Results for artificial intelligence
On this page:
 
Dictionary:

artificial intelligence


n. (Abbr. AI)
  1. The ability of a computer or other machine to perform those activities that are normally thought to require intelligence.
  2. The branch of computer science concerned with the development of machines having this ability.

 
 
Sci-Tech Encyclopedia: Artificial intelligence

The subfield of computer science concerned with understanding the nature of intelligence and constructing computer systems capable of intelligent action. It embodies the dual motives of furthering basic scientific understanding and making computers more sophisticated in the service of humanity.

Many activities involve intelligent action—problem solving, perception, learning, planning and other symbolic reasoning, creativity, language, and so forth—and therein lie an immense diversity of phenomena. Scientific concern for these phenomena is shared by many fields, for example, psychology, linguistics, and philosophy of mind, in addition to artificial intelligence. The starting point for artificial intelligence is the capability of the computer to manipulate symbolic expressions that can represent all manner of things, including knowledge about the structure and function of objects and people in the world, beliefs and purposes, scientific theories, and the programs of action of the computer itself.

Artificial intelligence is primarily concerned with symbolic representations of knowledge and heuristic methods of reasoning, that is, using common assumptions and rules of thumb. Two examples of problems studied in artificial intelligence are planning how a robot, or person, might assemble a complicated device, or move from one place to another; and diagnosing the nature of a person's disease, or of a machine's malfunction, from the observable manifestations of the problem. In both cases, reasoning with symbolic descriptions predominates over calculating.

The approach of artificial intelligence researchers is largely experimental, with small patches of mathematical theory. As in other experimental sciences, investigators build devices (in this case, computer programs) to carry out their experimental investigations. New programs are created to explore ideas about how intelligent action might be attained, and are also developed to test hypotheses about concepts or mechanisms involved in intelligent behavior.

The foundations of artificial intelligence are divided into representation, problem-solving methods, architecture, and knowledge. To work on a task, a computer must have an internal representation in its memory, for example, the symbolic description of a room for a moving robot, or a set of features describing a person with a disease. The representation also includes all the knowledge, including basic programs, for testing and measuring the structure, plus all the programs for transforming the structure into another one in ways appropriate to the task. Changing the representation used for a task can make an immense difference, turning a problem from impossible to trivial.

Given the representation of a task, a method must be adopted that has some chance of accomplishing the task. Artificial intelligence has gradually built up a stock of relevant problem-solving methods (the so-called weak methods) that apply extremely generally.

An important feature of all the weak methods is that they involve search. One of the most important generalizations to arise in artificial intelligence is the ubiquity of search. It appears to underlie all intelligent action. In the worst case, the search is blind. In heuristic search extra information is used to guide the search.

Some of the weak methods are generate-and-test (a sequence of candidates is generated, each being tested for solutionhood); hill climbing (a measure of progress is used to guide each step); means-ends analysis (the difference between the desired situation and the present one is used to select the next step); impasse resolution (the inability to take the desired next step leads to a subgoal of making the step feasible); planning by abstraction (the task is simplified, solved, and the solution used as a guide); and matching (the present situation is represented as a schema to be mapped into the desired situation by putting the two in correspondence).

An intelligent agent—person or program—has multiple means for representing tasks and dealing with them. Also required is an architecture or operating framework within which to select and carry out these activities. Often called the executive or control structure, it is best viewed as a total architecture (as in computer architecture), that is, a machine that provides data structures, operations on those data structures, memory for holding data structures, accessing operations for retrieving data structures from memory, a programming language for expressing integrated patterns of conditional operations, and an interpreter for carrying out programs. Any digital computer provides an architecture, as does any programming language. Architectures are not all equivalent, and one important scientific question is what architecture is appropriate for a general intelligent agent.

In artificial intelligence, the basic paradigm of intelligent action is that of search through a space of partial solutions (called the problem space) for a goal situation. Each step offers several possibilities, leading to a cascading of possibilities that can be represented as a branching tree. The search is thus said to be combinatorial or exponential. For example, if there are 10 possible actions in any situation, and it takes a sequence of 12 steps to find a solution (a goal state), then there are 1012 possible sequences in the exhaustive search tree. What keeps the search under control is knowledge, which suggests how to choose or narrow the options at each step. Thus the fourth fundamental concern is how to represent knowledge in the memory of the system so it can be brought to bear on the search when relevant.

An intelligent agent will have immense amounts of knowledge. This implies another major problem, that of discovering the relevant knowledge as the solution attempt progresses. Although this search does not include the combinatorial explosion characteristic of searching the problem space, it can be time consuming and hard. However, the structure of the database holding the knowledge (called the knowledge base) can be carefully tailored to suit the architecture in order to make the search efficient. This knowledge base, with its accompanying problems of encoding and access, constitutes the final ingredient of an intelligent system.

