Artificial Intelligence: An Emerging Technology

Artificial Intelligence: An Emerging Technology
How far has it progressed? Where will it go from here?

Samantha A. Edwards
December 13, 1993

Copyright © 1993 - 1996 by Samantha A. Edwards
All rights reserved. This paper may not be reproduced without the expressed written consent of the author.

Topics

Original Hopes and Expectations
Present Applications
Future Applications
Footnotes

Since the invention of the computer, science-fiction films and books have been predicting that these machines would become intelligent enough to take over the world, then reduce humankind to servitude. This pessimistic prediction is far from reality: Computers have a huge capacity for processing information quickly but scientists have yet to create a machine with human-like "common sense" and adaptability in new situations. As Anatol Holt, a researcher in the field of artificial intelligence, once noted: "A brilliant chess move while the room is filling with smoke because the house is burning down does not show intelligence. If the capacity for brilliant chess moves without regard to life circumstances deserves a name, I would call it "artificial intelligence."¹ Artificial Intelligence (AI) scientists have been struggling for decades to perfect their intelligent creations but as Herbert Simon (one of the pioneers of AI research at Carnegie-Mellon University) recalls, "the biggest surprise of his decades spent trying to recreate human intelligence was 'how easy the 'hard' things were to do, and how hard the 'easy' things."² In order to illustrate this dilemma, Simon describes his first AI program, completed in 1955, that could work out logical theorems but since then, no-one has built a machine that can navigate across a crowded room or understand a children's story. This "curious combination of brilliance and stupidity" continues to confound the AI scientists who strive to create humanoids like Data from Star Trek: The Next Generation.³

It is interesting to note that even though AI's present limitations frustrate the researchers, the American artificial intelligence industry has thrived from AI's progress so far. There are numerous applications on the market that promise to facilitate every aspect of our lives, from business, to manufacturing, to home. This paper will discuss the progress of AI to date, as well as forecasting its uses in the future. AI is a complex subject with varied fields of concentration, but the potential for future business opportunities make it an emerging technology well worth harnessing and exploiting. Before describing AI's present and potential applications, it is essential to review the development and expectations of the AI field.

THE ORIGINAL HOPES AND EXPECTATIONS OF ARTIFICIAL INTELLIGENCE
In the beginning.... In 1956, Marvin Minsky(computer scientist), John McCarthy (mathematician) and other young scientists set up the assumptions that have guided AI research since then. During the two month colloquium at Darmouth, Minsky formed the "classical top-down postulate that any thinking machine must emulate the problem-solving ability associated with rational thought. For a machine to be truly intelligent, ... it must possess the symbol-manipulating, rule-implementing proficiency that characterizes human reasoning."⁴ This conclusion was based on the assumption that computers could be equivalent to human minds if they could pass the Turing test as efficiently as a human. Alan Turing, the father of AI, devised his test to prove that computers could think if given the right programs. The "right" program "can perform in such a way that an expert cannot distinguish its performance from that of a human who has a certain cognitive ability, then the computer also has that ability."⁵ This definition implies that if a computer performs like a human expert, it literally has a mind like a human. This definition is flawed if you consider that a program can simulate the process of thought and come to the correct conclusion without understanding. Searle, a philosopher who disagrees with Turing's "strong AI" beliefs, argues: "Just manipulating the symbols is not by itself enough to guarantee cognition, perception, understanding, thinking .... Since computers ... are symbol-manipulating devices, merely running the computer program is not enough to guarantee cognition."⁶ Many "weak AI" scientists believe, as Searle does, that computers and their processes enable us to study the mind even though they do not have minds themselves. This less literal approach to AI probably led the development of the "expert systems" that we will discuss later.

If it can think, then walking will be easy.... Minsky and his associates took their theory one step further by saying that once a machine had the thinking program as intelligent as a human, it could easily learn to move in and interact with its environment. Unfortunately, when they first attached their "thinking" computers to robotic arms and television cameras, "the early results were like a cold shower. While pure reasoning programs did their jobs about as well as college freshman, the best robot control programs ... took hours to find and pick up a few blocks on a table top, and often failed completely, performing much worse than a six-month-old child."⁷ They discovered that the computers could analyze and implement their tasks if there was no limit to the time taken to do it. Real-time experiments and real robotic arms had to be replaced by simulation programs to simplify the process: The computers were taking so much time going through every possible movement variation that they could hardly move. This branch of AI research has progressed substantially over the last two decades but some scientists were frustrated enough to come up with another approach.

