APPLIED ARTIFICIAL INTELLIGENCE
∑ Artificial Intelligence (AI) is a subdivision of computer science devoted to creating computer software and hardware that attempt to produce results such as those produced by people.
∑ AI technologies advance us one step further: they enable us to automate or enhance more complex tasks that were done manually.
∑ AI technologies are being integrated with other computer-based information systems so that the capabilities and applicability of computers are greatly increased.
∑ Most experts agree that AI is concerned with two basic ideas.
∑ First, it involves studying the thought processes of humans (to understand what intelligence is); second, it deals with representing those processes via machines (computers, robots, etc.).
∑ One well-publicized definition of AI is as follows:
AI is behavior by a machine that, if performed by a human being, would be called intelligent.
∑ A thought-provoking definition is provided by Rich:
AI is the study of how to make computers do things at which, at the moment, people are better.
∑ Mark Fox of Carnegie-Mellon University often says that AI is basically a theory of how the human mind works.
∑ Winston and Prendergast list three objectives of AI:
1. Make machines smarter (primary goal)
2. Understand what intelligence is (the Noble laureate purpose)
3. Make machines more useful (the entrepreneurial purpose)
∑ What is the meaning of the term intelligent behavior?
∑ Several abilities are considered signs of intelligence:
Learn or understand from experience
Make sense out of ambiguous or contradictory messages
Respond quickly and successfully to a new situation (different responses, flexibility)
Use reason in solving problems and directing conduct effectively
Deal with perplexing situations
Understand and infer in ordinary, rational ways
Apply knowledge to manipulate the environment
Acquire and apply knowledge
Think and reason
Recognize the relative importance of different elements in a situation
AI is that branch of computer science dealing with symbolic, nonalgorithmic methods of problem solving.
o This definition focuses on two characteristic of computer programs:
1. Numeric versus Symbolic:
o Computers were originally designed specifically to process numbers (numeric processing).
o Peoples, however, tend to think symbolically; our intelligence seems to be based, in part, on our mental ability to manipulate symbols rather than just numbers.
o Although symbolic processing is at the core of AI, this does not mean that AI does not involve math; rather, the emphasis in AI is on manipulation of symbols.
2. Algorithmic versus Nonalgorithmic:
o An algorithm is a step-by-step procedure that has well-defined starting and ending points and that is guaranteed to reach a solution to a specific problem.
o Computer architecture readily lends itself to this step-by-step approach.
o Many human reasoning processes, however, tend to be nonalgorithmic; in other words, our mental activities consist of more than just following logical, step-by-step procedures.
o Heuristics (rules of thumb) are included as a key element of AI in the following definition:
AI is the branch of computer science that deals with ways of representing knowledge using symbols rather than numbers and with rules-of-thumb, or heuristics, methods for processing information.
o People frequently use heuristics, consciously or otherwise, to make decisions.
o By using heuristics one does not have to rethink completely what to do every time a similar problem is encountered.
o AI involves an attempt by machines to exhibit reasoning capabilities.
o The reasoning consists of inferencing from facts and rules using heuristics or other search approaches.
o AI is unique in that it makes inferences by employing the pattern-matching (or recognition) approach.
o The following definition of AI focuses on pattern-matching techniques:
AI works with pattern-matching methods which attempt to describe objects, events, or processes in terms of their qualitative features and logical and computational relationships.
Artificial versus Natural Intelligence
∑ According to Kaplan, AI has several important commercial advantages:
AI is more permanent. Natural intelligence is perishable from a commercial standpoint in that workers can change their place of employment or forget information. AI, however, is permanent as long as the computer systems and programs remain unchanged.
AI offers ease of duplication and dissemination. Transferring a body of knowledge from one person to another usually requires a lengthy process of apprenticeship; when knowledge is embodied in a computer system, it can be copied from that computer and easily moved to another computer.
AI can be less expensive than natural intelligence. There are many circumstances in which buying computer services costs less than having corresponding human power carry out the same tasks (over the long run).
AI, being a computer technology, is consistent and thorough. Natural intelligence is erratic because people are erratic.
AI can be documented. Decisions made by a computer can be easily documented by tracing the activities of the system. Natural intelligence is difficult to reproduce; for example, a person may reach a conclusion but at some later date may be unable to re-create the reasoning process that led to that conclusion or to even recall the assumptions that were a part of the decision.
∑ Natural intelligence does have several advantages over AI:
Natural intelligence is creative, whereas AI is rather uninspired.
Natural intelligence enables people to benefit from and use sensory experience directly, whereas most AI systems must work with symbolic input.
