I just saw a headline mentioning IBM Watson being used in Healthcare. Do we suffer from Deep-O-Philia in Healthcare in the same way as the rest of AI applications? Presumably NOT.
Healthcare correlations are not about intelligent action so much as powerful multi-variate statistical tools. So hype is not a problem and healthcare is exactly where "deep" data mining could be useful. So the question becomes: how good is Watson's multi-variate correlation engine?
I know from personal experience (implementing Data Equilibrium) that tabulating and storing things like 5 variable correlations is not an easy algorithm. I wonder if they know how to do it?
NB: I implemented this algorithm inside a SourceForge project called "Data Equilibrium" and I filed a patent for the algorithm that got rejected because I used the word "dotted" - as in dotted line - and the reviewer tossed the application.
Update: See later posts about deep AI and healthcare. I think what we learn is that correlation is better when supported by a model. And we learned that IBM's Watson may be using such actual models which, amazingly, they extract somehow from correlations found in language.
Those bastards! That's the right direction.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment