# Introduction to Computers

## II. Expert Systems

### Consider possible classes of Sailboats

Figure 13-3 Sailing Craft Identified by Expert System These nine vessels with their sail plans are uniquely identified by clues the user provides the expert system in an interactive query system. [From Firebaugh, M., Artificial Intelligence, Boyd & Fraser (1988)]

### Using 17 IF-THEN Rules such as the following:

RULE NUMBER: 2

IF:

The number of masts is two
and The position of the Mainmast is forward of the
shorter mast (mizzenmast)
and The position of the mizzenmast, wrt the helm (rudder
post), is forward of the helm

THEN:

The type of vessel is a ketch

NOTE:

A ketch has a mainmast and a shorter mizzenmast which is just forward of the helm.

----------------------------------------

.... RULE NUMBER: 15

IF:

The type of vessel is a ketch
and The sail plan is gaff-headed

THEN:

NOTE:

A ketch with gaff-sail is a gaff-headed ketch.

----------------------------------------

## <<SAILING CRAFT IDENTIFIER>>

by: --> MORRIS W. FIREBAUGH

Press any key to start:

<CR> The SAILING CRAFT IDENTIFIER helps you identify various classes of sail boats from a brief description of their mast structure and sail plan. Please answer the following questions on the number and position of masts and shape of sails:

Press any key to start:
<CR>

The number of masts is

1 one
2 two

2

The position of the Mainmast is

1 forward of the shorter mast (mizzenmast)
2 aft the shorter mast (foremast)
3 about 40% aft the bow
4 about 25-30% aft the bow

1

The position of the mizzenmast, wrt the helm (rudder post), is

1 forward of the helm
2 aft the helm

1

The shape of the mainsail is

1 triangular
3 triangular with two foresails

2

From the information you have given me, I conclude that the sail boat you describe is a:

Values based on 0 - 10 system VALUE

### Examples of Expert Systems

• The Campbell Soup Company's system called Aldo

• Bowing Aircraft's electrical connector assembly system

• United Airlines gate/flight coordinator

• Medical Diagnostic systems (MYCIN)

## III. Robotics

### Consider the definition suggested by the Robotics Industries Association:

A robot is a re-programmable, multi-functional, manipulator designed to move material, parts, tools, or other specialized devices through various programmed motions for the performance of a variety of tasks.

Figure 16-2 T3 – An articulated robot with 6 Degrees of Freedom Note that 6 degrees of freedom are required for completely generalized motion. Three degrees are required for position and three for orientation. [From Firebaugh, M., Artificial Intelligence, Boyd & Fraser (1988)]

### What should robots do?

Machines do not breathe – This fact makes robots well suited for operation in a contaminated, poisonous, dangerous, or extreme environments where one (human) would not want to work or could work only with difficulty. Examples include the extreme conditions of space, nuclear reactors, and sea-bed exploration. The industrial application of spray-painting is an example in which robots have a competitive edge over humans.

Machines are strong – The mechanical forces available from hydraulic, pneumatic, and electrical motors give machines an enormous mechanical advantage over humans. This advantage is the basis of conveyor belts, cranes, and forklifts many of which presently require human operators but which could in many cases be converted to operate robotically. Large industrial spot welders weighing over 100 pounds are an illustration of such an application which is yielding rapidly to robotics.

Machines are precise, repetitive, and tireless – These features which provide the basis for automated assembly lines also provide the basis for successful robot applications. A prime example of a task that is tiresome, repetitive (boring), and requires high precision is that of “board stuffing” or inserting integrated circuits on printed circuit boards. Circuit boards assembled by robots have a much lower defect rejection ratio than those assembled by humans. This increased productivity has made board assembly a rapid growth area for robotics.

## IV. Critiques of the Turing Test

### Turing's own objections to AI

1. The Theological Objection --> "Thinking is a function of man's immortal soul." Turing refutes the above argument by noting: "It appears to me that the argument quoted above implies a serious restriction of the omnipotence of the Almighty. That is, could God have not given a soul to an elephant or a machine had he wanted to?"

2. The "Heads In The Sand" Objection --> "The consequences of machines thinking would be too dreadful. Let us hope and believe that they cannot do so." Turing notes that this objection "... is likely to be quite strong in intellectual people, since they value the power of thinking more highly than others ," but then dismisses it as insufficiently substantial to require refutation.

3. The Mathematical Objection --> (Gödel's Theorem) " in any sufficiently powerful logical system, statements can be formulated which can neither be proved nor disproved within the system, unless possibly the system itself is inconsistent." Turing acknowledges the validity of this argument, noting that he himself had published similar results. He then observes that there is no proof that the human intellect does not suffer the same limitations and concludes, "We too often give wrong answers to questions ourselves to be justified in being very pleased at such evidence of fallibility on the part of the machines."