Human reasoning is able to make use at all times of a wide context of experience and bring that to bear on individual problems; in contrast, AI systems typically gain their power by having a very narrow focus.
Knowledge in AI
o In the field of information systems it is customary to distinguish between data, information, and knowledge.
o The term data refers to numeric (or alphanumeric) strings that by themselves do not have a meaning.
o They can be facts or figures to be processed.
o Information is data organized so that it is meaningful to the person receiving it.
o Knowledge has several definitions.
o For example, according to the Websterís New World Dictionary of the American Language, knowledge is:
a clear and certain perception of something.
all that has been perceived or grasped by the mind.
practical experience, skill.
acquaintance or familiarity.
organized information applicable to problem solving.
o Data, information, and knowledge can be classified by their degree of abstraction and by their quantity.
o Knowledge is the most abstract and exists in the smallest quantity.
o Another definition of knowledge is that given by Sowa:
Knowledge encompasses the implicit and explicit restrictions placed upon objects (entities), operations, and relationships along with general and specific heuristics and inference procedures involved in the situation being modeled.
o Knowledge is also information that has been organized and analyzed to make it understandable and applicable to problem solving or decision making.
o The collection of knowledge related to a problem (or an opportunity) to be used in an AI system is called a knowledge base.
Knowledge Bases and Knowledge-Based Organizations
o With a knowledge base and the ability to draw inferences from it, the computer can now be put to some practical use as a problem solver and decision-maker.
o By searching the knowledge base for relevant facts and relationships, the computer can reach one or more alternative solutions to the given problem.
o The computerís knowledge base and inferencing capability augment those of the user.
o This ability can be applied to different areas ranging from problem solving to the interpretation of languages and scenarios.
o The importance of AI and knowledge bases is rapidly increasing.
o Therefore, many people believe that we are moving from the information age into the knowledge age, and some even talk about knowledge-based organizations and societies.
How AI Differs from Conventional Computing?
∑ AI is software that permits a computer to duplicate some functions of the human brain in a limited way.
∑ Although special AI hardware can be built, most AI is software that runs on mainframes, minicomputers, workstations, and even on personal computers.
∑ AI programs can be written in virtually any computer language and have been written in assembly language, BASIC, Fortran, Pascal, C and Forth.
∑ However, some programming languages have been developed especially for AI applications.
∑ The two most popular AI programming languages are LISP and PROLOG.
o In conventional computing we tell the computer how to solve the problem.
o The computer is given data and a step-by-step program that specifies how the data is to be used to reach an answer.
o In AI, the computer is given knowledge about the subject area plus some inferencing capability.
o The AI program determines the specific procedure for arriving at a solution.
o Conventional computer programs are based on an algorithm, which is a clearly defined, step-by-step procedure for solving a problem.
o It may be a mathematical formula or a sequential procedure that will lead to a solution.
o The algorithm is converted into a computer program (sequential list of instructions or commands) that tells the computer exactly what operations to carry out.
o The algorithm then uses data such as numbers, letters, or words to solve the problem (a typical example is the preparation of a payroll in an organization).
o This process is limited to very structured applications.
o If we give the computer a knowledge base and inferencing capability, however, the computer can be used to support nonstructured applications that so far have been restricted to manual solutions.
o As a result, the computer may become a far more useful tool that supplements and enhances human capabilities.
o AI software is based on symbolic representation and manipulation.
o In AI, a symbol is a letter, word, or number that is used to represent objects, processes, and their relationships.
o Objects can be people, things, ideas, concepts, events, or statements of fact.
o By using symbols, it is possible to create a knowledge base that states facts, concepts, and the relationships among them.
o Then various processes are used to manipulate the symbols to generate advice or a recommendation for solving problems.
o The process in AI is qualitative rather than quantitative as it is in algorithmic computing.
o Virtually all digital computers are algorithmic in their operation, based on the von Neuman concept: instructions stored in memory are executed sequentially to perform some desired operation.
o Algorithmic software can be written in such a way as to permit symbolic representation and manipulation.
o Once a knowledge base is built, some means of using it to solve problems must be developed.
o How does the AI software reason or infer with this knowledge base?
o The basic techniques are search and pattern matching.
o Given some initial start-up information, the AI software searches the knowledge base looking for specific conditions or patterns.
o It looks for matchups that satisfy the criteria set up to solve the problem.
o The computer literally hunts around until it finds the best answer it can give based on the knowledge it has.
o Even though AI problem solving does not take place directly by algorithmic processes, algorithms, of course, are used to implement the search process.
o A word of caution! Some people believe that AI is magic. It is not.
o AI is basically a different approach to programming computers, and it should be treated as such.
o Therefore, although there are some profound differences between AI and conventional computing, there are also many similarities.
o For example, developing an AI system should be treated like the development of any other computer system.
o We use system analysis and design in AI in a manner very similar to the way we use it in other computer systems.
o We may use different terminology, but an AI system is a computer-based information system (CBIS), although it has some different characteristics.