An example of artificial intelligence is computer perception. Perception is the formation, from a sensory signal, of an internal representation suitable for intelligent processing. Though there are many types of sensory signals, computer perception has focused on vision and speech. Perception might seem to be distinct from intelligence, since it involves incident time-varying continuous energy distributions prior to interpretation in symbolic terms. However, all the same ingredients occur: representation, search, architecture, and knowledge. Speech perception starts with the acoustic wave of a human utterance and proceeds to an internal representation of what the speech is about. A sequence of representations is used: the digitization of the acoustic wave into an array of intensities; the formation of a small set of parametric quantities that vary continuously with time (such as the intensities and frequencies of the formants, bands of resonant energy characteristic of speech); a sequence of phons (members of a finite alphabet of labels for characteristic sounds, analogous to letters); a sequence of words; a parsed sequence of words reflecting grammatical structure; and finally a semantic data structure representing a sentence (or other utterance) that reflects the meaning behind the sounds.

A class of artificial intelligence programs called expert systems attempt to accomplish tasks by acquiring and incorporating the same knowledge that human experts have. Many attempts to apply artificial intelligence to medicine, government, and other socially significant tasks take the form of expert systems. Even though the emphasis is on knowledge, all the standard ingredients are present.

In careful tests, a number of expert systems have shown performance at levels of quality equivalent to or better than average practicing professionals (for example, average practicing physicians) on the restricted domains over which they operate. Nearly all large corporations and many smaller ones use expert systems. A common application is to provide technical assistance to persons who answer customers' trouble calls. Computer companies use expert systems to assist in configuring components from a parts catalog into a complete system that matches a customer's specifications, a kind of application that has been replicated in other industries tailoring assembled products to customers' needs. Troubleshooting and diagnostic programs are commonplace. Another widespread use of this technology is in software for home computers that assists taxpayers. One important lesson learned from incorporating artificial intelligence software into ongoing practice is that its success depends on many other aspects besides the intrinsic intellectual quality, for example, ease of interaction, integration into existing workflow, and costs.

Expert systems have sparked important insights in reasoning under uncertainty, causal reasoning, reasoning about knowledge, and acceptance of computer systems in the workplace. They illustrate that there is no hard separation between pure and applied artificial intelligence; finding what is required for intelligent action in a complex applied area makes a significant contribution to basic knowledge. See also Expert systems.

In addition to the subject areas mentioned above, significant work in artificial intelligence has been done on puzzles and reasoning tasks, induction and concept identification, symbolic mathematics, theorem proving in formal logic, natural language understanding and generation, vision, robotics, chemistry, biology, engineering analysis, computer-assisted instruction, and computer-program synthesis and verification, to name only the most prominent. As computers become smaller and less expensive, more and more intelligence is built into automobiles, appliances, and other machines, as well as computer software, in everyday use. See also Automata theory; Computer; Control systems; Cybernetics; Digital computer; Intelligent machine; Robotics.


 
Modern Science: artificial intelligence
artificial intelligence or AI

The means of duplicating or imitating intelligence in computers, robots, or other devices, which allows them to solve problems, discriminate among objects, and respond to voice command.

 
Real Estate Dictionary: Artificial Intelligence

The ability of a computer program to evaluate information and make decisions according to pre-established criteria. This ability is exploited in such applications as Automated Mortgage Underwriting.Example: Excell Mortgage Company, using artificial intelligence, can pre-approve a large volume of mortgage applications in a short period of time. The process is totally computerized and takes applications through use of an Internet site. The computer can interact with applicants and solicit information as required for the underwriting process.

 
Accounting Dictionary: Artificial Intelligence (AI)

Umbrella terminology for several main categories of research. They include natural language systems, visual and voice recognition systems, robotic systems, and Expert Systems. Artificial intelligence generally is the attempt to build machines that think, as well as the study of mental faculties through the use of computational models. A reasoning process is involved with self-correction. Significant data are evaluated and relevant relationships, such as the determination of a warranty reserve, uncovered. The computer learns which kind of answers are reasonable and which are not. Artificial intelligence performs complicated strategies that compute the best or worst way to achieve a task or avoid an undesirable result. An example of an application is in tax planning involving tax shelter options given the client's financial position.

 
Business Encyclopedia: Artificial Intelligence

Computer systems are becoming commonplace; indeed, they are almost ubiquitous. We find them central to the functioning of most business, governmental, military, environmental, and health-care organizations. They are also a part of many educational and training programs. But these computer systems, while increasingly affecting our lives, are rigid, complex and incapable of rapid change. To help us and our organizations cope with the unpredictable eventualities of an ever-more volatile world, these systems need capabilities that will enable them to adapt readily to change. They need to be intelligent. Our national competitiveness depends increasingly on capacities for accessing, processing, and analyzing information. The computer systems used for such purposes must also be intelligent. Health-care providers require easy access to information systems so they can track health-care delivery and identify the most recent and effective medical treatments for their patients' conditions. Crisis management teams must be able to explore alternative courses of action and support decision making. Educators need systems that adapt to a student's individual needs and abilities. Businesses require flexible manufacturing and software design aids to maintain their leadership position in information technology, and to regain it in manufacturing. (Grosz and Davis, 1994)

The history of artificial intelligence (AI) predates the development of the first computing machines. On a general level, intelligence has been the subject of philosophical study for 2000 years. At the computational level, mathematician Alan Turing constructed a framework for AI during the era of analog computers.

