Simulating environments

First things first, we will need a simulated environment. An environment is defined as the interaction space for a learning agent. For humans, an environment can be any play you go to in the course of a day. For an artificial agent, this will often be a simulated environment that we have engineered. Why is it simulated? Well, we could ask the agent to learn in real time, like ourselves, but it turns out that this is quite impractical. For one, we would have to design each agent a body and then precisely engineer its actions and the environments that they are to interact with. Moreover, an agent can train much faster in a simulation, without requiring it to be restricted to human time frames. By the time a machine completes a single task in reality, its simulated version could have completed the same task several times over, providing a better opportunity...