Applications of AI
o Expert systems are computerized advisory programs that attempt to imitate the reasoning processes and knowledge of experts in solving specific types of problems.
o They are used more than any other applied AI technology.
o Expert systems are of great interest to organizations because of their potential to enhance productivity and to augment work forces in many specialty areas where human experts are becoming increasingly difficult to find and retain.
o Current applications are restricted to relatively limited and narrowly defined areas of expertise (called domains).
o Human experts tend to specialize in relatively narrow problem-solving areas or tasks.
o Typically, human experts possess these characteristics: they solve problems quickly and fairly accurately, explain what (and sometimes how) they do, judge the reliability of their own conclusions, know when they are stumped, and communicate with other experts.
o They can also learn from experience, change their points of view to suit a problem, and transfer knowledge from one domain to another.
o Finally, they use tools, such as rules of thumb, mathematical models, and detailed simulations to support their decisions.
o Knowledge is a major resource, and it often lies with only a few experts.
o It is important to capture that knowledge so others can use it.
o Experts get sick or become unavailable and knowledge is thus not always available when needed.
o Books and manuals can capture some knowledge, but they leave the problem of a particular application up to the reader.
o Expert systems can provide a direct means of applying expertise.
o An expert system permits the knowledge and experience of one or more experts to be captured and stored in a computer.
o This knowledge can then be used by anyone who requires it.
o The purpose of an expert system is not to replace the experts, but simply to make their knowledge and experience more widely available.
o Typically, there are more problems to solve than there are experts available to handle them.
o An expert system permits others to increase their productivity, improve the quality of their decisions, and solve problems when an expert is not available.
Natural Language Processing
o Natural language technology gives computer users the ability to communicate with the computer in their native language.
o This technology allows for a conversational type of interface, in contrast to one of computer jargon, syntax and commands.
o Limited success in this area is typified by current systems that can recognize and interpret written sentences relating to very restricted topics.
o Although this ability can be used to great advantage with some applications, a general natural language processing (NLP) system is not yet possible.
o The field of NLP is divided into two subfields:
Natural language understanding investigates methods of allowing the computer to comprehend instructions given in ordinary English so that computers can understand people more easily.
Natural language generation strives to have computers produce ordinary English language so that people can understand computers more easily.
Speech (Voice) Understanding
o Speech understanding is the recognition and understanding by a computer of spoken language.
o Speech understanding is a process that allows one to communicate with a computer by speaking to it.
o The term speech recognition is sometimes applied only to the first part of the process: recognizing the words that have been spoken without necessarily interpreting their meanings.
o The other part of this process, in which the meaning of the speech is ascertained, is called speech understanding.
o It may be possible to understand the meaning of a spoken sentence without actually recognizing every word.
o NLP is an attempt to allow computers to interpret normal statements expressed in a natural human language, such as English or Japanese.
o The process of speech understanding, in contrast, attempts to translate the human voice into individual words and sentences understandable by the computer.
o A combination of speech understanding and NLP will be required to realize the capability of the computer to converse in a manner normal to humans.
Robotics and Sensory Systems
o Sensory systems, such as vision systems, tactile systems, and signal processing systems, when combined with AI define a broad category of systems generally referred to as robotics.
o A robot is an electromechanical device that can be programmed to perform manual tasks.
o The Robotic Institute of America formally defines a robot as "a reprogrammable multifunctional manipulator designed to move materials, parts, tools, or specialized devices through variable programmed motions for the performance of a variety of tasks."
o Not all of robotics is considered to be part of AI.
o A robot that performs only the actions that it has been preprogrammed to perform is considered to be a "dumb" robot, possessing no more intelligence than, say, a dishwasher.
o An "intelligent" robot includes some kind of sensory apparatus, such as a camera, that collects information about the robotís operation and its environment.
o The intelligent part of the robot allows it to respond and adapt to changes in its environment, rather than just to follow instructions "mindlessly."
o Robots combine sensory systems with mechanical motion to produce machines of widely varying intelligence and abilities.
o The research and application areas under the sensory systems umbrella include machines that sense, move, and manipulate their environment.
o Assembly line operations, particularly those that are highly repetitive or hazardous, are beginning to be performed by robots.
o The difference between the automatic machine and an intelligent robot is that the robot senses its environment and modifies its behavior as a result of the information gained.
o The intelligent robot is thought to have humanlike capabilities and attributes.
o For example, some robots are distinguished from regular automation by their ability to deal with uncertainty.
