Running head: COMPUTERS THAT "THINK"

Computers That "Think:"

Is it Live or is it Memorex?

Tabatha W. Davis

University of Maryland, European Division

Bowie State University

Abstract

Examination of neural networks, how they work, and the possibility of creating computers that "think." The results indicated the optimism and pessimism of building computers that "think." However, the pessimists out-weighed the optimists. Because of the uncertainty of the brain’s functions, the findings are consistent with and in support of the position that computers do not "think."

Table of Contents

1. Introduction to Neural Networks

1. What is a neural network?
2. Historical background of neural networks
3. Why use neural networks
4. Neural networks versus conventional computers

2. Human and Artificial Neurons

1. How the human brain learns?
2. From human neurons to artificial neurons
3. How neural networks learn?

3. The Learning Process

1. The teaching procedure
2. The back propagation algorithm

4. Applications of Neural Networks

1. Neural networks in practice
2. Neural networks in business

5. The Future of Neural Networks
1. Conclusion
2. References

Thesis Statement

An increasing number of computer scientist have indicated that through Artificial Intelligence, which is a branch of science that attempts to mimic the cognitive and symbolic skills of humans (Caudill, Butler, 1990; Forsyth, 1986), and the use of neural networks, described as a new approach to computing that roughly mimics the brain’s unmatched ability to recognize and understand patterns-faces, voices, and written characters (Bylinsky, 1993; see also Chitra, Bulson, & Morrell, 1995; O’Reilly, 1989), computers will be able to do everything that humans can do including the ability to "think." Many of these scientists are influenced by revolutionary research taking place around the world aimed at creating devices that are more like living brains instead of computers (Freedman, 1994). Optimistic researchers contend that machines are physical things that can perform certain tasks; people are biological machines; so, since people can think, machines can think ("Artificial Intelligence or Maybe Not," 1990).

According to Hawkins (1988), "think" can be defined as to exercise the mind in an active way, to form connected ideas, to have as an opinion, to judge, or to take into consideration. "Think" is also regarded as a chain reaction in which activity in one network stimulates associated responses in another (Editors of Time-Life Books, 1991). The latter is considered to be a biological definition of "think." In order to fully, or even partially, understand the biological meaning of the word "think," one must examine the human brain and the tools that make it work.

The most important building block of the brain is the nerve cell, or neuron, which transmits information in the form of electrical impulses (Lemonick, 1995). Some scientist believe that by imitating the structure of the brain cells and the three-dimensional lattice of conditions among them (Bylinsky, 1993), they can reveal the steps of the learning process which is otherwise called, for this study, the "thinking process." A close examination of this process will follow in future sections of this report.

In this theoretical study, the author has attempted to examine neural networks, how neural networks work, and the possibility of creating computers that "think" through neural networks. It is felt that neural networks, data processing systems modeled on some properties of the brain and nervous system (Cooper, 1989), are in their infancy and this will be an art for a long time before it becomes a science (O’Reilly, 1989). The author also believes that although computers add, subtract, and execute instructions with incredible speed and accuracy, they fail to be good models for the brain (Cooper, 1989). The author will attempt to prove that the process of thinking is to use one’s power of conception, judgement, or inference, or to form or have in the mind (Webster’s New Collegiate Dictionary, 1973). Since a computer is an automatic electronic machine for performing calculations (Webster’s, 1973) and does not have the power of conception, judgement, or inference and does not have a mind, it can not possibly "think."

Outline

Computers That "Think:" Is it Live or is it Memorex?

