Lecture 8 Artificial Intelligence

Question 1

What is the name of the architecture used as the basis of most standard computers?

Answer 1

Von Neumann programmable computer architecture

Question 2

What is the Turing Test of Intelligence?

Answer 2

The Turing test for intelligence has an interrogator respond to 2 people via typed message. One of these people is a computer and the interrogator has to guess which one is which. The program is said to be intelligent if the interrogator can’t tell between the two.

Question 3

What did the Eliza program try to simulate? What was the surprisingly finding about this?

Answer 3

Rogerian psychoanalyst counselling. Some psychiatrists couldn’t tell between Eliza and a real counsellor and suggested that it be implemented in hospitals with large waiting lists.

Question 4

What are the two varieties of artificial intelligence?

Answer 4

Weak artificial intelligence- simulating intelligence (getting a machine to do things that would require intelligence done by a person)
Strong artificial intelligence - duplicating intelligence

Question 5

How did Searle (1990) suggest computers do when it comes to AI?

Answer 5

Computers don’t think. What they do is fundamentally different to thinking. They rely on production rules.

Question 6

How did searle explain his ideas on computers not thinking?

Answer 6

Searle used the Chinese box thought experiment to explain this concept. He said that a box in China containing an English speaking man received notes from the chinsese. The man in box had a rule book to decipher these notes and reply. The man does not understand the Chinese as a computer does not understand what it is inputted.

Question 7

How invented the machine “bombe” and what did it solve during world war 2?

Answer 7

Alan Turing invented the machine bombe which was designed to solve the unbreakable enigma code used by the Germans.

Question 8

What was the only valid rebuttal made about Searles Chinese box thought experiment?

Answer 8

That the box is reliant on today’s computer architecture. If the architecture was more advanced perhaps the Chinese man could understand.

Question 9

What were the main conclusions from Searles thought experiment ?

Answer 9

Can a machine think? Yes, us.
Can a man-made machine think? Yes, if duplicates us.
Can a digital computer think? Yes, we are digital computers.
Can a computer with the right program lead to understanding? No.

Question 10

What is an expert system?

Answer 10

An expert system is a computer program that emulates the decision-making ability of an expert

Question 11

What do expert systems rely on when making judgements ?

Answer 11

Production rules (IF… THEN…)

Question 12

What are expert systems an example of?

Answer 12

Weak AI. Expert systems can’t be described as no more than an automated textbook

Question 13

What is the physical symbol hypothesis? What is an example of this?

Answer 13

That any facet of human intelligence can be represented by symbolic structures. Take for example language which is a symbolic system that uses words to represent things.

Question 14

Describe the basic Von Neumann architecture.

Answer 14

Input - comes into the system ie by key strokes
CPU - hard processing
Memory
Output- ie what’s displayed on the screen

Question 15

Describe the connectionism/neural network system.

Answer 15

The system is biologically inspired and consisted of simple processing nodes which mirror neurones. Nodes are connect by weighted links and can strengthen or inhibit signals passed between.

Question 16

How can a neural network learn through back propagation?

Answer 16

Feed the network with the sort of data it is to process and then tell it what the expected output is. The network adjusts weights so that this output is given.

Question 17

What did McClelland and Rumelhart (1988) find when they employed the connectionist network ?

Answer 17

Used the connectionist network to produce past tense versions of English words. Found that errors that system made were similar to that of a child who was learning past tense.

Question 18

What the advantages of the connectionist system?

Answer 18

Overcome classification problems- ie. different font As.
Overcome context issues - processes things in parallel so context is taken into consideration.
Retrieval from partial descriptions- can recognise things from partial info.
Graceful degradation (doesn’t fail when there’s an error) - degrades gradually as errors are introduced.
Can produce vague generalisations.

Question 19

What are the criticisms of the connectionist system ?

Answer 19

Similarities to real neurons are superficial.
Rule based symbolic systems better capture some aspects of human language.
Neural nets generally implemented in serial, symbolic bog-standard computers anyways.
Many of the aspects of connectionist systems can be captured using more symbolic techniques

Question 20

What are genetic algorithms ?

Answer 20

Genetic algorithms are inspired by evolutionary theory. Involves selecting the strongest and cross breeding them through trial and error until the best solution is come up with. They are able to produce efficient and workable solutions but the solutions are often incomprehensible with no clear design principles.

Question 21

What are some examples of AI currently in use?

Answer 21

Siri. 
Credit card fraud. 
Face recognition. 
Computer games and special effects in movies. 
Self flying spacecrafts.

Question 22

What type of AI is advancing fast?

Answer 22

Weak AI

Question 23

Are we getting closer to strong AI?

Answer 23

Doesn’t look like it.