Real/Simulated World Dichotomy

While one of the goals of artificial intelligence is to design an intelligent agent that is fully functional in our world, the real world poses so many challenges that it is difficult to address them all in a single system. For example, agents that operate in the real world must be able to deal with imperfect sensing, and be able to react quickly to sensor data since the world is often dynamic and unpredictable. The agent must contend with the difficult issue of motor control which does not conform well to discrete operator application. Furthermore, the dynamic nature of the world and imperfect motor control means that an agent that operates in the real world must be able to constantly replan .

By designing an agent to operate in a simulated world, the researcher can focus more on cognitive functions such as planning and learning while carefully controlling the environment in a scientifically deliberate way. The hope is that research obtained by experimentation in simplified simulated environments will provide information about aspects of operation in the real world.


Back to List of Dichotomies