I. Introduction to Neural Networks

A. Definition of neural networks

2. Historical background of neural networks

1. First Attempts

2. Promising and Emerging Technology

3. Period of Frustration and Disrepute

4. Innovation

5. Re-Emergence

3. Uses of neural networks
1. Neural networks vs. Conventional computers

2. Human and Artificial neurons

A. How the human brain learns

1. Components of the neuron
2. The "Thinking Process"

2. From human neurons to artificial neurons

3. The Learning Process

1. The teaching process
2. The back propagation algorithm

4. Applications of neural networks

1. Neural networks in practice
2. Neural networks in business

5. The future of neural networks
1. Conclusion
2. References

A neural network is an information processing paradigm that is inspired by the way biological nervous systems, such as the brain, process information (Stergiou, Siganos, 1996). Another definition of neural networks is mathematical systems that are comprised of a number of "processing units" that are linked via weighted interconnections (BioComp Systems, 1995-1998). Still, another definition of neural networks by the inventor of the first neurocomputer, Dr. Robert Hecht-Nielson is as follows: "a computing system made up of a number of simple, highly interconnected processing elements, which process information by their dynamic state response to external inputs." (Caudill, 1989). A neural network is a processing device that is loosely modeled after the neuronal structure of the mammalian cerebral cortex but on a much smaller scale. A large neural network might have hundreds or thousands of processor units, whereas a mammalian brain has billions of neurons (A Basic Introduction to Neural Networks).

2. Historical background of neural networks

Neural network simulations appear to be a recent development. However, this field was established before the advent of computers, and has survived at least one major setback and several eras. Following an initial period of enthusiasm, the field survived a period of frustration and disrepute. During this period when funding and professional support was minimal, important advances were made by relatively few researches. These pioneers were able to develop convincing technology which surpassed the limitations identified by earlier researchers. Currently, the neural network field enjoys a resurgence of interest and a corresponding increase in funding. The history of neural networks can be divided into several periods:

First Attempts: There were some initial simulations using formal logic. McCulloch and Pitts (1943) developed models of neural networks based on their understanding of neurology. These models made several assumptions about how neurons worked. Their networks were based on simple neurons which were considered to be binary devices with fixed thresholds. The results of their model were simple logic functions such as "a or b" and "a and b". Another attempt was by using computer simulations. Two groups (Farley and Clark, 1954; Rochester, Holland, Haibit and Duda, 1956). The first group (IBM researchers) maintained closed contact with neuroscientists at McGill University. So whenever their models did not work, they consulted the neuroscientists. This interaction established a multidisciplinary trend which continues to the present day.

Promising & Emerging Technology: Not only was neuroscience influential in the development of neural networks, but psychologists and engineers also contributed to the progress of neural network simulations. Rosenblatt (1958) stirred considerable interest and activity in the field when he designed and developed the Perceptron. The Perceptron had three layers with the middle layer known as the association layer. This system could learn to connect or associate a given input to a random output unit.

Another system was the ADALINE (ADAptive LInear Element) which was developed in 1960 by Widrow and Hoff (of Stanford University). The ADALINE was an analogue electronic device made from simple components. The method used for learning was different to that of the Perceptron, it employed the Least-Mean-Squares (LMS) learning rule.

Period of Frustration & Disrepute: In 1969 Minsky and Papert wrote a book in which they generalized the limitations of single layer Perceptrons to multilayered systems. In the book they said: "...our intuitive judgment that the extension (to multilayer systems) is sterile". The significant result of their book was to eliminate funding for research with neural network simulations. The conclusions supported the disenchantment of researchers in the field. As a result, considerable prejudice against this field was activated.

Innovation: Although public interest and available funding were minimal, several researchers continued working to develop neuromorphically based computational methods for problems such as pattern recognition. During this period several paradigms were generated which modern work continues to enhance. Grossberg's (Steve Grossberg and Gail Carpenter in 1988) influence founded a school of thought which explores resonating algorithms. They developed the ART (Adaptive Resonance Theory) networks based on biologically plausible models. Anderson and Kohonen developed associative techniques independent of each other. Klopf (A. Henry Klopf) in 1972, developed a basis for learning in artificial neurons based on a biological principle for neuronal learning called heterostasis.

Werbos (Paul Werbos 1974) developed and used the back-propagation learning method, however several years passed before this approach was popularized. Back-propagation nets are probably the most well known and widely applied of the neural networks today. In essence, the back-propagation net is a Perceptron with multiple layers, a different threshold functions in the artificial neuron, and a more robust and capable learning rule.

