On the potential of molecular computing