It thinks like an cockroach.... Rodney Brooks, considered a radical in AI circles, focuses his research on the bottom-up approach. He found his inspiration while watching flies buzzing around a room. If a "stupid" fly could bomb around without bumping into anything, he questioned, then what could he remove from the classical top-down approach to help the "intelligent" robots to move. His argument is simple: "sensorimotor skills, not higher-level thought processes, are the foundation on which intelligence is built. In other words, robots are going to have to learn to crawl before they learn to think .... Intelligent inorganic life will evolve, like organic life before it, from simpler organisms.⁸ Brooks ideas may seem reasonable to the average person but classical AI researchers oppose Brooks' insect approach. Minsky criticized Brooks' methodology with irony: "Hey, maybe we should all just devote ourselves to replicating insect intelligence."⁹ This conflict developed because Brooks sees his way as the only way whereas other researchers see his breakthroughs as "merely a way of building a competent mobile platform that could carry a conventional, cogitating brain."¹⁰ It is interesting to note that Brooks' staunchest critics cannot deny that his robots (Genghis and Attila are the most recent additions to his menagerie at MIT's Mobile Robot Project) move about and adapt faster to their environment than any of their "intelligent" predecessors.

Brooks' "subsumption architecture" strategy enables robots, like Attila, to walk to their destinations without using the usual symbolic models of the world required by classical AI. David Freedman describes this process as: "a menu of primitive instincts and knee-jerk reactions such as TRACK PREY, MOVE FORWARD, and BACK OFF. His robots have no central brain that chooses and blends these simple behaviors. Instead, each behavior acts as an individual "intelligence" that competes for control of the robot. The winner is determined by what the robot's sensors detect at any particular moment, at which point all other behaviors are temporarily subsumed."¹¹ Brooks's critics argue that his machines are stupid but he points out that for what we want robots to do, they are complex enough. More importantly, they are "faster, cheaper, and more reliable" and therefore, have many more immediate product possibilities.¹² Brooks envisions legions of his robots working together to complete a shared goal. He is presently building twenty identical small robots in the hope that they will react to each other like unthinking social insects, such as bees or termites. Once they have mastered these behaviors, Brooks believes that they will easily assimilate more complex demands because of the "modular nature of subsumption architecture." He explains that "distributing the problem-solving ... makes adding higher level competencies simpler and simpler, because you're not always trying to have this central intelligence understand everything."¹³ If Brooks is correct, these simple interacting robots could work in packs to nibble our lawns to the right length, remove the dust from our carpets at night, or unclog our blocked arteries.¹⁴

While Brooks thinks smaller, others think bigger.... From the examples above, it is evident that Brooks wants to incorporate the simplicity and size of insects into his robots. He eventually wants his "gnat robots" to be "carved from a single crumb of silicon -- brains, motors, and all -- at an eventual cost of pennies apiece."¹⁵ This idea contrasts sharply with the AI projects that presently receive the most government funding. It seems as though AI researchers have taken the government's preference for "big science" projects quite literally: Hans Moravec (affiliated with Carnegie-Mellon's Field Robotics Center) received NASA funding for his 12-foot-high, six-legged rover.¹⁶ His other projects, the Tesselator and the NavLabs, are equally large and absorb equally enormous amounts of government money. While they may have been developed to perform dangerous space shuttle maintenance for NASA or to transport materials in war zones for DARPA (now ARPA), it is hard to justify such an expense when you consider the risk inherent in new technology.¹⁷ Brooks' cheaper, simpler robots seem like lower risk investments since he can build a few machines for what it costs Moravec to produce one. Unfortunately for the US economy, foreign high-tech companies have been more supportive of this emerging technology: This year, ARPA will fund one of Brooks's projects for the first time while the project's co-sponsor, Matsushita, has been funding MIT's Mobile Robot Project since its conception. American basic research has been historically more advanced than other nations and the venture capital industry will have to maintain that position. Robot applications of AI may not be explicitly profitable but any advances made in understanding human or computer thought processes will affect the AI field for decades.