While precise definitions are still the subject of debate, AI may be usefully thought of as the branch of computer science that is concerned with the automation of intelligent behavior. The intent of AI is to develop systems that have the ability to perceive and to learn, to accomplish physical tasks, and to emulate human decision making. AI seeks to design and develop intelligent agents as well as to understand them. Currently, the main fields of research and development include the following:

  1. Natural languages: These studies focus on problems related to natural language interface, machine translation, understanding spoken language, and so forth.
  2. Expert systems: No generalizable solutions are researched, but expertise is used to deal with ill-defined problems and relationships.
  3. Cognition and learning: Investigations are being made into modes of thinking, learning, and problem solving.
  4. Computer vision: Efforts are being made to develop principles and algorithms for machine vision and the interpretation of visual data.
  5. Automatic deduction: This area deals with the resolution of problems, theorem proving, and logic programming.

Foundations

The term "AI" was applied about 1956, giving a formal name to work that had been developing over the previous five or six years. Individuals and organizations have an abiding interest in AI for several important reasons, including the following:

  1. To preserve expertise that might be lost when an acknowledged expert is unavailable.
  2. To create organizational knowledge bases so that others may learn from past problem-solving successes.
  3. To help decision makers be consistent in their evaluation of complex problems.

During its early years AI was dominated by reliance on logic as a means of representing knowledge and on logical inference as the primary mechanism for intelligent reasoning. In the 1990s other paradigms arrived on the scene, some of which had a dramatic impact. Artificial neural networks (ANNs) were motivated by assumptions about how the brain functions— particularly the ideas of massively parallel connections, each of which performs simple computational tasks. Taken together, they represent knowledge as a property of patterns of relationships. Genetic algorithms apply principles of biological evolution to the problems of searching complex solution spaces. The programs do not use logical reasoning either, but evolve toward better and better solutions to complex problems.

Multiagent systems have recently come to the fore of AI research. This emergence has been driven by a recognition that intelligence may be reflected by the collective behaviors of large numbers of very simple interacting members of a community of agents. These agents can be computers, software modules, or virtually any object that can perceive aspects of its environment and proceed in a rational way toward accomplishing a goal.

A variety of disciplines have influenced the development of AI. These include philosophy (logic), mathematics (intractibility, computability, algorithms), psychology (cognition), engineering (computer hardware and software), and linguistics (knowledge representation and natural-language processing).

Long before the development of computers, the notion that thinking was a form of computation motivated the formalization of logic. These efforts continue today. Graph theory provided the architecture for searching a solution space for a problem. Operations research, with its focus on optimization algorithms, used graph theory and other methods to solve complex decision-making problems.

In 1950, Alan Turing proposed what has become known as the Turing Test for defining intelligent behavior. The idea was to specify requirements that a computer would have to exhibit in order to demonstrate intelligence. Briefly, the Turing Test proposes that the computer should be interrogated via telecommunications by a human. Intelligence is exhibited by the computer if the interrogator cannot tell whether there is a human or a computer at the other end. In order to pass the test, a computer would need to have capabilities for natural-language processing, knowledge representation, automated reasoning, and machine learning.

An Evolution of Applications

While computer systems have become commonplace, they are generally rigid, complex, and incapable of rapid change. According to A Report to ARPA on Twenty-First Century Intelligent Systems, for us and our organizations to cope with the unpredictable eventualities of an ever-more volatile world, these systems need capabilities that will enable them to adapt readily to change. The report argues that our national competitiveness depends increasingly on capacities for accessing, processing, and analyzing information (Grosz and Davis, 1994).

One of the early milestones in AI was Newell and Simon's General Problem Solver (GPS). The program was designed to imitate human problem-solving methods. This and other developments such as Logic Theorist and the Geometry Theorem Prover generated enthusiasm for the future of AI. Simon went so far as to assert that in the near-term future the problems that computers could solve would be coextensive with the range of problems to which the human mind has been applied.

Soon difficulties in achieving this objective began to manifest themselves. In scaling up from earlier successes, problems of intractability were encountered. A search for alternative approaches led to attempts to solve typically occurring cases in narrow areas of expertise. This prompted the development of expert systems. A seminal model was MYCIN, developed to diagnose blood infections. Having about 450 rules, MYCIN was able to perform as well as many experts. This and other expert-systems research led to the first commercial expert system, R1, implemented at Digital Equipment Corporation (DEC) to help configure orders for new computer systems. Sub-sequent to R1's implementation, it was estimated to save DEC about $40 million a year.

Other classic systems include the PROSPECTOR program for determining the probable location and type of ore deposits and the INTERNIST program for performing medical diagnosis in internal medicine.

