PREDICTIONS – 10 Years Later - Santa Fe Institute

1. SCIENCE AND FORETELLING
In my experience the simplest and best Time Series approach is to
graph the data and extrapolate their form by eye. At any rate, the endeavor—extrapolating
a series of data points without understanding
their shape—can shed light on only the very near future. It does little
more than the naive “meteo” method: Tomorrow will be the same as
today. Time Series models perform no better on long-range forecasts
than those of economists and meteorologists. Using science this way has
proved of little help.
The scientists who have systematically shied away from forecasting
are physicists. Until recently the best of them, attracted by glittering particle
physics, swarmed around nuclear physics laboratories, shunning
meteorology and economics as pseudo-sciences. Some venerated theorists
openly admitted physics’ inability to make predictions about
phenomena as complicated as turbulence, which is extremely sensitive
to initial conditions. In Gleick’s Butterfly Effect, a butterfly stirring the
air today in Peking can produce storms next month in New York. In a
science fiction story, a time traveler steps on an insect millions of years
ago and upon coming back to the present finds an entirely different
world. The awesome complexity of such systems turns away physicists
who seek elegant solutions.
Classical physics is excellent at describing billiard ball movements.
The difficulty comes from putting many balls together (many, for physicists,
can be anything greater than three). Molecules in a volume of gas
behave very much like billiard balls, but there are too many of them and
they bounce too often. Thermodynamics, the branch of physics which
studies gases, makes predictions by focusing on the macroscopic global
variables only: temperature, pressure, and volume. The bottom-up
approach, tracking individual molecules, taxed the ingenuity of the best
minds in physics for at least a hundred years and has served only in
understanding, corroborating, and justifying the relations established
experimentally between the overall variables.
Squeezed by diminishing returns, physicists have begun to scatter
outside the musty dungeons of particle accelerators in recent years. They
are showing up in unlikely places: neurophysiology, Wall Street, chaos
studies, and forecasting, to name a few. They carry with them detectors,
tools, techniques, and tricks, all built around the scientific method: observation,
prediction, verification. They are striving for understanding
24