Shower Thoughts

If the TV series House, MD was filmed today, it would only feature a bored 25-year-old medical student prompting a differential diagnosis LLM all day long with 100% accuracy

Aviator, diver, Etherean, nexialist, PhD in aviation ∩ climate science | Host of /diving /ethfinance /eth-staker /geopolitics /science /singapore /thomas

Beep boop ace inhibitors
Beep boop broad spectrum antibiotics
Beep boop cannot get answer if patient lie to me (literally every patient is lying in the show)

Startup artist seeking inspiration.  Champion of /bcard

They lie about causal factors (taking drugs, having an affair, etc) — they can’t lie about the symptoms in a clinical setting.

It’s precisely because the doctors in the show are human that they fall for those lies and fail to see objectively (except House, of course).

An LLM can’t get swayed by what the patient says — it will not only look at the symptoms with an open mind, but it can also be trained for lying risk profiles (e.g., a teenager is more likely to be experimenting with drugs, a middle-aged married adult is more likely to be having an affair, etc).

So I’d argue that being impervious to lying gives the AI more of an edge, not less

But soon Alexa, your Ring doorbell, and your Roomba will also contribute to the training dataset, so the LLM doesn’t even need to break into your home 😅

I dunno - episodes where they have to break into a patient's home/car and find out the source of infection doesn't hit as hard with an LLM.