This paragraph should have read: I hereby conjecture that for any epsilon > 0 there is some 2^K and some 3^L such that either of them, expressed in either base 2 or base 3, has a fraction <= epsilon of digits that are nonzero. --Dan