Amari (A. Shun-Ichi 1967) was involved with theoretical developments: he published a paper which established a mathematical theory for a learning basis (error-correction method) dealing with adaptive pattern classification. While Fukushima (F. Kunihiko) developed a step wise trained multilayered neural network for interpretation of handwritten characters. The original network was published in 1975 and was called the Cognitron.

Re-Emergence: Progress during the late 1970s and early 1980s was important to the re-emergence on interest in the neural network field. Several factors influenced this movement. For example, comprehensive books and conferences provided a forum for people in diverse fields with specialized technical languages, and the response to conferences and publications was quite positive. The news media picked up on the increased activity and tutorials helped disseminate the technology. Academic programs appeared and courses were introduced at most major Universities (in US and Europe). Attention is now focused on funding levels throughout Europe, Japan and the US and as this funding becomes available, several new commercial with applications in industry and financial institutions are emerging.

Today: Significant progress has been made in the field of neural networks-enough to attract a great deal of attention and fund further research. Advancement beyond current commercial applications appears to be possible, and research is advancing the field on many fronts. Neurally based chips are emerging and applications to complex problems are developing. Clearly, today is a period of transition for neural network technology. (Stergiou, 1996)

3. Why use neural networks

Neural networks, with their remarkable ability to derive meaning from complicated or imprecise data, can be used to extract patterns and detect trends that are too complex to be noticed by either humans or other computer techniques (Stergiou, Siganos, 1996). Neural networks are universal approximators, and they work best if the system you are using them to model has a high tolerance to error. Neural networks work well for capturing associations or discovering irregularities within a set of patterns; where the volume, number of variables or diversity of the data is very great; the relationships between variables are vaguely understood; or, the relationships are difficult to describe adequately with conventional approaches (A Basic Introduction to Neural Networks). Neural networks are good pattern recognition engines and robust classifiers, with the ability to generalize in making decisions about imprecise input data. They offer ideal solutions to a variety of classifications problems such as speech, character and signal recognition, as well as functional prediction and system modeling where the physical processes are not understood or are highly complex (Battelle Memorial Institute, 1997). According to Battelle Memorial Institute, 1997, the advantage of neural networks lies in their resilience against distortions in the input data and their capability of learning.

4. Neural networks versus conventional computers

Neural networks take a different approach to problem solving than that of conventional computers. Conventional computers use an algorithmic approach i.e. the computer follows a set of instructions in order to solve a problem. Unless the specific steps that the computer needs to follow are known the computer cannot solve the problem. That restricts the problem solving capability of conventional computers to problems that we already understand and know how to solve.

Neural networks process information in a similar way the human brain does. The network is composed of a large number of highly interconnected processing elements (neurons) working in parallel to solve a specific problem. Neural networks learn by example. They cannot be programmed to perform a specific task. The examples must be selected carefully otherwise useful time is wasted or even worse the network might be functioning incorrectly. The disadvantage is that because the network finds out how to solve the problem by itself, its operations can be unpredictable (Stergiou, Siganos, 1996). Neural networks provide an analytical alternative to conventional techniques which are often limited by strict assumptions of normality, linearity, variable independence, etc. Because a neural network can capture many kinds of relationships it allows the user to quickly and relatively easily model phenomena which otherwise may have been very difficult or impossible to explain otherwise (A Basic Introduction to Neural Networks).

Much is still unknown about how the brain trains itself to process information, so theories abound. In the human brain, a neuron collects signals from others through a host of fine structures called dendrites. The neuron sends out spikes of electrical activity through a long, thin stand known as an axon, which splits into thousands of branches. At the end of each branch, a structure called a synapse converts the activity from the axon into electrical effects that inhibit or excite activity from the axon into electrical effects that inhibit or excite activity in the connected neurons. When a neuron receives excitatory input that is sufficiently large compared with its inhibitory input, it sends a spike of electrical activity down its axon. Learning occurs by changing the effectiveness of the synapses so that the influence of one neuron on another changes (Stergiou, 1996). This process is shown below.

