Artificial Intelligence & Expert Systems

Document Sample
Artificial Intelligence & Expert Systems Powered By Docstoc
					Artificial Intelligence
   Expert Systems
   Definition of artificial Intelligence
Conventional data processing is concerned with inputting and processing data in the form
of fact and figures in order to produce operational or management information. Artificial
intelligence, on the other hand, is based on knowledge. A widely accepted definition of
artificial intelligence is based on a test devised by Alan Turing in 1950:

Suppose there are two identical terminals in a room, one connected to a computer, and the
other operated remotely by a person. If someone using the two terminals is unable to tell
which is connected to the computer and which is operated by thee person, the computer
can be credited with intelligence.

Competitions are regularly held in which judges “talk” via terminals to a mixture of
computer systems and people and try to guess which is which. In one such contest held in
Boston in August 1991, where questions were limited to a small range of specific topics, two
People and six programs were tested against six judges.

Four programs convinced at least one judge that they were people. Two judges mistook one
of the people, Cynthia Clay, for a program. They didn‟t believe anyone would know that
much about Shakespeare.

Q1: Write down three questions you could pose in order to distinguish between the
computer and the human.
                      Neural Computing
The technology that makes this possible is neural computing, the revolutionary process that
mimics the way the human brain works. By imitating the way that thought processes pass
through the neurons and synapses of the brain, reinforcing some signals and dissipating
others, neural computing enable machines to retain the most important information and
discard the rest. This self-teaching ability sets them apart from traditional computers which
cannot reprogram themselves.
Neural computing is being used in a wide variety of applications. Bass the brewers for
example, has built (with the aid of scientists at Warwick University) a machine that can sniff
the beer and decide if it is up to scratch. Its artificial nose makes a complex series of
judgements based on electro-chemical stimuli received by its gas sensors. By detecting an
over-active yeast or a weak crop of hops at an early stage in the fermentation process, the
“nose” could save the brewers from having to throw away a whole batch of 345,600 pints of
Bankers, stockbrokers and insurance agents are all experimenting with the possibilities of
the neural computing. TSB has been applying the system to forecasting the risk on
insurance premiums, and is now seeing how it can be used to detect fraud. The computer
can be trained to alert bank staff when an unusual transaction appears in a client‟s account.
Bankers also hope to train computers to tell them how much money load into their cash
Not all banks, however, have had happy experiences with neural computing. One large bank
reportedly lost millions of pounds before closing down its neural network in 1993. the
problem with a technology still in its infancy is its unpredictability; it is difficult to know
whether results will justify the expenditure.
                     Expert systems
Expert systems are computer programs that attempt to replicate the performance of a
human expert on some specialised reasoning task. Also called knowledge based systems,
they are able to store and manipulate knowledge so that they can help a user to solve a
problem or make a decision.

The main features of an expert system are:
It is limited to a specific domain (area of expertise)

It is typically rule based

It can reason with uncertain data (the user can respond “don‟t know” to a question)

It delivers advice

It explains its reasoning to the user.

An expert system has the following constituents:
„the knowledge base‟ that contains the facts and rules provided by a human expert

Some means of using the knowledge (an „inference mechanism‟ or „inference engine‟)

A means of communicating with the user (the „man-machine interface‟ or „human
computer interface‟)
                       The knowledge base
The knowledge base will store knowledge in different forms, namely FACTS and RULES. For example:
THE BASIC FEE FOR A 10 WEEK COURSE IS £25                                                            FACT
THE COURSE CODE FOR C PROGRAMMING IS EEC5012                                                         FACT
IF STUDENT IS UNEMPLOYED THEN THE BASIC FEE IS WAIVED                                                RULE
IF STUDENT IS UNDER 18 THEN ALL FEES ARE WAIVED                                                      RULE
CHARGE OF £15                                                                                        RULE
EEC5012 IS A 10 WEEK COURSE                                                                          FACT

Knowledge like this can be stored in a knowledge base, and the expert system should then be able to make
deductions. If we supply the information that Jo Bloggs wishes to enrol for C Programming, the expert system
should guide us through a series of relevant questions and deduce the fee to be charged.
In practice there may be hundreds or thousands of facts and rules. When the program runs it dose not simply
start at the first rule and run through them all; it makes deductions as it goes along, finding out what else it
needs to know before providing a answer. The way in which it does this is called the „inference mechanism‟.
Part of the inference mechanism may be represented in rules – for example.


Methods of reasoning fall into two categories: deduction and induction. Using deduction, we start with
statements which are true in general, and make specific deductions from them e.g. given that
              All mammals suckle their young
And           whales are mammals
We can deduce that
              Whales suckle their young.
Using induction, we use facts like “All the swans I have ever seen are white” to induce that “All swans are
white”. Of course sometimes, as in this case, these generalizations turn out not to be true.

Some types of knowledge are rather indefinite and difficult to store in an expert system. For example:
“This patient reminded me of one I saw a few months ago”
“Sometimes coffee seems to keep me awake at night”
“he is tall”
            The man-machine interface
This describes the way in which the user interacts with the system by asking questions, supplying answers,
requesting explanations, and so on. An inflexible interface will be like the traditional menu-drive programs
where the user has to know what to enter and has little opportunity for error or variation. A better interface
will be nearer to natural language and will let the user feel in control and well-informed throughout.

Software that allows you to build an expert systems for some particular purpose is sometimes called an
“Expert System Shell”, and the package Crystal is one example. This package allows you both to build the
expert system and the use it to seek advice.

