In his lectures on computer science, Feynman talks about finite state machines:
he present a simple delay finite state machine
Let me now give a specific example of an FSM that actually does something, albeit something pretty trivial - a delay machine. You feed it a stimulus and, after a pause, it responds with the same stimulus. That's all it does. Figure 3.4 shows the "state diagram" of such a delay machine.
I just don't see the utility of using two states here, why not just use a machine with one state?