At the moment, for all intent and purposes, yes they do smile
  emoticon We can still try to predict them and toss all the math
  we want at it and I believe we _should_ [just as we should with
  the mind] but I believe we have a LONG LONG way to go before we
  can predict things like this with near-certainty. I honestly
  believe it may well be impossible utilizing the ways-of-thinking
  we have been/currently are using. There are fundamental flaws in
  our approaches towards it that reflect in our predictive
  capabilities. Not that i think impure functional programming in
  parallel message passing processing will fix it, but it'd be a
  start. But with millions of lines of FORTRAN to go through, it's
  the best we got for now smile emoticon