In summery, a good expert system should be able to
          Allow a user to specify the parameters for a problem
          Ask relevant questions and draw inferences from the replies
          Be capable of handling incomplete or imprecise information
          Allow the user to change parameters to explore „what if‟ situations
          Make reasonable guesses or deductions
          Explain how it reaches its conclusions
Robotic and Embedded controls
There are many factories where robots are used on a production line. The jobs
   that the robots do are generally jobs which humans would find boring and
   repetitive, or jobs that may be dangerous.

These jobs would previously have been done by humans, but they may have
   been re-trained for other jobs - or possibly have been made redundant.

There will be some initial costs in purchasing the equipment and installing it,
   but the advantages of a robotic system would include..

   robots work 24 hours a day - no need for breaks.
   robots do not need to be paid.
   robots work consistently - they do not get tired or make mistakes.

Example :

Computer-controlled robots may be found in car-making factories. The jobs
   they do may include..
  welding or assembling parts
  paint spraying
  moving heavy parts around the factory.

The robots will have been programmed to perform the jobs they do.
                                   Stock Control
The stock level of an item is the number of that item in store.

When a shop sells items, it is sometimes important that the shop..

does not stock too many of an item e.g. The item might be perishable and those not sold will become

does not stock too few. A customer may want to buy one and there are none available.

For these reasons shops set reorder levels and reorder quantities. When the number of items is reduced to
     the reorder level, then more of that item are ordered. The number of items ordered is determined by the
     reorder quantity.

Stock Control is the administration of stock levels.

 A good stock control system will...
keep track of exactly how many of each item are in stock.
be able to say which items need re-ordering.
analyse which items are selling well or needed most and which are not.
 All businesses which store items (in shops, warehouses etc) will need a stock control system.

 Some shops use barcodes and POS terminals for automatic stock control. A computer stores a master file
     with records of every item held in stock. One field would be the item's stock level.
Every time an item is bought, the barcode is scanned and the computer will deduct 1 from the stock level for
     that item. The computer will know exactly how many of that item are in stock.

A similar system would operate in a warehouse. Each item removed would be logged and the stock level
     automatically adjusted.

This would be an example of a real-time system. Data is processed s soon as it is received and the system is
     always up-to-date.
                            Order Processing
Customers may buy goods from a business by sending an order - which will have details of which items they
    want to buy and how many.

Orders may be received...

...through the post on order forms. These will need to be entered onto the computer as records on the orders
      file. Customers send orders by email or over the Internet. These orders would be saved in the orders file.

The orders file would have fields with information about ...
the date of the order
customer details
item details and quantities
money owing or received from the customer.

Appropriate data validation will take place when the orders are entered.

When the goods are dispatched, an invoice is sent requesting the amount of money due for the order. The
    customer may then pay by cherub.

Online orders are usually paid for using credit (or debit) cards.

If goods are paid for online over the Internet, there would need to be a secure system in place to avoid
      problems with credit card details being stolen. When an order is processed....
stock levels are adjusted for the items sent.

financial information is stored and analysed. Who owes what and how much has been received etc
statistical information is updated - numbers of sales, best selling products etc...
                          Expert systems
   ...has a large database of knowledge.
   ...allows the database to be interrogated.
   ...has a set of rules (inference engine) for making deductions.
An expert system is a computer system which simulates the knowledge and expertise of a
    human expert. For example, in Medicine, expert systems are being used for disease
    diagnosis. The patient's details and symptoms are input, and the system outputs
    probable diagnoses, recommended treatments or drugs which may be prescribed.
Expert systems are not really replacing doctors but are being used to help them. There are
    ethical and legal reasons for this - if a computerised diagnosis is wrong, who do you
Some patients would feel happier typing medical information into a computer than
    discussing it with a human doctor...but others would prefer the 'human' touch.

The advantages of an expert system over a doctor are...
    ...a large database of knowledge can be added to and kept up-to-date - it can store
     more knowledge than a person.
    ...the system cannot 'forget' or get facts wrong.
 survives forever. There is no loss of knowledge as there is when a doctor retires.
    ...the computer can access specialist knowledge that a doctor may not have.
  An expert system would be programmed using an AI (Artificial Intelligence) language such
     as PROLOG.
                             Booking Systems
It is now possible to make bookings on-line for holidays, trains, planes, hotel rooms, theatre
       performances...and many others.
 A travel agent for example, may have computers in all its branches directly connected to a central computer
       where a database of all bookings is stored. This is an example of a multi-access system. When a
       booking is made, the customer will need to provide input details (name, date, place, number of people

These details may be entered ...
    at a computer terminal in a multi-access system.
    on a form on an Internet web page.
    by transcribing them from a paper booking form.
    by typing them in when in telephone communication with the customer.

This form (on an Internet web page) is used for collecting details of a passenger's booking on the Euro tunnel.
      These details are then used as input data when the booking is made.

This data is validated to check if the details are sensible. The computer will check to see if the booking is
      available, and, if it is, the booking is made and it will then store the booking details in the database.

Documents will need to be output to give to the customer, confirming the booking and giving details about it.

As soon as a customer makes a booking it has to be processed immediately, so that no other customer can
     make the same booking. This means it is a real-time (transaction processing) system.

It is essential that no data is lost, so the database will have to be regularly backed up - possibly using a tape

If payments for the booking are required, then these can generally be done on-line using a Credit card or a
      Debit card. If details of these are transmitted over the Internet, the website must be secure so that this
      information cannot be stolen.

Shared By:
pptfiles pptfiles