Is wrestling my mind
I've got a strange feeling
I'm going to die before my time
-- Chris Hillman, Roger McGuinn, and Michael Clarke
OK, so a bit of misdirection. As if missive titles in these endeavors ever did that? A piece on Stat News contains this nifty graph:
Healthcare is a prime venue for AI?
So, why might that be? Kind of simple: what passes for much of what is called AI is just minimized squared differences in many ways. Computers are really good at that sort of thing. Given that health diagnosis is all about mapping myriad symptoms to a single cause, doesn't massive correlation analysis have a chance to top human experience and cognition? Of course it can. The thing about health diagnosis is that it might appear to violate these endeavor's prime directive: correlation analysis, esp. time series version, dealing with human venues is a waste of time. Just read the New York Times and Wall Street Journal and professional journals. Right? Well, no. These data are generated by Mother Nature, so it's merely a matter of sufficient understanding of Her rules. Humans are the target of the data generating process, not the source. Which process, in most of the remaining dark corners of medicine, is really dark.
No comments:
Post a Comment