Computer Vision and Scene Recognition
o Visual recognition has been defined as the addition of some form of computer intelligence and decision making to digitized visual information received from a machine sensor.
o The combined information is then used to perform, or control, such operations as robotic movement, conveyor speeds, and production-line quality.
o The basic objective of computer vision is to interpret pictures rather than generate pictures (which preoccupies computer graphics).
o What "interpreting pictures" means differs depending on the application.
o For example, in interpreting pictures taken by satellite, it may be sufficient to roughly identify regions of crop damage.
o On the other hand, robot vision systems may find it necessary to precisely identify assembly components to correctly affix the components to the item being assembled.
o Research in machine vision may enhance the abilities of automated systems by handle the manipulation of unlike objects in multiple orientations, such as forms lying on a table or parts moving on a conveyor belt.
o Optical recognition systems, for example, can retrieve handwritten or typed data from a form and reformat it for storage.
Intelligent Computer-Aided Instruction
o Intelligent computer-aided instruction (ICAI) refers to machines that can tutor humans.
o To a certain extent, such a machine can be viewed as an expert system.
o However, the major objective of an expert system is to render advice, whereas the purpose of ICAI is to teach.
o Computer-assisted instruction, which has been in use for many years, brings the power of the computer to bear on the educational process.
o Now AI methods are being applied to the development of intelligent computer-assisted instruction systems in an attempt to create computerized "tutors" that shape their teaching techniques to fit the learning patterns of individual students.
o ICAI applications are not limited to schools; as a matter of fact, they have found a sizable niche in the military and corporate sectors.
o ICAI systems are being used today for various tasks such as problem solving, simulation, discovery, learning, drill and practice, games, and testing.
o Such systems are also being used to support impaired people.
o Even though ICAI programs are user interactive, use knowledge bases, and employ some AI technologies (like natural language interfaces), there is some debate about whether the programs themselves are really examples of AI.
o Often these programs are databases structured to respond to specific inputs with specific answers within a predetermined structure.
Translation from One Language to Another
o Computer programs are able to translate words and simple sentences from one language to another.
o For example, a package called LOGOS is used for translating from English to German (and German to English).
The Future of AI
∑ So what does the future hold?
∑ First, AI research and development will continue, and all of the various subfields will evolve and improve.
∑ New software techniques will be discovered.
∑ Improved software development tools will be created for easier development of expert systems and other AI applications.
∑ Advances will also occur in hardware.
∑ In addition to the usual ongoing developments in semiconductor technology that will bring us larger and faster microprocessors and RAM chips, entirely new devices will be created.
∑ Special search, pattern-matching, and symbolic computing chips are being developed.
∑ The new parallel computing and especially neural computing architectures, with multiple CPUs operating simultaneously, will bring a whole new dimension to AI.
∑ For the immediate future, you can expect to see AI added to existing software.
∑ Natural language interfaces will become a common feature on many applications programs, and intelligent databases are being developed.
∑ Internally, programs will use segments of AI to make some performance improvements.
∑ Expert systems that advise on many important topics will become widely available.
∑ Generally speaking, however, there will be relatively few stand-alone AI application products.
∑ Expert systems are the exception.
∑ Predictions are that AI software will, in most applications, be combined with conventional algorithmic software; that is, AI subroutines, including expert systems, will be embedded in traditional software.
∑ AI will be virtually transparent to the user.
∑ There is a great concern within the AI and software communities that too much will be expected of AI.
∑ AI researchers are particularly upset over the excessive amount of media hype that AI has been receiving during the past few years.
∑ They fear that the publicís expectations will be far beyond what AI can really deliver.
∑ That is certainly a legitimate concern, but such media buildups occur with many new high-tech methods and products.
∑ So, although AI is an excellent technology, it is not the panacea we might like it to be.
∑ Look at it for what it really is: some special software techniques now developed to the point that can be practical and useful.
∑ Where possible, consider AI another computer-based information system that can expand the applicability of computers and increase productivity and compatibility.
∑ Look for ways to use it, but do not expect miracles.
∑ On rare occasions, AI may give a miracle like solution, but more likely it will not.
∑ It will deliver evolutionary rather than revolutionary improvements.