Asked by: Ken Rehill, Canada
Computers are often required to produce random numbers as they’re useful for a host of tasks, from taking random samples of data to simulating the formation of galaxies. But computers produce these numbers using mathematical formulas, which means they aren’t truly random. This isn’t as bad as it seems, as true randomness is prone to producing bizarre patterns that can fool researchers into seeing effects that don’t exist. To avoid this, the so-called pseudo-random number generator (PRNGs) formulas used in computers undergo statistical tests to keep the risk of long ‘runs’ below a certain threshold. Even so, some computer-generated random numbers have still caused problems. Subtle patterns in the output of the so-called RANDU generator created by IBM in the 1960s is suspected to have undermined the reliability of many research projects over the years.