4. The Argument From Consciousness --> "Not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not be a chance fall of symbols, could we agree that machine equals brain - that is, not only write it but know that it had written it." This argument leads to the classical solipsist position that the only way to know how a person thinks is to be that person. Rather than get into the endless circular arguments into which such a position leads, Turing notes that "... it is usual to have the polite convention that everyone thinks."

5. Arguments From Various Disabilities --> "I grant you that you can make a machine do all the things you have mentioned, but you will never be able to make one to do X" where X is the ability to "be kind, resourceful, beautiful, friendly, like ice cream." Turing refutes this argument by correctly observing that "Many of these limitations are associated with the very small storage capacity of most machines." Perhaps it is significant that increased memory capacity has facilitated the feature we call "user friendliness" in modern operating systems.

6. Lady Lovelace's Objection --> "The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform." Another version of this objection states: "Machines never do anything really new" or can never "take us by surprise." Turing says, "Machines take me by surprise with great frequency." Most of us who program certainly share this experience.

7. Arguments from Continuity of the Nervous System --> "The nervous system is certainly not a discrete state machine." Since small errors in the size of an impulse striking a neuron may make a large difference in the size of the outgoing impulse one cannot expect to mimic the behavior of the nervous system with a discrete state system. Turing fails to offer a convincing rebuttal to this argument, but we should note that present day computers can simulate the behavior of non-linear, analog devices for which the transformation is known, to almost any desired accuracy by standard computational procedures. This argument, however, is the central tenet of those working in the field of neural network research as the basis for the validity of their approach.

8. The Argument from Informality of Behavior --> "It is not possible to produce a set of rules purporting to describe what a man should do in every conceivable set of circumstances." For instance, if a red light means "stop" and a green light means "go," what would a computer do when faced with a simultaneous red and green light due to a fault in the system? This is a tough argument to refute, but Turing suggests that we can experimentally determine the "laws of behavior" (as opposed to the "rules of conduct") which may be able to resolve such dilemmas.

9. The Argument from Extrasensory Perception --> Here Turing cites four manifestations of extrasensory perception: telepathy, clairvoyance, precognition, and psychokinesis and sadly notes, "How we should like to discredit them! Unfortunately the statistical evidence, at least for telepathy, is overwhelming." Here at last Turing has blundered in interpreting telepathy as a valid phenomenon. The "overwhelming statistical evidence" has evaporated like the morning mist exposed to the harsh sunlight of full disclosure and independent confirmation. It is tempting to criticize Turing for not having the courage of his convictions, but we might note that many great scientists have had at least one faux pas. Maxwell refers to the "luminiferous ether" as the medium for his light waves. Einstein did not accept the statistical interpretation of quantum theory ("God does not play dice with the Universe!"), and Enrico Fermi observed both the neutron and nuclear fission but failed to discover them.

## V. Examples of AI Programs

### Example 1: Continuous Speech, User-Independent Voice Recognition

• Each of the Centris 660 AV machines in our lab is capable of responding to voice commands:
• "Computer, open the trash"
• "Computer, close folder"
• "Computer, what time is it?"

• Upon receiving the last query, the computer responds in perfect English
• "The time is now two fifty-six p.m."
• This problem, long a goal of AI research, has now been solved for standard phrases
• Involves recognizing a pattern of phonemes (vocal sounds)
• Real "understanding" of what it is learning is still missing

### Example 2: A Neural Network that Learns to Recognize Letters

• Each of these layers of neurons is totally connected to the next layer
• A set of 26 training letters is presented to the Input Layer
• Initially, random patterns appear on the middle and output layers
• The output layer activity is compared to the "Desired" Activity (binary 5)
• Error signals are generated and used to adjust the connections between layers
• This cycle is repeated until the circuit "learns" the letters

 Activites after 25 Cycles Activites after 50 Cycles Activites after 161 Cycles

• Note that after 161 cycles of learning the network has been trained
• The network uses only the "Rules for Learning"
• The network then adjusts its connection weights automatically
• Using the same algorithm, the network can learn any pattern

• This is an example of automatic (machine) learning

### Example 3: The Deep Blue Chess Program

• An IBM computer called Deep Blue beat world chess champion Garry Kasparov February 10, 1996, in the first game of their 6-game match
• This was the first machine victory over the human chess champion under classic tournament rules
• Deep Blue is a System 6000 Scalable POWER parallel RISC processor system
• It has 256 processors working in parallel
• It can compute 50-100 billion moves within the 3 minute game limit

• Kasparov went on to win the 1996 match 4-2.

• In 1997, however, the tide turned from humans to machines in chess. An account is given in:
• Mr. Kasparov said this about the match:

"For me, the match was over yesterday. I had no real strength left to fight. And today's win by Deep Blue was justified."

• IBM does not plan any further Deep Blue vs. Kasparov competition.