CMSC 471

Artificial Intelligence -- Fall 2002

HOMEWORK FOUR
out 10/16/01 due 10/30/01

http://www.cs.umbc.edu/671/fall01/hw/hw4.html

Part I. Philosophy of AI (35 points)

For questions 2 and 3, you should write approximately 200-500 words in response to each. (Please don't count words; just think about whether your answer makes sense, answers the question, and is understandable.) There are no right or wrong answers, only arguments and opinions that are presented well or poorly. Aim for the former! Please visit the Writing Center in the library if you think you need help in this area. If you lose points for excessive grammatical errors (or I return your essay ungraded due to lack of readability), I will give you one retry: that is, you may rewrite and resubmit your answer one time (unless you've already used that one-time resubmit this semester!)

John Searle's Chinese Room argument essentially boils down to this: a symbol processing system is insufficient to represent "true" intelligence. He uses a "reductio ad absurdum" argument to analogize an AI system to a room containing a non-Chinese-speaking person who is answering questions in Chinese by manipulating symbol tables with no understanding of the content of these tables. (Effectively, the person in the room is running the program, by acting as the CPU, with pieces of paper as the "memory" of the CPU.) Searle claims that it is absurd to believe that the person -- or the whole system -- "understands" Chinese.

1. Do you agree, disagree, or partially agree with Searle's assertion?

2. If you agree, how would you describe the nature of the difference between the symbol processing system and the brain's mental processes? That is, what is it about human mental processing that makes it inherently different from physical symbol processing in a computer? If you disagree, what do you think are the key similarities of symbol processing and human mental processing that you think make it (at least theoretically) possible to model the latter using the former?

3. In general, people are not willing to accept simulations of processes as equivalent to those processes. For example, I could write a simulation of a car's engine, but you wouldn't say that that simulation actually is a car's engine. Supporters of AI seem to be claiming that a simulation of intelligence is intelligent. Do you agree with this position? Why or why not? In particular, why is a simulation of intelligence like (or unlike) a simulation of a car's engine?

PART II. Logic Warm-Up (15 points)

(Russell & Norvig 7.6) Using the basic relations Child (x,y), Sibling (x,y), Female (x), Male(x), and Spouse (x,y):

(10 points) Write axioms describing the predicates GrandChild, Sister, Brother, Daughter, Son, Aunt, Uncle, and FirstCousin. (A first cousin is a parent's sibling's child.)
(5 points) A second cousin is a child of one's parent's first cousin; a third cousin is a grandchild of one's grandparent's first cousin; and so on. Write a definition of Nth-cousin (N, x, y), as a recursive expression in terms of the predicates defined above.

PART III. Resolution and Crime (50 points)

Dipsy has been murdered and Tinkey-Winkey, Laa-Laa, and Po are the only suspects.

Chief detective Barney is on the case, and brings them in for questioning. Each one tells the truth except for the culprit, who may be lying. Here is what they told Barney.

Tinkey-Winkey says that she is innocent. She says that Laa-Laa and Dipsy were not friends, and Po and Dipsy were not friends.
Laa-Laa says that he is innocent, plus that he and Dipsy were friends.
Po says that she is innocent, and that Laa-Laa and Dipsy were not friends.

As an astute detective, Barney makes the following assumptions about the world:

Friends don't murder each other.
There is exactly one killer, and it is one of the Teletubbies.
The killer may lie, but the other Teletubbies always tell the truth.

Your task is to do the following:

(b) Write a set of FOL sentences representing the general knowledge assumptions you've made. If necessary for (e) below, include here any other general knowledge sentences that you need.

(d) State the goal to be solved as an FOL sentence and solve the crime using a resolution refutation proof, showing your result as a proof tree that derives False.