|
The GCP site has a lot of information about how the project is run.
The hardware is pretty much three different types of basically off the shelf cryptographically secure random number generators. They do produce random output. Furthermore, they XOR the data coming out of the RNGs with a balanced mask (I.E. equal parts ones and zeros) to normalize the data. They also have cryptolevel pseudoRNGs operating as controls.
If you flip a coin (or a bit) over and over and keep tally, after a given number of flips there is a probability that you'll have X number of tails (or 1s) more than heads (or 0s). Over a long enough timescale, you expect them to keep parity, for the odds to remain 50:50, for it to be a fair coin. Over short timescales, though, in a long enough trial, you should expect there to be runs of all heads or all tails (Popular assignment in introductory CS classes is to tell the students to record a such a heads and Tails flip for 100 trials. You should expect to see a run of 6 all heads or all tails in a truly random trail of this sort. But that doesn't "look random" to most people. So the teacher collects the results, and gives failing grades to everyone that doesn't have the run for just writing down a bunch of Hs and Ts rather than really flipping the coin.).
What the project is basically noting when the RNGs start flipping more heads or more tails than they should, and try to correlate those events with something happening in world events.
There is a spooky "pure data" angle to it. The RNGs are showing significant departures from normal random behaviour... presumably for some reason. Whether it's global conciousness or something else is up for debate, but there is a real effect that has to be explained.
__________________
Simple Machines in Higher Dimensions
|