Re:Ask YC: Real World Applications of Machine Learning?
by rms on Saturday March 08, 2008 @07:59PM
(#132266)
I met a guy at a university commercialization seminar that was trying to make ML/decision tree stuff for medical diagnoses and such. I don't think he was thinking of triage but much more basic decisions. I hope he got some funding, but this was also in Pittsburgh.
Medical diagnosis is an area where ML should be used, but isn't. It's been discovered that a decision tree beats a cardiologist for heart attack triage, but no one will touch it (1). In research, it is often used to discover genetic regulatory networks, and other such things. Basically, ML is replacing what would be a boring job for a grad student.
Also look at marketing. Thanks to their ML systems, Walmart knows that people buy poptarts to prepare for a hurricane.
(1) If a cardiologist kills 10/1000 people due to human error (whoops, forgot the aspirin), it happens. If a computer program kills 5/1000 people due to classification error it's 5 lawsuits waiting to happen, provided you can find a cardiologist to say "In hindsight, I would have gotten that right."