Update on IBM Watson

This forum made possible through the generous support of SDN members, donors, and sponsors. Thank you.

pgoyal

Full Member
10+ Year Member
Joined
Jan 22, 2012
Messages
128
Reaction score
0
Once you get to the wards, you'll find that the challenge in medicine isn't really the decision-making that watson is supposed to replace. The real challenge is getting the correct story from the patient, correctly assessing very-imperfect clinical data with large error margins, and finally putting everything together in the context of the patient's social situation.

And then there's clinical intuition to factor in as well... Just today I heard an attending say "The data says we should do ____ but if that were me on the table, I'd want ______ instead."

I can see NPs and PAs being replaced by Watson before MDs. The real value of the MD is our ability to interpret subtleties and make decisions when the evidence isn't black or white.
 
In my pre-med years, I privately wondered how useful my training was actually going to be. I mean, these things aren't that complicated, and spending ten minutes on WebMD should give you a pretty good idea of what the diagnosis is, right? I didn't see how much value I would add, at least for my educated, internet-savvy patients.

...yeah.

I think Watson will be useful, and I'd be very interested to hear more about it from Dr. Mehta's perspective. But I don't think it will replace doctors- rather, I think that instead of being trained to tune a patient out immediately upon hearing, "but I read online...", in twenty years we will be tuning out administrators saying, "but according to Watson..."

Same stuff, different day.
 
This will increase costs like no one can imagine. I can come up with a differential diagnosis of 10 things. Watson can come up with thousands. Start working them up...
 
What would distinguish Watson from something like a WebMD symptom checker. Put info in.... check database.... use an algorithm..... get diagnosis.....
 
What would distinguish Watson from something like a WebMD symptom checker. Put info in.... check database.... use an algorithm..... get diagnosis.....

WebMD ignores a ton of parameters. This would (hopefully) have much greater specificity and sensitivity.
 
Top