Clouds strongly influence Earth's weather and climate, but they are difficult to represent accurately in computer simulations. High-altitude clouds are especially hard to model because the ice particles that make up these clouds come in a wide variety of shapes and properties. Over the years, millions of images of these ice particles have been taken using special cameras mounted on research airplanes. In our work, we train an AI model to uncover patterns in this large dataset of millions of images, without any human labels or explicit instructions on how to sort this data. Using the patterns our AI model discovers from these images, we hope to learn fundamental characteristics of the ice particle population in Earth's atmosphere. This will help us improve our representations of clouds in weather and climate models.