Seen on the classiccmp mail list, where @bdk6 quotes from Wikipedia on Chaos Theory:
“Edward Lorenz was an early pioneer of the theory. His interest in chaos came about accidentally through his work on weather prediction in 1961. Lorenz was using a simple digital computer, a Royal McBee LGP-30, to run his weather simulation. He wanted to see a sequence of data again, and to save time he started the simulation in the middle of its course. He did this by entering a printout of the data that corresponded to conditions in the middle of the original simulation. To his surprise, the weather the machine began to predict was completely different from the previous calculation.”
@jecel clarifies that this was a discovery, not an error:
He dumped the full state of the simulation to paper with six digits after the decimal point even though the internal calculations used eight digits (I don’t remember the actual precisions involved). So when he restarted the simulation from the middle he introduced errors of less than 1 per million and fully expected the results to be the same for the days he had already simulated so he could continue a little further. But he was shocked that the simulation went in a different direction and the results were totally different after only a few days.
This is an absurd sensitivity to initial conditions that had never been noticed in any system before. He compared it to whether a butterfly flapped its wings or not in the middle of the Amazon making a difference on there being a nice day or a huge storm on the other side of the world a week later. This is the infamous “butterfly effect”.
All this came after eliminating all kinds of possible errors, of course. The first thing we thought back then when something like this happened was not “I found a new theory” but “the hardware is probably flaky or there is a compiler bug”.
There are links in the thread to Lorenz’ original paper, a later more substantial paper, and an online copy of James Gleick’s popular science book “Chaos”
We’ve mentioned the LGP-30 previously.