Posted today by Richard Gill, of the Mathematical Institute:
"I point out a simple algebraic error in Joy Christian's refutation of Bell's theorem. In substituting the result of multiplying some derived bivectors with one another by consultation of their multiplication table, he confuses the generic vectors which he used to define the table, with other specific vectors having a special role in the paper, which had been introduced earlier. The result should be expressed in terms of the derived bivectors which indeed do follow this multiplication table. When correcting this calculation, the result is not the singlet correlation any more. Moreover, curiously, his normalized correlations are independent of the number of measurements and certainly do not require letting n converge to infinity. On the other hand his unnormalized or raw correlations are identically equal to -1, independently of the number of measurements too. Correctly computed, his standardized correlations are the bivectors - a . b - a x b, and they find their origin entirely in his normalization or standardization factors; the raw product moment correlations are all -1. I conclude that his research program has been set up around an elaborately hidden but trivial mistake. "
It is interesting to add this note, addressed to those who suggest Jaynes is the only person who properly understands how probability applies to Bell's Theorem, entanglement, etc: Gill is also an expert in statistical theory, and has done extensive research in this area (including the application of Bayes). He apparently does not see the issue Jaynes does. Gill frequently collaborates with the top scientists in the study of entanglement, so I think it is safe to say this area has been well considered and has not been overlooked somehow.