SCXT 350

 

Exam #2

 

 

 

 

 

Name ___________________________________

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Friday, April 2

100 pts.

 

 

The story so far

 

In exploring the idea that intelligent activity, language, cognition, perception and the like is computational in nature, we have started to look at approaches to building computational implementations of intelligent actions.  This gives rise to the field of artificial intelligence (AI).

 

We can categorize work in AI as symbolic AI (GOFAI) and connectionist AI.

 

Connectionist AI:

 

1.         (10 pts.)           Give a brief definition of the connectionist approach to AI.  Include some details (that is, a brief one-sentence response will probably not get full credit).

 

 

 

 

 

 

 

 

 

 

 

2.         (10 pts.)           Consider a  perceptron (Rosenblatt, 1958) with two inputs x1 and x2 (together with the bias input x0 set always to 1) and weights w0 = 1, w1 = -2, w2 = 2.  Sketch the perceptron (using the diagram we have used in class), and fill in the following table:

 

                                   

x1

x2

out

0

0

 

0

1

 

1

0

 

1

1

 

 

 

 

 

 

 

 

 

 

 

 

 

GOFAI

 

Symbolic AI, or good old-fashioned AI (GOFAI) is characterized by the commitment to the physical symbol system hypothesis (PSSH). 

 

3.         (10 pts.)           What is a physical symbol system?

 

GOFAI Continued

 

4.         (10 pts.)           State the physical symbol system hypothesis.  What is meant by ‘necessary’ and ‘sufficient’ in the statement?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(GOF)AI = KR (knowledge representation) + Search.  Newell and Simon also remark on the importance of heuristic search in their Turing award paper.

 

We begin by recalling some definitions:

 

5.         (5 pts. each)  Give brief definitions of each of the following terms related to using search in  problem-solving.  It might help to use one of our examples (for example, the wine-pouring puzzle) for illustrations, but please give general definitions.

 

            state

 

 

 

 

 

 

            operator

 

 

 

 

 

 

 

(continued next page)

(definitions continued)

 

            precondition

 

 

 

 

 

 

 

 

            postcondition

 

 

 

 

 

 

 

 

            heuristic

 

 

 

 

 

Search can be divided into uninformed search and informed (heuristic) search.  Uninformed search generally comes in two varieties:  depth-first and breadth-first search.

 

6.         (10 pts.)  You are in the center of a labyrinth with a (very long) ball of string (which will help you to back up to a branch in the labyrinth if you hit a dead end).  Give a brief description of depth-first search, and say how it would apply to your efforts to escape the labyrinth.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

7.         (5 pts.) One aid in escaping a labyrinth is to follow the left-hand wall.  What sort of aid is this?  What names do we give to rules like this?

 

 

 

 

 

Knowledge Representation

 

The final part of the equation ‘GOFAI = KR + Search’ is knowledge representation.  While we have just begun our discussion of KR, two questions might reasonably be asked:

 

8.         (10 pts.)           We have referred to many of the knowledge representation schemes as “slot and filler structures”.  Say briefly what we mean by this, and give two examples.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

9.         (10 pts.)           What is a production?  Give an example of a production that might be useful in taking an exam.