The Future

A Report to ARPA on Twenty-First Century Intelligent Systems identified four types of systems that will have a substantial impact on applications: intelligent simulation, intelligent information resources, intelligent project coaches, and robot teams (Grosz and Davis, 1994).

Intelligent simulations generate realistic simulated worlds that enable extensive affordable training and education that can be made available any time and anywhere. Examples may be hurricane crisis management, exploration of the impacts of different economic theories, tests of products on simulated customers, and technological design—testing features through simulation that would cost millions of dollars to test using an actual prototype.

Intelligent information resources systems (IRSS) will enable easy access to information related to a specific problem. For instance, a rural doctor whose patient presents with a rare condition might use IRSS to help assess different treatments or identify new ones. An educator might find relevant background materials, including information about similar courses taught elsewhere.

Intelligent project coaches (IPC) could function as co-workers, assisting and collaborating with design or operations teams for complex systems. Such systems could remember and recall the rationale of previous decisions and, in times of crisis, explain the methods and reasoning previously used to handle that situation. An IPC for aircraft design, for example, could enhance collaboration by keeping communication flowing among the large, distributed design staff, the program managers, the customer, and the subcontractors.

Robot teams could contribute to manufacturing by operating in a dynamic environment with minimal instrumentation, thus providing the benefits of economies of scale. They could also participate in automating sophisticated laboratory procedures that require sensing, manipulation, planning, and transport.

Conclusion

AI is a young field and faces many complexities. Nonetheless, the Spring 1998 issue of AI Magazine contained articles on the following innovative applications of AI: This is suggestive of the broad potential of AI in the future.

  1. "Case- and Constraint-Based Project Planning for Apartment Construction"
  2. "CREWS–NS: Scheduling Train Crews in The Netherlands"
  3. "An Intelligent System for Case Review and Risk Assessment in Social Services"
  4. "CHEMREG: Using Case-Based Reasoning to Support Health and Safety Compliance in the Chemical Industry"
  5. "MITA: An Information-Extraction Approach to the Analysis of Free-Form Text in Life Insurance Applications"

Bibliography

AI Magazine. (Spring 1998).

Grosz, Barbara, and Davis, Randall, eds. (1994). A Report to ARPA on Twenty-First Century Intelligent Systems.

Luger, George F., and Stubblefield, William A. (1998). Artificial Intelligence: Structures and Strategies for Complex Problem Solving, 3d ed. Reading, MA: Addison-Wesley.

Russell, Stuart J., and Norvig, Peter. (1995). Artificial Intelligence: A Modern Approach. Upper Saddle River, NJ: Prentice-Hall.

[Article by: JAMES V. HANSEN]

 

Abbreviation for ‘Artificial Intelligence’, so common that the full form is almost never written or spoken among hackers.


 
Geography Dictionary: artificial intelligence

The use of computer software in, among others, data collection and processing, analysis, searching for patterns and detecting anomalies, modelling, and problem solving. Applications include genetic algorithms and neural networks.

 
Britannica Concise Encyclopedia: artificial intelligence

Ability of a machine to perform tasks thought to require human intelligence. Typical applications include game playing, language translation, expert systems, and robotics. Although pseudo-intelligent machinery dates back to antiquity, the first glimmerings of true intelligence awaited the development of digital computers in the 1940s. AI, or at least the semblance of intelligence, has developed in parallel with computer processing power, which appears to be the main limiting factor. Early AI projects, such as playing chess and solving mathematical problems, are now seen as trivial compared to visual pattern recognition, complex decision making, and the use of natural language. See also Turing test.

For more information on artificial intelligence, visit Britannica.com.

 
Philosophy Dictionary: artificial intelligence

(or AI) The science of making machines that can do the kinds of thing that humans can do. Topics of research have included speech recognition, visual recognition, and the more familiar problem solving and game-playing. Modelling a psychological phenomenon on a computer is a way of showing how the phenomenon is possible in a physical world, and is also a way of bringing out the complexities involved in apparently simple tasks. A central concept in much AI research is that of a representation, with programs designed to construct, adapt, and link representations in the production of intelligent responses. This research has been responsible for a considerable retreat from dogmatic behaviourism, in which the idea of mental manipulations was thought to be unscientific, since it is exactly the storage and manipulation of representations of the world that is demanded in the problems AI approaches. Strong AI is the philosophical thesis that appropriately programmed computers have minds in exactly the same sense that we do. Weak AI is the methodological belief that the best way to explore the mind is to proceed as if this were true, without commenting on the legacies of dualism that lead to discomfort with the strong thesis. See also Chinese room, connectionism, frame problem, Turing test.

