The Butterfly Effect

In 1961, Edward Lorenz was trying to prove that a relatively new invention called the computer could be used to predict the weather. He set up a program with a very simplified representation of the atmosphere. This consisted of a few heat exchange equations that were iterative--input one number, get an output, the output is then input, and so on.

After a the program had been running for a little while, Edward actually started to see a pattern in numbers. This was really exciting news! It meant that it was perhaps not impossible, as people had so far thought, to accurately predict the weather even a long time in advance! Imagine never getting rained out of a ball game ever again, or imagine knowing when a hurricane would strike New Orleans.

Then the computer crashed one day. Once it was back up and running, Edward took the last number from the paper readout of his program and input it back into the program. The mock weather patterns in the computer should have behaved just as predictably as they had before, but they went seemingly haywire after a few iterations and behaved entirely differently.

What happened? Surely with the input of the last output of numbers, the program should have resumed its familiar patterns. Actually, Edward soon realized that the paper output only printed the numbers to three decimal spaces in order to save space. The computer program itself recorded the numbers to six decimal spaces. Would a miniscule millionth of a unit really have an effect? Yes. Over time, that tiny fraction would multiply and make a bigger and bigger difference.

Thus, Edward Lorenz concluded that it would be impossible to accurately predict the weather in the real world because nobody could ever account for all of the tiny variations in air currents that would multiply to affect weather patterns. Put more eloquently, a butterfly flapping it's wings in Brazil could cause a tornado in Texas a month later.

This idea of little variations that make big differences came to be known as the Butterfly Effect.