And then there is Fuzzy Logic.... If we view AI programming as either symbolic (Minsky) or connectionist (Brooks), Fuzzy Logic is neither one. Both AI programming methods use traditional computer reasoning: that is, precise membership requirements shuffle inputs into specific sets with rigid rules. For example, if a traditional program has one set representing "old" and another representing "young," the age inputted is considered old or young. With Fuzzy Logic, it is possible to have sets that represent more gradual degrees of age. Fuzzy sets have "more flexible membership requirements that allow for partial membership in a set. The degree to which an object is a member of a fuzzy set can be any value between 0 and 1, rather than strictly 0 or 1 as in a traditional set. With a fuzzy set, there is a gradual transition from membership to nonmembership."¹⁸ The Fuzzy Logic branch of mathematics, invented in 1965 by Dr. Lofti Zadeh, is popular because it is modelled after how humans respond to the real world in real-time. For instance, humans say a person is short or tall, old or young, thin or fat, but these sets are not precise since everyone has a slightly different definition for each word (see Figure A): Fuzzy Logic permits computers to categorize in an equally imprecise way.¹⁹

[Crisp vs. Fuzzy Sets]

Figure A: Crisp vs. Fuzzy Sets

It is ironic that this seeming imprecision makes products that corporate Fuzzy Logic "intelligent" or "smart." Zadah's mathematics is not imprecise, as its nickname implies, and in fact, US industry now lags behind Japan in its use of Fuzzy Logic technology because of this American misnomer. While Japanese companies created products (such as rice cookers, microwaves, and washing machines²⁰) which incorporate their newly developed fuzzy-expert systems, US manufacturers shied away from it because its name implied uncertainty. American firms understood that Fuzzy Logic output an undesirable fuzzy result. The opposite was true: "a fuzzy system takes the combined fuzzy output and converts it into a crisp, numerical result through a process called defuzzification. The procedure is mathematically complex; it involves finding the center of gravity, or the centroid, of the combined fuzzy output set.(see Figure B)"²¹ This mistaken interpretation indicates that the managers in American industry were either too uneducated to appreciate the subtlety of Fuzzy Logic, or were too comfortable with present technology to consider something new.

[Centroid Defuzzification]

Figure B: the Centroid Defuzzification Method

How do expert-systems fit into AI?.... "Weak AI" researchers developed these systems that have become the most visible commercial applications of artificial intelligence. The simplest definition of an expert system is "a program which embodies 'expertise' about something and which allows the user to ask the computer certain questions about that expertise."²² Expert systems can take months or years to construct because a knowledge engineer must extract knowledge from a human expert, think of a useful way to present the information, then write the program so that the computer can utilize the program's expertise. All of the components of an expert system, namely the user interface, the knowledge-acquisition module, the knowledge base, the inference engine, and the explanation system, work together to extract and explain the program's expertise.²³ The knowledge base and inference engine are the key components since this separation of control (inference engine) and knowledge (knowledge base) is unique to expert systems.²⁴ There are various types of expert systems in use -- advisory systems, clerical checking systems, ordering and configuring systems, real-time monitoring systems, and battlefield systems -- but they all work in basically the same way. First of all, the user "consults the knowledge base asking and responding to on-screen menus and questions, the inference engine processes the expertise in the knowledge base, then it send its conclusion to the explanation system so that it can explain the reasoning behind the inference engine's conclusion."²⁵ Since traditional AI programs cannot trace the reasoning behind their results, the explanation is also a feature unique to expert systems. Users feel more comfortable acting on advice given to them by an expert system because they know the reasoning process behind the conclusion. Expert systems are fallible, like any other rule-based program, so the explanation system allows the user to rule out implausible conclusions quickly. On the other hand, it also permits the user to analyze a line of reasoning that he/she had not considered.

It is important to remember that an expert system is limited to the set of rules in its program, therefore it only has expertise when "limited to simple, self-contained jobs which require no common sense reasoning."²⁶ Often referred to as "the bureaucrats of AI," these systems can misdiagnose a problem if the user's inputs stray, even slightly, from the program's parameters.²⁷ MYCIN, one of the first expert systems, demonstrates the limited expertise that these programs have. Edward Shortliffe (MYCIN's creator at Stanford) hoped that his program would aid doctors in their analyses of bacterial infections. When doctors inputted the symptoms of an infection, MYCIN's treatment recommendations were often better than newly qualified doctors. Unfortunately, since its goal was only to interpret symptoms in terms of bacterial infections, MYCIN "cannot distinguish between a patient who needs treatment for a bacterial infection and one who needs a midwife."²⁸ These significant inadequacies have not yet been overcome so it is important that expert systems are treated as people's assistants, not their masters as AI fanatics originally believed. An expert system can increase an educated user's productivity, but it is useless if its user is ignorant of its extensive knowledge base and applications. This point is essential to the development of the future competitiveness of American expert system producers: If American workers are not trained to take advantage of their intelligent computers, companies in the market will lose international competitiveness due to their limited domestic market. The US educational system must train its students to use up-to-date technologies if the US is to remain ahead in the global AI market. In order to forecast future business opportunities in AI, we must look at specific AI applications already in use.