 
US History Encyclopedia: Artificial Intelligence

Artificial Intelligence, a branch of computer science that seeks to create a computer system capable of sensing the world around it, understanding conversations, learning, reasoning, and reaching decisions, just as would a human. In 1950 the pioneering British mathematician Alan Turing proposed a test for artificial intelligence in which a human subject tries to talk with an unseen conversant. The tester sends questions to the machine via teletype and reads its answers; if the subject cannot discern whether the conversation is being held with another person or a machine, then the machine is deemed to have artificial intelligence. No machine has come close to passing this test, and it is unlikely that one will in the near future. Researchers, however, have made progress on specific pieces of the artificial intelligence puzzle, and some of their work has had tangible benefits.

One area of progress is the field of expert systems, or computer systems designed to reproduce the knowledge base and decision-making techniques used by experts in a given field. Such a system can train workers and assist in decision making. MYCIN, a program developed in 1976 at Stanford University, suggests possible diagnoses for patients with infectious blood diseases, proposes treatments, and explains its "reasoning" in English. Corporations have used such systems to reduce the labor costs involved in repetitive calculations. A system used by American Express since November 1988 to advise when to deny credit to a customer saves the company millions of dollars annually.

A second area of artificial intelligence research is the field of artificial perception, or computer vision. Computer vision is the ability to recognize patterns in an image and to separate objects from background as quickly as the human brain. In the 1990s military technology initially developed to analyze spy-satellite images found its way into commercial applications, including monitors for assembly lines, digital cameras, and automotive imaging systems. Another pursuit in artificial intelligence research is natural language processing, the ability to interpret and generate human languages. In this area, as in others related to artificial intelligence research, commercial applications have been delayed as improvements in hardware—the computing power of the machines themselves—have not kept pace with the increasing complexity of software.

The field of neural networks seeks to reproduce the architecture of the brain—billions of connected nerve cells—by joining a large number of computer processors through a technique known as parallel processing. Fuzzy systems is a subfield of artificial intelligence research based on the assumption that the world encountered by humans is fraught with approximate, rather than precise, information. Interest in the field has been particularly strong in Japan, where fuzzy systems have been used in disparate applications, from operating subway cars to guiding the sale of securities. Some theorists argue that the technical obstacles to artificial intelligence, while large, are not insurmountable. A number of computer experts, philosophers, and futurists have speculated on the ethical and spiritual challenges facing society when artificially intelligent machines begin to mimic human personality traits, including memory, emotion, and consciousness.

Bibliography

Kurzweil, Ray. The Age of Spiritual Machines. New York: Viking, 1999.

Partridge, Derek. A New Guide to Artificial Intelligence. Norwood, N.J.: Ablex, 1991.

Shapiro, Stuart C., ed. Encyclopedia of Artificial Intelligence. 2d ed. New York: Wiley, 1992.

Turbam, Efraim. Expert Systems and Applied Artificial Intelligence. New York: MacMillan, 1992.

—Vincent Kiernan/A. R.

 
Columbia Encyclopedia: artificial intelligence
(AI), the use of computers to model the behavioral aspects of human reasoning and learning. Research in AI is concentrated in some half-dozen areas. In problem solving, one must proceed from a beginning (the initial state) to the end (the goal state) via a limited number of steps; AI here involves an attempt to model the reasoning process in solving a problem, such as the proof of a theorem in Euclidean geometry. In game theory (see games, theory of), the computer must choose among a number of possible “next” moves to select the one that optimizes its probability of winning; this type of choice is analogous to that of a chess player selecting the next move in response to an opponent's move. In pattern recognition, shapes, forms, or configurations of data must be identified and isolated from a larger group; the process here is similar to that used by a doctor in classifying medical problems on the basis of symptoms. Natural language processing is an analysis of current or colloquial language usage without the sometimes misleading effect of formal grammars; it is an attempt to model the learning process of a translator faced with the phrase “throw mama from the train a kiss.” Cybernetics is the analysis of the communication and control processes of biological organisms and their relationship to mechanical and electrical systems; this study could ultimately lead to the development of “thinking” robots (see robotics). Machine learning occurs when a computer improves its performance of a task on the basis of its programmed application of AI principles to its past performance of that task.

In the public eye advances in chess-playing computer programs have become symbolic of progress in AI. In 1948 British mathematician Alan Turing developed a chess algorithm for use with calculating machines—it lost to an amateur player in the one game that it played. Ten years later American mathematician Claude Shannon articulated two chess-playing algorithms: brute force, in which all possible moves and their consequences are calculated as far into the future as possible; and selective mode, in which only the most promising moves and their more immediate consequences are evaluated. In 1988 Hitech, a program developed at Carnegie-Mellon Univ., defeated former U.S. champion Arnold Denker in a four-game match, becoming the first computer to defeat a grandmaster. A year later, Gary Kasparov, the reigning world champion, bested Deep Thought, a program developed by the IBM Corp., in a two-game exhibition. In 1990 the German computer Mephisto-Portrose became the first program to defeat a former world champion; while playing an exhibition of 24 simultaneous games, Anatoly Karpov bested 23 human opponents but lost to the computer. Kasparov in 1996 became the first reigning world champion to lose to a computer in a game played with regulation time controls; the Deep Blue computer, developed by the IBM Corp., won the first game of the match, lost the second, drew the third and fourth, and lost the fifth and sixth. Deep Blue used the brute force approach, evaluating more than 100 billion chess positions each turn while looking six moves ahead; it coupled this with the most efficient chess evaluation software yet developed and an extensive library of chess games it could analyze as part of the decision process. Subsequent matches between Vladimir Kramnik and Deep Fritz (2002, 2006) and Kasparov and Deep Junior (2003) have resulted in two ties and a win for the programs. Unlike Deep Blue, which was a specially designed computer, these more recent computer challengers are chess programs that run on powerful personal computers. Such programs have become an important tool in chess, and are used by chess masters to analyze games and experiment with new moves.

