2 Comments
User's avatar
Sahar Leene's avatar

Machine learning. Not really. Someone has to write the script for the AI/LLM. Our current script for sepsis (whatever that word means these days) lacks a disease model. Using "AI" to correlate thousands of sepsis data points is like trying to retrace a house fire without a degree in structural engineering. We need to go back to hard science and research microcirculation from the front end. I think p hacking and Vibe coding large data sets of models we dont understand is a dead end. I hope it isnt as we've exhausted and optimized the macroscopic model of reperfusion and infection source control. But data crunching is not a substitute for understanding replicable human physiology.

PulmCCM's avatar

I like the analogy of retracing a house fire. I have used a storm system analogy before. Both highly complex events that can’t be analyzed from the residue that remains after they are well in progress.