PRESENT APPLICATIONS OF ARTIFICIAL INTELLIGENCE
Fuzzy Logic and Expert Systems lead the way.... Fuzzy Logic and Expert Systems are already on the market, helping people to simplify their daily routines. Here are a few of the firms that incorporate AI applications into their current products:

NISSAN has recently added the Laurel, a fuzzy-logic five-speed automatic auto, to its Japanese sedan line. Having fuzzy logic in the gear control system gives the driver a "freer choice of gears ratios (i.e. a lower first and a higher fifth). The fuzzy logic prevents premature shifting from fourth to fifth gear on uphill sections, even when the accelerator pedal is eased." This fuzzy feature preserves the auto's transmission and reduces revving that increases fuel consumption. Many other auto makers have added fuzzy systems to their latest production lines.²⁹
FUJITSU LTD. has incorporated fuzzy-logic control systems into air conditioners, heating and refrigeration units, washing machines, microwave ovens, vending machines, and automated currency exchange machines. Fuzzy Logic, called fuaji riron ('fuaji' being a nonsense word that sounds like the English word, fuzzy and 'riron' meaning theory), has become the latest buzz word in high-tech Japanese electronics.³⁰ Fuzzy logic became so popular that Fujitsu even claimed to have incorporated it into products when it had not. The company recently acknowledged its erroneous marketing strategy and is working to replace "unsophisticated software that merely mimics the real thing" with actual fuzzy controls.³¹ In an effort to clean its soiled fuzzy software image, Fujitsu and other industry giants have promised the Japanese consumers that they will also develop fuzzy hardware..³²
ARRIS PHARMACEUTICAL CORP. is using expert-systems to enhance their chemists' knowledge of biology. The program scans "hundreds of thousands of chemicals, recording each one's individual shape and size" so that the chemist can use the programs knowledge base to cross-reference any new molecular discoveries. Arris hopes that the program will take some of the guess work usually involved in drug discovery. Since the expert-system speeds up the chemists' work, Arris hopes that it will soon develop a drug that can decrease the progression of diseases, such as atherosclerosis, which clog and harden arteries.³³
KURZWEIL APPLIED INTELLIGENCE INC. has a speech recogition system that doctors use to fill out medical charts quickly and accurately while still in the emergency room.³⁴ This $27,000 expert-system has many benefits: "The system costs far less than a clerical worker or two and pays for itself in six months .... And [its] built-in 'knowledge base' retrieves and feeds medical data to the doctor, prompting him to check for symptoms he may have missed."³⁵ Doctors love the systems accuracy in this era of ever-increasing malpractice anxieties. Kurzweil's system is more advanced than most voice-recognition systems in use since its database stores the most common phonemes, or normal speech sounds. Since it stores frequently repeated sounds, rather than words, the program requires much less computer power than tradition programs.³⁶ This storage method also enables the system to "process words that are articulated at a normal conversational cadence."³⁷ This breakthrough will surely speed Kurzweil's method to the forefront of the speech-recognition field.
VERBEX VOICE SYSTEMS INC. have also developed a speech-recognition system that helps traders at Lehman Brothers record their transactions as soon as they occur. They take special phones onto the trading floor and as they speak into them, their trade orders appear on their computer screens.³⁸ The speech-recognition program's only drawback is that it cannot understand what the trader is saying if the noise in the room gets too loud. The only way to adapt the system for such extreme situations is to train it to respond to the user's voice when the most possible noise is in the background. Expert systems such as these have varying degrees of intelligence since they can either be "speaker-trained," "speaker-independent" or "speaker adaptive." Speaker-trained systems require that a user record the vocabulary that the computer will then recognize when he/she next uses those words; speaker-independent systems have a limited vocabulary but recognize any user's voice; and speaker-adaptive systems begin by being independent but become trained as they adapt to the user's voice.³⁹ According to Voice Information Associates, this speech-recognition application of AI promises to grow by as much as 55.5% a year. If these business-oriented products are successful, we may soon have speech-reactive computer systems in our homes: Apple Computer Inc. have already announced that some of their systems will offer speech recognition as an option.⁴⁰

