Wednesday, March 11, 2015

Written in Our Stars

Many of the debates I have had here with Lee concern the effectiveness of predictions based on ad hoc models. If I understand correctly, Lee always dismisses these, sometimes saying that they are no better than astrology. There are other kinds of models, of course, like Newton's celestial mechanics, that are precisely predictive and well tested, and I'm pretty sure we agree on these.

Astrology is an interesting case. It has a definite model, with lots of history and data, and it's based on celestial mechanics - sort of. It holds that our fates are written in the stars, or rather, in the positions of the planets among those stars. It has a couple of flaws, the first being that it is based on an obsolete conception of the planets as gods in the Indo-European pantheon, when in fact they are just balls of rock and gas. The second and quite fatal flaw is that it is utterly refuted by the data. Our fates have an utterly negligible correlation with the positions of the planets at our birth.

My interest in more ad hoc models is based on my experience that they are frequently but hardly invariably quite effective, and my belief that prediction from such models is exactly what the human brain was evolved for, and more importantly, that one can predict something about the how effective such models are likely to be based on their ingredients. Moreover, such predictions are the basis of all our planning, personal, business, national and global.

That's why I'm impatient with critiques of predictions that are simply dismissive. I don't find that very interesting. More interesting is asking why von Neumann, for example, was wrong about some important developments in the future of computing (though right about many more) while Gordon Moore was so right about the future of electronics. The answer, I think is clear, and it has nothing to do with their relative IQs. It's that Moore knew about the integrated circuit - the key enabling technology for ubiquitous electronics and computing.

Any prediction about the future can be invalidated by a drastic change in circumstances. That kind of disruptive change happens, but usually not too often. Moreover, in technology it seems to happen in only one direction, in that technology is far more likely to open new opportunities for us than foreclose old ones. Thus negative predictions (like von Neumann allegedly thinking computers would always cost many millions, or Monod thinking we would never be able to manipulate the genome are far more likely to be invalidated than positive predictions.

Of course positive predictions (We will all fly around in rocket cars by the year 2000) are worthless unless they are based on sound physics, economics, etc.