This is indeed old news, and I have posted something similar before, but since repetition is the very purpose of the Internet, here it is again:
Some years ago I did participate in a "camera in the cockpit" experiment, which as a previous poster suggested was aimed at creating training material. Whilst it is possible to "act" a scripted scenario with grand thespian gestures at instruments to draw attention to what is going on, a much more telling video was recorded during a real high-workload incident which was not scripted at all. I and my copilot would later agree that our effort had been 100%, we were working at the peak of CRM performance, and a tricky situation was well managed. Watching the video however, all you see is two statues! Movements were so economical that you might think we never even twitched a muscle, let alone operated controls and switches (but we did). In a heightened state you can draw your colleagues attention to an instrument with the tiniest nod of the head or flick of the eye, and in the real world that is how it works. As a record of the incident, the video was therefore completely useless. For detailed operational metrics, the FDR is much more likely to explain what was going on.
Having said that, there have been some accident reports recently that would have benefited from cameras - but they are ones that have such major unknowns as "Who was sitting in which seat?". It is surprising how often an accident chain starts with someone leaving/entering the cockpit, and the reports end up with lines like "Unknown Voice Number 7"! These are questions that can best be answered with cameras; hoping for insight into detailed operation may be hoping for too much.