Components of a neuron

The synapse

According to Caudill and Butler (1990), an artificial neural network has only three building blocks: neurodes (an artificial model of the biological neuron), interconnects (the paths or links between neurodes), and synapses (the junction where an interconnect meets a neurode). These correspond to the neurons, axons, and synapses of biological systems.

The basic procedure of computing is a three-step process: input-activation-output. First, the neurodes must compute its total incoming stimulation from all sources. Next, it must decide the level of activity that total incoming stimulation corresponds to. Finally, it must generate the appropriate output signal for the resulting level of activity and transmit it along its output interconnects (Caudill, Butler, 1990).

Contrary to the above-mentioned process, Chitra, Bulson & Morrell (1995), believe there are four steps in the "thinking process:" 1) various signals are received from other neurons, 2) a weighted sum is calculated, 3) the calculated sum is transferred by a function, and 4) the transformed result is sent to other neurons. This process, as described above, is illustrated

below.

Four-steps in "thinking process"

Input signals to a neuron Output signals to other neurons

It is unclear as to which process is correct. In fact, because of the uncertainty surrounding the brain and how it functions, it might be safe to say that neither is the "thinking process."

Neural networks are conducted by trying to deduce the essential features of biological neurons and their interconnections. The computer is then typically programmed to simulate these features. However, because the knowledge of biological neurons is incomplete and computing power is limited, models are necessarily gross idealizations or real networks of neurons (Stergiou, 1996).

Artificial neural networks are typically composed of interconnected "units", which serve as model neurons. The function of the synapse is modeled by a modifiable weight, which is associated with each connection. Each unit converts the pattern of incoming activities that it receives into a single outgoing activity that it broadcasts to other units. It performs this conversion in two stages:

1. It multiples each incoming activity by the weight on the connection and adds together all these weighted inputs to get a quantity called the total input.
2. A unit uses an input-output function that transforms the total input into the outgoing activity.

The behavior of an artificial neural network depends on both the weights and the input output function (transfer function) that is specified for the units. This function typically falls into one of three categories:

• linear
• threshold
• sigmoid

For linear units, the output activity is proportional to the total weighted output.

For threshold units, the output is set at one of two levels, depending on whether the total input is greater than or less than some threshold value.

• The activity of the input units represents the raw information that is fed into the network.
• The activity of each hidden unit is determined by the activities of the input units and the weights on the connections between the input and the hidden units.
• The behavior of the output units depends on the activity of the hidden units and the weights between the hidden and output units.

The simplest type of network is interesting because the hidden units are free to construct their own representations of the input. The weights between the input and hidden units determine when each hidden unit is active, and so by modifying these weights, a hidden unit can choose what is represents (Stergiou, 1996).

A three-layer network can be taught to perform a particular task by using the following procedure:

1. Present the network with training examples, which consist of a pattern of activity for the input units together with the desired pattern of activities for the output units.
2. Determine how closely the actual output of the network matches the desired output.
3. Change the weight of each connection so that the network produces a better approximation of the desired output.

2. The Back Propagation Algorithm

To train a neural network to perform some task, one must adjust the weights of each unit in such a way that the error between the desired output and the actual output is reduced. This process requires that the neural network compute the error derivative of the weights (EW). In other words, it must calculate how the error changes as each weight is increased or decreased slightly. The back propagation algorithm is the most widely used method for determining the EW (Stergiou, 1996).

