The amount of blind faith people put in computer models (and in “experts”) never ceases to amaze me. Personally, I’m from Missouri. Show me. Prove it to me.
I was reading an article in a recent issue of “The Economist” that explained how the recent financial meltdown in CDO (collateralized mortgage obligations) occurred because the “complex computer models” proved inadequate, i.e. extrapolations of the past did not equal the future. Computer programers and other supposed experts proved no better at forecasting the future than your local palm reader.
(For an excellent entertaining read on the limits of complex computer models, read “When Genius Failed.” With the perspective of the recent financial crisis, it is also instructional on how quickly the lessons of the past fade from memory.)
Lesson learned, we hope. Computer models are no better than the assumptions built into them. A computer is no more infallible than the humans who program them: “Garbage In, Garbage Out.”
So, I then flipped a few pages forward in the SAME issue of “The Economist” and what did I find? An article about a complex computer model of the human body and its reactions is being created that will run simulations to speed up the testing of medicinal drugs.
I’ve got no more faith in the medical “complex computer model” than I did in the disastrous financial
“complex computer model.” Trouble is, the medical one could get people killed, not just broke.
I’m all for pushing back the frontiers of human knowledge via “complex computer models,” I just wish we would learn to be appropriately skeptical of black boxes and magical solutions.
I close with an excellent quote from Howard Marks of Oaktree Capital: “… while computers can find, verify and extrapolate relationships that have held in the past, they can’t tell when those relationships will cease to work and what new relationships will take their place. Put another way, computers know a lot about the past but much less about the future.”