Here's an attempt to explain the principles of the new experiment and compare it with the usual ones. In terms of two variants of "the Bell game".
Probably it is impossible to explain this to science-journalists. I think I explained once to an audience of Buddhists and neuro-scientists. But I had an hour and there was a very lively discussion. http://www.slideshare.net/gill1109/bell-43072906
Standard Bell game
==================
You control three computers A, B, C
You may write computer programmes for them.
Many times, the following happens:
Computer C sends messages to A and B
No more communication allowed between A, B and C
I toss two fair coins and deliver outcomes (H/T) to A and B
A and B output binary outcomes +/-1
Your aim: when A and B both receive "H" the outcomes should be the same (both +1 or both -1)
When either or both of A and B receive "T" the outcomes should be different (+1 and -1)
We call each of these exchanges a "trial". Each trial, you either win or lose
(you either achieve your per-trial aim or you don't).
The whole game: your overall aim is to "win" statistically significantly more than 75% of the trials (say: 80%, with N pretty large). Bell says it can't be done. (Well - just once in a while it could happen purely by chance, obviously, but you can't write those computer programs so that you systematically win).
Modifed Bell game
=================
You control three computers A, B, C
You may write computer programmes for them.
Many times, the following happens:
Computers A and B send messages to C
No more communication allowed between A, B and C
I toss two fair coins and deliver outcomes (H/T) to A and B
A and B output binary outcomes +/-1
C delivers a binary outcome "go"/"no-go"
Your aim: when C outputs "go" *and* A and B both receive "H" the outcomes should be the same (both +1 or both -1)
When C outputs "go" *and* either or both of A and B receive "T" the outcomes should be different (+1 and -1)
We call each of these exchanges a "trial". Each trial in which C says "go", you either win or lose
(you either acheive the aim or you don't).
The whole game: your aim is to "win" statistically significantly more than 75% of the trials for which C said "go"(say: 80%, with N pretty large). Bell says it can't be done. (Well - just once in a while it could happen purely by chance, obviously, but you can't write those computer programs so that you systematically win).
===============================
Relevance to QM. Einstein seems to think that the world essentially operates like a discrete automaton (e.g. game of life). ie like a network of classical computers. There can be randomness but it is local (the local rules may be probabilistic rules). QM says it doesnâ€™t. You can't model the universe with a classical discrete automaton.