Still, it seems like if you wanted reduced power, and could accept reduced precision, wouldn't you just do the same thing in binary by using only a few bits? N

Still, it seems like if you wanted reduced power, and could accept reduced precision, wouldn't you just do the same thing in binary by using only a few bits? Not even 8 bits, but like 3 or 4 bits?