The back propagation algorithm is easiest to understand if all the units in the network are linear. The algorithm computes each EW by first computing the EA, the rate at which the error changes as the activity level of a unit is changed. For output units, the EA is simply the difference between the actual and the desired output. To compute the EA for a hidden unit in the layer just before the output layer, one first identifies all the weights between that hidden unit and the output units to which it is connected. One then multiplies those weights by the EA of those output units and add the products. This sum equals the EA for the chosen hidden unit. After calculating all the EAs in the hidden layer just before the output layer, one can compute in like fashion the EAs for other layers, moving from layer to layer in a direction opposite to the way activities propagate through the network. This is what gives back propagation its name. Once the EA has been computed for a unit, it is straight forward to compute the EW for each incoming connection of the unit. The EW is the product of the EA and the activity through the incoming connection. For non-linear units, the back propagation algorithm includes an extra step. Before back propagating, the EA must be converted into the EI, the rate at which the error changes as the total input received by a unit is changed (Stergiou, Siganos, 1996).

According to Stergiou and Siganos, 1996, Neural networks have broad applicability to real world business problems. In fact, they have already been successfully applied in many industries. Since neural networks are best at identifying patterns or trends in data, they are well suited for prediction for forecasting needs including: sales forecasting, industrial process control, customer research, data validation, risk management, and target marketing. Artificial neural networks are also used in the following specific paradigms: recognition of speakers in communications, diagnosis or hepatitis, recovery of telecommunications from faulty software, interpretation of multimeaning Chinese words, undersea mine detection, texture analysis, three-dimensional object recognition, hand-written word recognition, and facial recognition.

There are a number of new applications using neural networks. These applications are as follows: Pen PC’s - PC’s where one can write on a tablet, and the writing will be recognized and translated into (ASCII) text. Speech and Vision recognition systems - Not new, but neural networks are becoming increasingly part of such systems. They are used as a system component, in conjunction with traditional computers. White goods and toys - as neural network chips become available, the possibility of simple cheap systems which have learned to recognize simple entities (e.g. walls looming, or simple commands like Go, or Stop), may lead to their incorporation in toys and washing machines, etc. Already the Japanese are using a related technology, fuzzy logic, in this way. There is considerable interest in the combination of fuzzy and neural technologies (Smith, 1998).

2. Neural Networks in Business

Business is a diverted field with several general areas of specialization such as accounting or financial analysis. Almost any neural network application would fit into one business area or financial analysis. There is some potential for using neural networks for business purposes, including resource allocation and scheduling. There is also a strong potential for using neural networks for database mining, that is, searching for patterns implicit within the explicitly stored information in databases. Most of the funded work in this area is classified as proprietary. Thus, it is not possible to report on the full extent of the work going on. Most work is applying neural networks (Stergiou, Siganos, 1996).

Because gazing into the future is somewhat like gazing into a crystal ball, it is better to quote some "predictions." Each prediction rests on some sort of evidence or established trend which, with extrapolation, clearly takes us into a new realm (Stergiou, 1996).

Prediction 1:

Neural Networks will fascinate user-specific systems for education, information

processing, and entertainment. "Alternative realities", produced by comprehensive

environments, are attractive in terms of their potential for systems control,

education, and entertainment. This is not just a far-out research trend, but

is something which is becoming an increasing part of our daily existence, as

witnessed by the growing interest in comprehensive "entertainment centers" in

each home.

This "programming" would require feedback from the user in order to be effective

but simple and "passive" sensors (e.g. fingertip sensors, gloves, or wristbands to

sense pulse, blood pressure, skin ionization, and so on), could provide effective

feedback into a neural control system. This could be achieved, for example, with

sensors that would detect pulse, blood pressure, skin ionization, and other

variables which the system could learn to correlate with a person's response state.

Prediction 2:

Neural networks, integrated with other artificial intelligence technologies,

methods for direct culture of nervous tissue, and other exotic technologies such

as genetic engineering, will allow us to develop radical and exotic life-forms

whether man, machine, or hybrid.

Prediction 3:

Neural networks will allow us to explore new realms of human capability realms

previously available only with extensive training and personal discipline. So

a specific state of consciously induced neurophysiologically observable awareness

is necessary in order to facilitate a man machine system interface.

