Knowledge representation in AI has been a bottleneck for years. And the difficulty is uncertainty hidden in qualitative concepts, that is the randomness and fuzziness. At this junction, this paper presents a new concept of cloud models with three digital characteristics： expected value Ex, entropy En, and hyper entropy He. This methodology has effectively made mapping between quantitative and qualitative knowledge much easier at any time. A cloud drop, that is a quantitative value, representing the qualitative concept can be measured by contributions. A new explanation for the 24 solar terms in lunar calendar is given as well. The cloud models have been used in data mining, intelligent control, hopping frequency technique, system evaluation, and so on.