Another example would be actions an autonomous entity could take in response to sensory data. You can represent the sensory data as a string of 1’s and 0’s for binary data (gets a little interesting if you want to use “fuzzy” GAs), as well as 1’s and 0’s for potential output actions. Using survival or some other fitness function, you can train a population that is able to “navigate” an environment or survive the longest.
Some previous work I did in GAs involved the use of the Quake II engine to create autonomous agents that were able to navigate the environment and search for food. The gene expression was a combination of distance to various environmental hazards as well as potential output actions like turn left if the last bit is a 1, turn right if its 0, etc.