As an exemplar of such programs, I want you to look at McCarthy's advice-taking program in more detail. Please read [Programs with Common Sense]. You should find it fairly easy to read. This is partly because it occurs very early in AI history, but also because it remains uncluttered with implementation details, being a proposal for a program that was never built (something not entirely untypical of later AI work). Despite this, the paper is worth taking seriously: McCarthy is one of the founders of AI, and went on to do a lot of work on AI and logic, as well as inventing the Lisp programming language. As with the papers above, you won't be expected to reproduce the workings of this program for Finals. What I want you to note is the general approach implied, that of representing the world as logical propositions.
A problem still largely unsolved is how such deductive systems should
distinguish just those premises that it needs from all the irrelevant
ones. Some critics still assert that this is impossible in principle,
and hence so is this kind of AI: see p 86.
