Ecosystems for Algorithm Development

 

The autonomous development of habitat and infrastructure on Mars will require a control system architecture that can handle intricate systems. Developing complex systems is a challenging problem. It is often unrealistic to expect the availability of a training and testing environment that can help us evaluate the performance. Many testing environments are complex but not terribly useful. Can we expect that a great chess-playing AI will be helpful in other tasks? Unlikely. However, some games are much better. For example, StarCraft is a complex ecosystem that contains many relationships that offer a way to test hypotheses about ecological and evolutionary processes. Thus, there is a feeling that some strategy games will be a good sandbox for complex cybernetic system design and development.

https://www.discovermagazine.com/technology/how-ai-accidentally-learned-ecology-by-playing-starcraft