Bibliography

See D. Freedman, Brainmakers: How Scientists Are Moving Beyond Computers to Create a Rival to the Human Brain (1994); D. Gelernter, The Muse in the Machine: Computerizing the Poetry of Human Thought (1994).


 
is short for:

American Indian/Alaska Native

 
Wikipedia: artificial intelligence
Garry Kasparov playing against Deep Blue, the first machine to win a chess match against a reigning world champion.
Enlarge
Garry Kasparov playing against Deep Blue, the first machine to win a chess match against a reigning world champion.


The modern definition of artificial intelligence (or AI) is "the study and design of intelligent agents" where an intelligent agent is a system that perceives its environment and takes actions which maximizes its chances of success.[1] John McCarthy, who coined the term in 1956,[2] defines it as "the science and engineering of making intelligent machines."[3] Other names for the field have been proposed, such as computational intelligence,[4] synthetic intelligence[4][5] or computational rationality.[6] The term artificial intelligence is also used to describe a property of machines or programs: the intelligence that the system demonstrates.

AI research uses tools and insights from many fields, including computer science, psychology, philosophy, neuroscience, cognitive science, linguistics, operations research, economics, control theory, probability, optimization and logic.[7] AI research also overlaps with tasks such as robotics, control systems, scheduling, data mining, logistics, speech recognition, facial recognition and many others.[citation needed]

History

See also: AI Winter

The field was born at a conference on the campus of Dartmouth College in the summer of 1956.[8] Those who attended would become the leaders of AI research for many decades, especially John McCarthy, Marvin Minsky, Allen Newell and Herbert Simon, who founded AI laboratories at MIT, CMU and Stanford. They and their students wrote programs that were, to most people, simply astonishing:[9] computers were solving word problems in algebra, proving logical theorems and speaking English.[10] By the middle 60s their research was heavily funded by DARPA[11] and they would make extraordinary predictions about their work:

  • 1965, H. A. Simon: "machines will be capable, within twenty years, of doing any work a man can do"[12]
  • 1967, Marvin Minsky: "Within a generation ... the problem of creating 'artificial intelligence' will substantially be solved."[13]

These predictions, and many like them, would not come true. They had failed to anticipate the difficulty of some of the problems they faced: the lack of raw computer power,[14] the intractable combinatorial explosion of their algorithms,[15] the difficulty of representing commonsense knowledge and doing commonsense reasoning,[16] the incredible difficulty of perception and motion[17] and the failings of logic.[18] In 1974, in response to the criticism of England's Sir James Lighthill and ongoing pressure from congress to fund more productive projects, DARPA cut off all undirected, exploratory research in AI. This was the first AI Winter.[19]

In the early 80s, the field was revived by the commercial success of expert systems and by 1985 the market for AI had reached more than a billion dollars.[20] Minsky and others warned the community that enthusiasm for AI had spiraled out of control and that disappointment was sure to follow.[21] Minsky was right. Beginning with the collapse of the Lisp Machine market in 1987, AI once again fell into disrepute, and a second, more lasting AI Winter began.[22]

In the 90s AI achieved its greatest successes, albeit somewhat behind the scenes. Artificial intelligence was adopted throughout the technology industry, providing the heavy lifting for logistics, data mining, medical diagnosis and many other areas.[23] The success was due to several factors: the incredible power of computers today (see Moore's law), a greater emphasis on solving specific subproblems, the creation of new ties between AI and other fields working on similar problems, and above all a new commitment by researchers to solid mathematical methods and rigorous scientific standards.[24]

Mechanisms

Expert systems were one of the earliest types of AI system. They are built around automated inference engines including forward reasoning and backwards reasoning. Based on certain conditions ("if") the system infers certain consequences ("then").

In terms of consequences, AI applications can be divided into two types: classifiers ("if shiny then diamond") and controllers ("if shiny then pick up"). Controllers do however also classify conditions before inferring actions, and therefore classification forms a central part of most AI systems.

Classifiers make use of pattern recognition for condition matching. In many cases this does not imply absolute, but rather the closest match. Techniques to achieve this divide roughly into two schools of thought: Conventional AI and Computational intelligence (CI).

