. No language model in the loop, results in under 100 milliseconds, and the same claim gets the same result every time.
𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗮𝗻 𝗮𝗻𝘀𝘄𝗲𝗿 𝗮𝗻𝗱 𝘃𝗲𝗿𝗶𝗳𝘆𝗶𝗻𝗴 𝗶𝘁 𝗮𝗿𝗲 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀, 𝗮𝗻𝗱 𝗲𝗮𝗰𝗵 𝗻𝗲𝗲𝗱𝘀 𝗶𝘁𝘀 𝗼𝘄𝗻 𝘁𝗼𝗼𝗹.
blog.doximity.com/articles/o…
Louis Mullie, MD retweeted
The most infuriating result in Mathematics will be when it turns out that P=NP but with a bounded constant so large that effectively NP will remain exponential for anything of practical importance.
Louis Mullie, MD retweeted
1/4
Meet Qwenputer: Qwen3.5-4B with a full RISC-V CPU inside its weights, using the model's own Gated DeltaNet layers as RAM.
It runs C and MicroPython in its forward passes, no external runtime.
On DeepMind math: base 55–90%, base + Python tool 95–100%, Qwenputer 97.5–100%.
Louis Mullie, MD retweeted
recipe from @_lyraaaa_ github.com/lyramakesmusic/ne…
Claude: "We wrote 5,000 questions..."
me: don't anthropomorphize yourself. 5,000 questions were synthetically generated.
me:
Louis Mullie, MD retweeted
my theory
being a founder comes with pain, ambiguity and emotional roller coasters
for ppl with trauma that's just default mode
so they can withstand more of it
and probably an unconscious drive to sublimate the pain to "fix" the original trauma
Louis Mullie, MD retweeted
The complexity theory result is actually huge.
P problems can be solved in polynomial time. It was widely believed that NP-complete problems require exponential time.
P = NP states that NP-complete problems can be solved in polynomial time. OpenAI didn't prove that. But, they proved that (some) NP-complete problems can be solved faster than exponential time, meaning:
- P = NP, or
- the complexity of NP problems is a "middle thing" between polynomial and exponential, eg. O(n^log n)
If it survives peer reviews, this could be one of the biggest CS breakthroughs in decades.
Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory).
AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…
Louis Mullie, MD retweeted
Here's a direct link to OpenAI's PDF claiming a refutation of ETH. In particular, it's claimed that 3SAT on n variables is in about exp(n/(log n)^{1/2+o(1)}) time
github.com/openai/math/blob/…
Louis Mullie, MD retweeted
💡@doximity + @UpToDate: UpToDate is coming to Ask!
🩺Clinicians at health systems using Doximity’s Ask will see an answer drawn from UpToDate’s topics and best-practice guides
Excited to try this out!
#Doximity #AIinMedicine #ClinicalAI #OncTwitter #MedTwitter
Louis Mullie, MD retweeted
Geoffrey Hinton once noticed that the tellers at his bank kept changing. When he asked one of them why, the teller didn't know, only that they got moved around a lot.
Hinton guessed the reason... if staff never work with the same people for long, it's much harder for a few of them to team up and cheat the bank.
That thought became one of the most useful tricks in deep learning. While a neural network trains, you randomly switch off about half its neurons at every step. No neuron can lean on a particular partner, so each one has to learn something useful on its own. The trick is called dropout, and it helped AlexNet win the 2012 ImageNet contest, which set off the modern AI boom. It's still built into many of today's networks.
Louis Mullie, MD retweeted
This happened because a radiologist is working their ass off to get cancer scans read at 11 pm on a Friday before the weekend and it is Federal law that results are immediately available to the patient in the electronic record
Healthcare is the only industry that drops your cancer scan results into an app at 11 PM on a Friday, then closes for the weekend.
No call. No context. Just a PDF full of words like "suspicious" and "cannot be excluded" and a little banner that says "Sleep well!"
Meanwhile billing found me in 4 hours 😐
Louis Mullie, MD retweeted
We have barely scratched the surface of GPUs, and we think GPUs are more flexible and valuable than most believe. Internally, we already see evidence of this at trillion-parameter pretraining scale. If we are right, this will drive substantially more demand for AI.
Louis Mullie, MD retweeted
"As Doximity ramps up its investment in clinical AI tools, the partnership with Wolters Kluwer positions [Doximity] Ask as the front door to clinical knowledge."
From @HeatherLandi at @FierceHealth on our partnership with @Wolters_Kluwer's @UpToDate.
fiercehealthcare.com/ai-and-…
Louis Mullie, MD retweeted
@UpToDate + Doximity = one search, more behind every answer.
AI Fast & Physicians First. Excited!