Conclusion There are many optimistic scientist who believe that through neural networks, computers can "think." In fact, according to Bylinsky (1993), neural networks have finally moved from a laboratory curiosity into everyday usage in a wide variety of industries. In addition, the network is the ultimate decision-maker; it does a better job than most analysts can and does it more consistently. Stergiou & Siganos, 1996, believes that the computing world has a lot to gain from neural networks. Their ability to learn by example makes them very flexible and powerful. Furthermore there is no need to devise an algorithm in order to perform a specific task; i.e. there is no need to understand the internal mechanisms of that task. They are also very well suited for real time systems because of their fast response and computational times which are due to their parallel architecture. Neural networks also contribute to other areas of research such as neurology and psychology. They are regularly used to model parts of living organisms and to investigate the internal mechanisms of the brain. Perhaps the most exciting aspect of neural networks is the possibility that some day ‘conscious’ networks might be produced. There are a number of scientists arguing that consciousness is a ‘mechanical’ property and that ‘conscious’ neural networks area realistic possibility.

In general, hopeful scientists believe that by studying neurons, creating neural networks which mimic the way the human brain copes with incomplete and confusing information (Chitra, Bulson & Morrell, 1995), and implementing these networks into computer hardware and/or software, they hold the key to understanding the brain and how it works. This will, in turn, make it possible to design computers that "think."

In contrast, pessimistic researchers view neural networks in a different light. Specialists in brain research scoff at the suggestion that artificial neural networks have anything at all to do with a brain, other than the contrivance of a common terminology (Brody, 1990). Contrary to popular conception, neural networks do not mimic the operation of the human brain. The best that can be said is that a simplistic analogy is possible: like the brain, neural nets can be thought of as consisting of interconnected neurons linked together by synapses. When enough of the input synapses send a signal into a neuron, it ‘fires’, causing signals to be sent down its output synapses, which in turn cause other neurons to fire, and so on. But this analogy is not worth pursuing too far. The brain’s mechanisms are vastly more subtle and complex than those in an artificial neural network, and it has orders of magnitude more connections. Most neural nets are software simulations rather than a hardware system. They are often nothing more than a sub-routine written in C which typically will have been automatically generated by a neural net development tool (Neural Networks). Jim Bower, a Caltech neurobiologist, states that the notion that neural networks mimic the brain is "nonsense" (Brody, 1990). Another scientists (e.g. Wilkes, 1992) believes intelligent behavior is outside the range of the computer. It is also believed that neural networks do not mimic the brain because an understanding of the brain’s functions is unknown. Scientists have not been able to figure out how the brain works. Michael Lemonick (1995) voices the brain is a hot topic and a complete understanding of the inner workings will be a long time coming. Before trying to duplicate the human brain, scientists will have to learn far more about the brain than they already know (Gorman, 1988). In essence, the opinion of the pessimistic researches can be summarized in the following statement ("Artificial Intelligence or Maybe Not", 1990):

No-one expects to get wet in a pool filled with

Ping-Pong-ball modes of water molecules. So

why would anything think a computer model of

thought processes would actually think? (p.89)

As hypothesized, computers are electronic machines for manipulating data. They can not possibly posses the tools needed to "think" because it is unclear what makes the brain "think." Therefore, computers do not "think."

The results of this research can generally be found in the following statement by the Editors of Time-Life Books (1991):

Computers will probably remain what machines

have always been - aids for specialized jobs.

For at least another century, the generalists-

the dreamers, and schemers, the problem posers

and problem solvers - will still be people,

relying on their peerless biological brains. (p.110)

References

A basic introduction to neural networks. [On-Line]. Available: http://blizzard.gis.uiuc.edu/htmldocs/Neural/neural.html

Artificial intelligence or maybe not. (1990, January 27). The Economist, 314, 89.