Conventional AI research focuses on attempts to mimic human intelligence through symbol manipulation and symbolically structured knowledge bases. This approach limits the situations to which conventional AI can be applied. Lotfi Zadeh stated that "we are also in possession of computational tools which are far more effective in the conception and design of intelligent systems than the predicate-logic-based methods which form the core of traditional AI." These techniques, which include fuzzy logic, have become known as soft computing. These often biologically inspired methods stand in contrast to conventional AI and compensate for the shortcomings of symbolicism.[25] These two methodologies have also been labeled as neats vs. scruffies, with neats emphasizing the use of logic and formal representation of knowledge while scruffies take an application-oriented heuristic bottom-up approach.[26]

Classifiers

Classifiers are functions that can be tuned according to examples, making them very attractive for use in AI. These examples are known as observations or patterns. In supervised learning, each pattern belongs to a certain predefined class. A class can be seen as a decision that has to be made. All the observations combined with their class labels are known as a data set.

When a new observation is received, that observation is classified based on previous experience. A classifier can be trained in various ways; there are mainly statistical and machine learning approaches.

A wide range of classifiers are available, each with its strengths and weaknesses. Classifier performance depends greatly on the characteristics of the data to be classified. There is no single classifier that works best on all given problems; this is also referred to as the "no free lunch" theorem. Various empirical tests have been performed to compare classifier performance and to find the characteristics of data that determine classifier performance. Determining a suitable classifier for a given problem is however still more an art than science.

The most widely used classifiers are the neural network, support vector machine, k-nearest neighbor algorithm, Gaussian mixture model, naive Bayes classifier, and decision tree.

Conventional AI

Conventional AI mostly involves methods now classified as machine learning, characterized by formalism and statistical analysis. This is also known as symbolic AI, logical AI, neat AI and Good Old Fashioned Artificial Intelligence (GOFAI). (Also see semantics.) Methods include:

  • Expert systems: apply reasoning capabilities to reach a conclusion. An expert system can process large amounts of known information and provide conclusions based on them.
  • Case based reasoning: stores a set of problems and answers in an organized data structure called cases. A case based reasoning system upon being presented with a problem finds a case in its knowledge base that is most closely related to the new problem and presents its solutions as an output with suitable modifications.[27]
  • Bayesian networks
  • Behavior based AI: a modular method of building AI systems by hand.

Computational intelligence

Computational intelligence involves iterative development or learning (e.g., parameter tuning in connectionist systems). Learning is based on empirical data and is associated with non-symbolic AI, scruffy AI and soft computing. Subjects in computational intelligence as defined by IEEE Computational Intelligence Society mainly include:

With hybrid intelligent systems, attempts are made to combine these two groups. Expert inference rules can be generated through neural network or production rules from statistical learning such as in ACT-R or CLARION (see References below). It is thought that the human brain uses multiple techniques to both formulate and cross-check results. Thus, systems integration is seen as promising and perhaps necessary for true AI, especially the integration of symbolic and connectionist models (e.g., as advocated by Ron Sun).

AI programming languages and styles

AI research has led to many advances in programming languages including the first list processing language by Allen Newell et al., Lisp dialects, Planner, Actors, the Scientific Community Metaphor, production systems, and rule-based languages.

GOFAI TEST research is often done in programming languages such as Prolog or Lisp. Matlab and Lush (a numerical dialect of Lisp) include many specialist probabilistic libraries for Bayesian systems. AI research often emphasises rapid development and prototyping, using such interpreted languages to empower rapid command-line testing and experimentation. Real-time systems are however likely to require dedicated optimized software.

Many expert systems are organized collections of if-then such statements, called productions. These can include stochastic elements, producing intrinsic variation, or rely on variation produced in response to a dynamic environment.

Research challenges

A legged league game from RoboCup 2004 in Lisbon, Portugal.
Enlarge
A legged league game from RoboCup 2004 in Lisbon, Portugal.

The 800 million-Euro EUREKA Prometheus Project on driverless cars (1987-1995) showed that fast autonomous vehicles, notably those of Ernst Dickmanns and his team, can drive long distances (over 100 miles) in traffic, automatically recognizing and tracking other cars through computer vision, passing slower cars in the left lane. But the challenge of safe door-to-door autonomous driving in arbitrary environments will require additional research.

The DARPA Grand Challenge was a race for a $2 million prize where cars had to drive themselves over a hundred miles of challenging desert terrain without any communication with humans, using GPS, computers and a sophisticated array of sensors. In 2005, the winning vehicles completed all 132 miles of the course in just under seven hours. This was the first in a series of challenges aimed at a congressional mandate stating that by 2015 one-third of the operational ground combat vehicles of the US Armed Forces should be unmanned.[28] For November 2007, DARPA introduced the DARPA Urban Challenge. The course will involve a sixty-mile urban area course. Darpa has secured the prize money for the challenge as $2 million for first place, $1 million for second and $500,000 for third.

