[paviva]

Unlikely.

Diagnosis is based on imperfect data with very high noise-to-signal ratio, and with auditory (patient's history), visual and tactile inputs. Treatment often need to be tailored for each patients unique needs, goals, and co-existing diseases.

[aaronwalker]

Right, but the unique profile (BMI, glucose levels, heart rate patterns) is something that an ML system could do better than any doctor. And right now the way medicine is distributed is by undergoing intense clinical trials on the general public. But, could drugs not be manufactured for the needs of each specific profile? Seems a matter of data intelligence and 3D printing on the manufacturing side to me.

[paviva]

You're assuming that more data will reveal a hidden pattern, but the fact is that most data, and especially most of the easily collectable data is useless junk.

[yrui]

I think the idea is that if you aggregate enough of this junk, some of the noise is canceled out--possibly enough to recover the signal.