Thank you! That's me.
As I also wrote on DPreview:
The huge boost in sensitivity claimed by the black silicon team is only for short-wave IR. Out around 2um to be specific. In that region of the spectrum, silicon has very low sensitivity, so there is significant margin for improvement.
However, 2um is way outside the human visual range, so this does not help visible light camera sensors, where there is only small margin for improvement anyway. What's more, the surface treatment used to make black silicon would turn that material in to a very poor image sensor indeed.
The treatment creates a surface that is rough at the nano-scale and absorbs light (IR light). In modern CMOS image sensors, the photodiode is deliberately burried slightly down from the surface of the silicon in order to move it away from the defects that result in dark current. With the absorption in black silicon occurring in a nano-scale surface layer, it precludes the use of pinned (burried) photodiodes. This would limit it to the older, 3T sensor technology which can not use pinned PDs. Conventional 3T sensors have dark currents that are typically around 2 to 3 orders of magnitude higher than can be achieved using 4T technology, and the dark signal non-uniformity (DSNU), which causes pixel fixed pattern noise, is correspondingly high.
I strongly suspect that dark current and DSNU in a black silicon sensor would be much, much higher, as the treatment will be introducing huge amounts of defects in to the surface.
While black silicon may have uses in perhaps solar panels or as detectors for IR lasers, but as a image sensor technology it's a no-go!