A popular challenge amongst AI research groups is the RoboCup and FIRA annual international robot soccer competitions. Hiroaki Kitano has formulated the International RoboCup Federation challenge: "In 2050 a team of fully autonomous humanoid robot soccer players shall win the soccer game, comply with the official rule of the FIFA, against the winner of the most recent World Cup."[29]

In the post-dot-com boom era, some search engine websites use a simple form of AI to provide answers to questions entered by the visitor. Questions such as What is the tallest building? can be entered into the search engine's input form, and a list of answers will be returned.

AI in other disciplines

AI is not only seen in computer science and engineering. It is studied and applied in various different sectors.

Philosophy


The strong AI vs. weak AI debate ("can a man-made artifact be conscious?") is still a hot topic amongst AI philosophers. This involves philosophy of mind and the mind-body problem. Most notably Roger Penrose in his book The Emperor's New Mind and John Searle with his "Chinese room" thought experiment argue that true consciousness cannot be achieved by formal logic systems, while Douglas Hofstadter in Gödel, Escher, Bach and Daniel Dennett in Consciousness Explained argue in favour of functionalism. In many strong AI supporters' opinions, artificial consciousness is considered the holy grail of artificial intelligence. Edsger Dijkstra famously opined that the debate had little importance: "The question of whether a computer can think is no more interesting than the question of whether a submarine can swim."

Epistemology, the study of knowledge, also makes contact with AI, as engineers find themselves debating similar questions to philosophers about how best to represent and use knowledge and information (e.g., semantic networks).

Neuro-psychology

Main article: Cognitive science

Techniques and technologies in AI which have been directly derived from neuroscience include neural networks, Hebbian learning and the relatively new field of Hierarchical Temporal Memory which simulates the architecture of the neocortex.

Computer Science

Notable examples include the languages LISP and Prolog, which were invented for AI research but are now used for non-AI tasks. Hacker culture first sprang from AI laboratories, in particular the MIT AI Lab, home at various times to such luminaries as John McCarthy, Marvin Minsky, Seymour Papert (who developed Logo there) and Terry Winograd (who abandoned AI after developing SHRDLU).

Business

Banks use artificial intelligence systems to organize operations, invest in stocks, and manage properties. In August 2001, robots beat humans in a simulated financial trading competition (BBC News, 2001).[30] A medical clinic can use artificial intelligence systems to organize bed schedules, make a staff rotation, and provide medical information. Many practical applications are dependent on artificial neural networks, networks that pattern their organization in mimicry of a brain's neurons, which have been found to excel in pattern recognition. Financial institutions have long used such systems to detect charges or claims outside of the norm, flagging these for human investigation. Neural networks are also being widely deployed in homeland security, speech and text recognition, medical diagnosis (such as in Concept Processing technology in EMR software), data mining, and e-mail spam filtering.

Robots have become common in many industries. They are often given jobs that are considered dangerous to humans. Robots have proven effective in jobs that are very repetitive which may lead to mistakes or accidents due to a lapse in concentration and other jobs which humans may find degrading. General Motors uses around 16,000 robots for tasks such as painting, welding, and assembly. Japan is the leader in using and producing robots in the world. In 1995, 700,000 robots were in use worldwide; over 500,000 of which were from Japan.[31]

Fiction

In science fiction AI is often portrayed as an upcoming power trying to overthrow human authority, usually in the form of futuristic humanoid robots. Best known examples include the films The Terminator and The Matrix, as well as TV shows such as the re-imagined Battlestar Galactica series.

Another common theme is the suspicion and hatred by humanity for AIs and the AIs attempt to gain human acceptance. Films include Bicentennial Man, Artificial Intelligence: A.I. and The Iron Giant. This concept is also explored in the Uncanny Valley hypothesis.

Isaac Asimov wrote stories where engineers understood these potential problems and designed their robots accordingly. Positive examples of AIs include Robby from Forbidden Planet, R2D2, C3PO and Data (Star Trek)

The inevitability of the integration of AI into human society is also argued by some science/futurist writers such as Kevin Warwick and Hans Moravec and the manga Ghost in the Shell

Toys and games

The 1990s saw some of the first attempts to mass-produce domestically aimed types of basic Artificial Intelligence for education, or leisure. This prospered greatly with the Digital Revolution, and helped introduce people, especially children, to a life of dealing with various types of AI, specifically in the form of Tamagotchis and Giga Pets, the Internet (example: basic search engine interfaces are one simple form), and the first widely released robot, Furby. A mere year later an improved type of domestic robot was released in the form of Aibo, a robotic dog with intelligent features and autonomy.

List of applications

Typical problems to which AI methods are applied
Other fields in which AI methods are implemented
Lists of researchers, projects & publications

See also

Main list: List of basic artificial intelligence topics

Notes

  1. ^ Textbooks that define AI this way include Poole, Mackworth & Goebel 1998, p. 1 and Russell & Norvig 2003, preface (who prefer the term "rational agent") and write "The whole-agent view is now widely accepted in the field" (Russell & Norvig 2003, p. 55)
  2. ^ Although there is some controversy on this point (see Crevier 1993, p. 50), McCarthy states unequivocally "I ca