Or maybe he does. Of course, most of us know by now that 6MP colour photos also really have only about 1.5 million-pixel's worth of unique information, since there are 6 million sensor elements, of which each captures just one colour component, and there are twice as many "green" as "blue" or "red".
The question is how much luminance information you can extract from a Bayer-pattern sensor. The simplest possible method I can think of would be to discard the red and blue data and interpolate the green pixels back into a square matrix - this would give you 3MP of information from a 6MP sensor. I don't have any way of testing this, but I'm happy to believe that more sophisticated algorithms can extract more luminance information by considering the other channels, and my personal estimate of the effective luminance resolution of a 6MP Bayer pattern sensor would be about 4MP. If Mishka meant "2MP-worth of information" when he said "2MP" then he was less incorrect, although I think it's an overly-pessimistic estimate. :-)
<snip>
I think, however, the main problem with B&W is that you decompose the light via filters, then put it back together, which is not quite the same thing as registering the amount of light (as opposed to colour) directly using a B&W film, or a plain CCD for that matter.
Agreed.
S