FUTURE APPLICATIONS OF ARTIFICIAL INTELLIGENCE
Some companies are developing new products that will incorporate existing AI technology in order to update their aging product lines. Other companies are investing much further into the future by funding basic research centers that are intended for advancing AI, as well as creating new technologies. Here are a few companies who have various degrees of interest in AI:

PANASONIC has used fuzzy logic technology in its developing product, the Home Cleaning Robot.⁴¹ The robot will learn the dimensions of your dusty carpet by travelling around the edge of the room. Then, it moves back to its starting point and vacuums until the carpet is clean, or its batteries are depleted. The robot will stop vacuuming when it only has enough power to travel back to its recharging stand. It will repeat this pattern until it finishes the room. Because of the built-in fuzzy logic technology, the Home Cleaning Robot can determine how much suction power and beater bar speed it needs to complete each job.⁴² This means that it can sense the different needs for cleaning a wood floor, as opposed to a shag carpet. It can also determine how much dust is on the floor. Panasonic has not yet announced the release date and price for its Home Cleaning Robot, but it should be a hit in the American market if it works well and does not cost too much. Panasonic already has vacuum cleaners out that use fuzzy sensors to find dust but these machines are not self-propelled.
UNITED TECHNOLOGIES CORP. envisions a "smart" air conditioner that will learn to regulate room temperature according to your exact specifications. Amal Kumar Naj explains: "Imagine an air conditioner that starts automatically ... the moment you walk into the room. And what's more, it recognizes you, remembers your preference for comfort and maintains the room's temperature, humidity and air flow to your liking."⁴³ Researchers are developing fuzzy and other AI systems so that, if more than one person is in the room, the air conditioner can regulate an environment which pleases everybody. Carrier (United Technologies air conditioning division) presently holds 11% of the global market share but its Japanese competitors, which hold between 8 to 4%, have already incorporated some fuzzy features into their products that Carrier is still developing.⁴⁴

RICOH CO. LTD. has focused its research on neural technology, the AI connectionists (Brooks) method of mimicking the physical aspects of the brain. It is different from fuzzy technology because it does not require complicated pre-programming before a neural-network system can work. "Neural networks can learn by themselves," as Brooks' robots can.⁴⁵ Ricoh's recently released experimental machine (that uses neural circuits) is predicted to give PCs more computational power than existing supercomputers. The experimental machine is "a general-purpose desktop neurocomputer that runs without software and acquires all computing capabilities through learning .... The unit is some 500 times faster than engineering workstations and about 4 times speedier than today's supercomputers," claims its developers.⁴⁶ Ricoh hopes that its neural chips will succeed the fuzzy logic chips already used in numerous products. A comparison study of the two technologies indicates that Ricoh's neural chips may overtake fuzzy chips in the next century, but fuzzy logic is expected to dominate until at least 1998 (see Figures C, D & E).⁴⁷ The number of fuzzy patents released in Japan demonstrates that this AI technology will dominate the market until the US capitalizes on this present development lead in neural networks.⁴⁸ So far, US companies have paid little attention to the innovative AI labs in academia whereas foreign companies are funding major projects. Matsushita's involvement in Brooks' MIT Mobile Robot Lab was an example used earlier in this discussion. US firms must follow the example of their foreign competitors if the want to maintain their international market position.

Figure C: Neural vs. Fuzzy Networks 1991	Figure D: Neural vs. Fuzzy Networks 1998
Figure E: Fuzzy Patents

NEC CORP. is another example of a foreign company that is taking advantage of American basic research advances. NEC's newly established Research Institute in New Jersey uses an old-fashioned philosophy in an attempt to attract the most innovative researchers, namely: "The best way to make money may be not to worry about making money. The company hopes that the freedom it gives researchers will lure some of the best young computer and biotechnology scientists to what experts is quickly becoming a world-class basic research lab."⁴⁹ NEC evaluates the progression of its researchers' work after three years to determine whether or not it can improve the company's competitiveness in or understanding of computers and communications.⁵⁰ NEC's basic research focus contrasts with the other US labs "fun-and profit" mentality. AT&T (Bell Labs), IBM and GE all fund larger labs than NEC but the mission of NEC's smaller Research Institute continues to attract scientists who dream "of being able to do what you want to do, unfettered by company bean counters."⁵¹ Since NEC focuses on funding the best computer scientists in the US, it is probable that one of its researchers will invent the next breakthrough in AI technology. If you fund the best minds, it is only a matter of time before you harness the best ideas: NEC hopes that this assumption will be profitable for them in the future.

