Kate Compton’s “10,000 Bowls of Oatmeal Problem” introduces the interesting idea that random generation, over a large set of possible outcomes, can eventually start producing outcomes that lack “perceptual uniqueness,” despite having different, randomized qualities. She described this through an example of 10,000 unique bowls of oatmeal. Although they would all be technically unique in their arrangement of oats, for example, over that large sample size many of the bowls would share many similarities and may only differentiate in one aspect.
In some situations, this phenomenon might actually come in handy, when doing tasks such as random map/environment generation, where one would want the different parts of the map to resemble the other without much noticeable change. In other scenarios, however, such as generating unique 2D designs, one would have to overcome this problem. Possible technical strategies could be to limit the number of possible similar values such that minute differences, although still technically different, wouldn’t fall under the same group visually. Another possibility would be to increase the amount of possible outcomes such that there would be more variation naturally.