This New York Times column contains what is no doubt the biggest (in magnitude, if not in importance) numerical error ever to appear in print, and it’s in a quote by a physicist:
For instance, if all the molecules of air in the room where you're sitting would suddenly cross to one side, you would not have any air to breathe. This probability is not zero. It is in the 10 to the minus-25 range.
10-25? It’s more like 10-1000000000000000000000000000 (unless you’re in a room that contains only about 80 air molecules, in which case you’re in trouble anyway). I wonder if a number in a reputable publication has ever been wrong by this large a factor before.