Thoughts on America's future.... Artificial Intelligence promises to revolutionize human/machine relations but US government and industry must fund more R&D if it wants to maintain our technological lead. The Japanese have capitalized on fuzzy logic technology, creating "smart" or "intelligent" features of numerous home and office appliances. America could use its advantage in general AI research to jettison ahead of Japan in global sales by the beginning of the next century. Unfortunately, the US government has focused its funding on "big science" space and defence applications so US industry lags behind Japan in smart appliance innovation. Artificial Intelligence has been put of the Department of Commerce's "Emerging Technologies List," as well as the Department of Defence's "Critical Technologies List" but these moves are not enough. ARPA and projects such as SEMATECH must increase their activity in AI research if the US is to remain ahead. Furthermore, the US must improve its secondary school education's record in mathematics and science or no amount of funding will sustain the domestic high-tech markets. If the American public is unable to appreciate new technologies, such as AI, we cannot expect industry to continually innovate.

There is a more insidious problem: American companies have become so accountable to their stockholders (and their need for bigger and bigger profit margins) that R&D funding has declined in recent years. The venture capital markets continue to be strong but the traditional high-tech companies must also invest in the American economy's future by supporting R&D in-house and in academia. Firms must stop resting on the profits created by decades old technology and concentrate on getting the economy back to being innovation-driven instead of being wealth-driven has it has been in recent years. US companies should benefit from the knowledge and innovative scientists that it has nurtured, instead of foreign investors. AI is one of the emerging technologies that can maintain the US technological edge for decades to come.