Battelle Memorial Institute (1997). What is an artificial neural network? [On-Line]. Available: http://www.emsl.pnl.gov:2080/proj/neuron/neural/what.html

BioComp Systems, Inc. (1995-1998). What are neural networks? [On-Line]. Available: http://www.bio-comp.com/pages/NNDefine.html

Black, I.B. (1991). Information in the brain: A molecular perspective. Cambridge, Massachusetts, London: The MIT Press.

Boden, M. (1877). Artificial intelligence and natural man. New York: Basic Books, Inc., Publishers.

Brody, H. (1990, August). The neural computer. Technology Review, 93, 42-48.

Bylinsky, G. (1993, September 6). Computers that learn by doing. Fortune, 128, 96-102.

Caudill, M., Butler, C. (1990). Naturally intelligent systems. Cambridge, Massachusetts, London: The MIT Press.

Chitra, S., Bulson, R.J., Morrell, D. (1995, February). Computer programs that learn from experience. Chemtech, 25, 18-25.

Clarkson, M. (1995, February). Eyes, ears, & brain on a chip. Byte, 20, 91-96.

Cooper, L. (1989, March 6). First word. Omni, 11, 6.

Editors of Time-Life Books. (1991). Understanding computers: Artificial Intelligence. Alexandria, Virginia: Author.

Forsyth, R. (1986). Science in action: Machines that think. New York, London: Warwick Press.

Freedman, D.H. (1994). Brainmakers: How scientists are moving beyond computers to create a rival tot he human brain. New York, London: Simon & Schuster.

Goldberg, L. (1994, October 3). Self-learning neural chips promise neural-network applications. Electronic Design, 42, 44-46.

Gorman, C. (1988, August 8). Putting brainpower in a box. Time, 132, 59.

Gross, N. (1992, June 8). A Japanese ‘flop’ that became a launching pad. Business Week, 0, 103.

Hawkins, J.M. (Ed.). (1988). The Oxford Large Print Dictionary. Oxford, New York, Toronto: Oxford University Press.

Kurzweil, R. (1992, November 15). The paradigms and paradoxes of intelligence: Building a brain. Library Journal, 117, 53-54.

Lemonick, M.D. (1995, July 17). Glimpse of the Mind: What is consciousness? Memory? Emotion? Science unravels the best-kept secrets of the human brain. Time, 146, 44-52.

Neural networks. [On-Line]. Available: http: //aiintelligence.com/aii-info/techs/nn.htm#What

O’Reilly, B. (1989, February 27). Computers that think like people. Fortune, 119, 90-93.

Penrose, R. (1994). Shadows of the mind: A search for the missing science of consciousness. Oxford, New York, Melbourne: Oxford University Press.

Penrose, R. (1989). The emperor’s new mind concerning computers, minds, and the laws of physics. New York, Oxford: Oxford University Press.

Peterson, I. (1993, February 27). Neural networks for learning verbs. Science News, 143, 141.

Port, O., Shiller, A., Miles, G.L., Schulman, A., bureau reports. (1989, May 8). Smart factories: America’s turn? A few advanced plants now-A competitive arsenal soon. Business Week, 0, 142-148.

Smith, Leslie Dr. (1998). An introduction to neural networks. [On-Line]. Available: http://www.cs.stir.ac.uk/~lss/NNIntro/InvSlides.html

Stergiou, Christos (1996). Neural networks, the human brain, and learning. [On-Line]. Available: http://www-dse.doc.ic.ac.uk/~nd/su…96/journal/vol2/cs11/article2.html

Stergiou, Christos (1996). What is a neural network? [On-Line]. Available: http://www-dse.doc.ic.ac.uk/~nd/su…96/vol1/cs11/article1.html

Stergiou, C., Siganos, D. (1996). Neural networks. [On-Line]. Available: http://www-dse.doc.ic.ac.uk/~nd/su…e96/journal/vol14/cs11/report.html

Webster’s New Collegiate Dictionary. (1973). Springfield, Massachusetts: G. & C. Merriam Company.

Wilkes, M.V. (1992, August). Artificial intelligence as the year 2000 approaches. Communications of the ACM, 35, 17-20.