Trial by jury gets one set of 10 (12, whatever), whereas the consensus model described would have n sets of 10.
I seriously doubt the wisdom of a crowd goes up when the total crowd size shrinks to 10. I suspect it plummets.
This does call into question the notion of a multiple-jury trial (maybe videoconferenced in, where the individual juries do not interact with each other). Based on this study's findings, it would appear that juries might get their verdicts wrong less frequently with this system.
Imagine if the first OJ trial or Trayvon Martin trial had been before multiple juries; would society have been better off if the results were less dependent on the biases of juries small enough to be cherry-picked?