FOOTNOTES
¹ John Browning, "Cogito, ergo something," The Economist: A Survey of Artificial Intelligence, London, March 14, 1992, p. 5.
² John Browning, "Cogito, ergo something," The Economist: A Survey of Artificial Intelligence, London, March 14, 1992, p. 5.
³ Mark Dery, "Terminators," Rolling Stone, June 10, 1993, p. 20.
⁴ Mark Dery, "Terminators," Rolling Stone, June 10, 1993, p. 20.
⁵ John R. Seale, "Is the Brain's Mind a Computer Program?," Scientific American, January, 1990, p. 26.
⁶ John R. Seale, "Is the Brain's Mind a Computer Program?," Scientific American, January, 1990, p. 26.
⁷ Mark Dery, "Terminators," Rolling Stone, June 10, 1993, p. 20.
⁸ Mark Dery, "Terminators," Rolling Stone, June 10, 1993, p. 16.
⁹ David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 47.
¹⁰ David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 47.
¹¹ David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 46.
¹² David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 47.
¹³ David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 47.
¹⁴ David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 47.
¹⁵ David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 43.
¹⁶ David Freedman, "Invasion of the Insect Robots," Discover, March, 1991, p. 49.
¹⁷ Mark Dery, "Terminators," Rolling Stone, June 10, 1993, p. 16.
¹⁸ Fuzzy Logic Primer: A Brief Introduction to Fuzzy Logic, Togai InfraLogic, Inc., 1988 - 1992, p. 1.
¹⁹ Fuzzy Logic Primer: A Brief Introduction to Fuzzy Logic, Togai InfraLogic, Inc., 1988 - 1992, p. 1-2.
²⁰ David Hulme, "Japan starts Fuzzy Logic II," Machine Design, September 10, 1992, p. 16.
²¹ Fuzzy Logic Primer: A Brief Introduction to Fuzzy Logic, Togai InfraLogic, Inc., 1988 - 1992, p. 5.
²² William bains and Jenny Raggett, Artificial Intelligence from A to Z, Chapman & Hall, 1992, p. 77.
²³ William bains and Jenny Raggett, Artificial Intelligence from A to Z, Chapman & Hall, 1992, p. 77-78.
²⁴ Dennis Mercadel, Dictionary of Artificial Intelligence, Van Nostrand Reinhold, New York, 1990, p. 96.
²⁵ William bains and Jenny Raggett, Artificial Intelligence from A to Z, Chapman & Hall, 1992, p. 79.
²⁶ "Bureaucrats of the mind," The Economist: A Survey of Artificial Intelligence, London, March 14, 1992, p. 12.
²⁷ "Bureaucrats of the mind," The Economist: A Survey of Artificial Intelligence, London, March 14, 1992, p. 11.
²⁸ John Browning, "Cogito, ergo something," The Economist: A Survey of Artificial Intelligence, London, March 14, 1992, p. 5.
²⁹ "Nissan Laurel offers fuzzy-logic five-speed auto," Automotive Engineering, May, 1993, p. 82.
³⁰ David Hulme, "Japan starts Fuzzy Logic II," Machine Design, September 10, 1992, p. 16.
³¹ David Hulme, "Japan starts Fuzzy Logic II," Machine Design, September 10, 1992, p. 16.
³² David Hulme, "Japan starts Fuzzy Logic II," Machine Design, September 10, 1992, p. 16.
³³ Udayan Gupta, "Designing Drugs: Computers promise to speed up the development, and improve the effectiveness, of new medicines," Wall Street Journal, April 6, 1992, p. 20.
³⁴ John J. Keller, "Computers get powerful �hearing� aids: Improved methods of voice recognition," Wall Street Journal, April 7, 1992, p. 1.
³⁵ John J. Keller, "Computers get powerful �hearing� aids: Improved methods of voice recognition," Wall Street Journal, April 7, 1992, p. 1.
³⁶ John J. Keller, "Computers get powerful �hearing� aids: Improved methods of voice recognition," Wall Street Journal, April 7, 1992, p. 1.
³⁷ John J. Keller, "Computers get powerful �hearing� aids: Improved methods of voice recognition," Wall Street Journal, April 7, 1992, p. 1.
³⁸ John J. Keller, "Computers get powerful �hearing� aids: Improved methods of voice recognition," Wall Street Journal, April 7, 1992, p. 1.
³⁹ John J. Keller, "Computers get powerful �hearing� aids: Improved methods of voice recognition," Wall Street Journal, April 7, 1992, p. 1.
⁴⁰ John J. Keller, "Computers get powerful �hearing� aids: Improved methods of voice recognition," Wall Street Journal, April 7, 1992, p. 1.
⁴¹ Andy Pargh, "Vacuum cleaners getting smarter," Design News, April 6, 1992, p. 160.
⁴² Andy Pargh, "Vacuum cleaners getting smarter," Design News, April 6, 1992, p. 160.
⁴³ Amal Kumar Naj, "In the lab: Air Conditioners learn to sense if you're cool," Wall Street Journal, August 31, 1993, p. 1.
⁴⁴ Amal Kumar Naj, "In the lab: Air Conditioners learn to sense if you're cool," Wall Street Journal, August 31, 1993, p. 1.
⁴⁵ Stuart Dambrot, "Neural Networks challenge fuzzy logic's reign," Electronics, July 13, 1992, p. 26.
⁴⁶ Stuart Dambrot, "Neural Networks challenge fuzzy logic's reign," Electronics, July 13, 1992, p. 26.
⁴⁷ James Manji, "Neural Networks, Fuzzy Logic Markets Will Skyrocket To Near $10 Billion By `98," Controls & Systems, June 1992, p. 12.
⁴⁸ James Manji, "Neural Networks, Fuzzy Logic Markets Will Skyrocket To Near $10 Billion By `98," Controls & Systems, June 1992, p. 12.
⁴⁹ John Keller, "Finding and feeling; Ignoring the bottom line -- NEC's US research lab has a theory: The Freedom not to worry about products may lead to the best products of all," Wall Street Journal, May 24, 1993, p. 12.
⁵⁰ John Keller, "Finding and feeling; Ignoring the bottom line -- NEC's US research lab has a theory: The Freedom not to worry about products may lead to the best products of all," Wall Street Journal, May 24, 1993, p. 12.
⁵¹ John Keller, "Finding and feeling; Ignoring the bottom line -- NEC's US research lab has a theory: The Freedom not to worry about products may lead to the best products of all," Wall Street Journal, May 24, 1993, p. 12.

Previous Page

samantha@nvrlnd.com - © 1996 Samantha Edwards. All Rights Reserved.