Let us consider the perception problem.  In perception, the world generates a percept, which is the input to the agent.  The goal in the perception task is to figure out which possible world generated the percept.  There are many different instantiations of this problem.
	However, this is a very hard problem.  In general, people have a hard time understanding how hard perception is for an automated agent, because it is so easy for people.  It’s obvious, looking at a chair, that it is a chair.  Why is it so hard?  To illustrate how hard this is for an agent, here is a perception task that is also hard for people.  What is this blob?  Even if I tell you that it’s a letter, it’s still hard to tell.  But what if I tell you that this letter is part of this word?