Open Questions: Artificial Intelligence

[Home] [Up] [Glossary] [Topic Index] [Site Map]

Introduction

History of artificial intelligence

Applications of artificial intelligence

Criticisms of artificial intelligence

Knowledge representation

Semantic nets

Bounded rationality and heuristics

"Common sense"

Category and topos theory

Recommended references: Web sites

Recommended references: Magazine/journal articles

Recommended references: Books

Introduction

AI is controversial
AI is both a science and a field of engineering
AI has had both successes and failures
AI cannot be separated from the problem of understanding what intelligence is and how it works

Recommended references: Web sites

Site indexes

Open Directory Project: Artificial Intelligence: Categorized and annotated links. A version of this list is at Google, with entries sorted in "page rank" order. May also be found at Netscape.
Artificial Intelligence: Part of the WWW Virtual Library.
AI, Cognitive Science, and Robotics: Guide to resources such as conferences, publishers, FAQs, bibliographies, and other guides and indexes.
Psychological Science on the Net: Artificial Intelligence: Extensive annotated list of links.
Brendan Kitts Hotlist: Large list of links related to artificial intelligence and also genetic algorithms, neural nets, and neuroscience.
AI on the Web: A huge categorized list of links to AI sites, by Stuart Russell, co-author of Artificial Intelligence, a Modern Approach.
Yahoo Artificial Intelligence Links: Annotated list of links.
Galaxy: Artificial Intelligence: Categorized site directory. Entries usually include descriptive annotations.

Sites with general resources

American Association for Artificial Intelligence: "The American Association for Artificial Intelligence (AAAI) is a nonprofit scientific society devoted to advancing the scientific understanding of the mechanisms underlying thought and intelligent behavior and their embodiment in machines." AAAI publishes AI Magazine. Their Web site contains information about the organization, and a very good collection of information on AI in general, called AI Topics.
AI Magazine: Flagship publication of AAAI, which calls it the "journal of record for the AI community". It covers "significant new research and literature across the entire field of artificial intelligence." The magazine has been published since 1980, and content of all issues (in PDF form) is available online. All but the most recent content is accessible, free of charge, to everyone. Many of the articles and other features are suitable for non-specialists in AI.
Marvin Minsky Home Page: Minsky is one of the founders of AI. His site contains abstracts, bibliography, and a number of online articles.
Bibliographies on Artificial Intelligence: Directory of many useful bibliographies. Part of the Collection of Computer Science Bibliographies.
John McCarthy's Home Page: McCarthy is one of the founders of AI and created the LISP programming language. His site contains pointers to many of his papers, and a very interesting and useful FAQ entitled What is Artificial Intelligence.
IEEE Intelligent Systems: Professional journal that focuses on applications of artificial intelligence. A small amount of the content is free, but most requires payment of a substantial fee.
MIT Artificial Intelligence Laboratory: The laboratory's "goal is to understand the nature of intelligence and to engineer systems that exhibit intelligence." Major emphasis is on vision, robotics, and language. Web site contains information on lab research activities and publications. The laboratory recently merged to form the MIT Computer Science and Artificial Intelligence Laboratory.

General surveys, overviews, tutorials

Artificial intelligence: Article from Wikipedia. See also Expert system, Knowledge representation, Frame problem, Machine learning, Natural language processing.
AI Topics: "A dynamic library of introductory information about artificial intelligence." The site is intended to provide information on AI that is understandable by students of high school level and above. It is provided by AAAI.
6.825 Techniques in Artificial Intelligence, Fall 2002: A graduate level course of MIT OpenCourseWare, developed by Leslie Kaelbling. Lecture notes are provided in PDF format. "Topics covered include: representation and inference in first-order logic, modern deterministic and decision-theoretic planning techniques, basic supervised learning methods, and Bayesian network inference and learning."
Intelligent Sytems and Their Societies: A free e-book by Walter Fritz dealing with intelligent systems, artificial intelligence, and robotics. Main page has some useful links.
What is Artificial Intelligence: Good list of frequently asked questions, by John McCarthy. It deals with basic questions, the branches of AI, and applications of AI.
Amplified Intelligence: Machines as Brain Boosters: August 2004 article from Space.com. Presents an interview with Ken Ford, discussing "amplified intelligence", a vision of how humans and machines interact.
Artificial Intelligence for the New Millennium: June 20, 2001 article from the New York Times, providing a very non-technical overview of the current state of applied artificial intelligence.
Falling Prey To Machines?: A discussion with AI expert John Holland sketches the current (2003) state of artificial intelligence.

Natural language

Natural Language Understanding: Online text of 1995 book by James Allen. A good survey of the field.
OpenCyc: "OpenCyc is the open source version of the Cyc technology, the world's largest and most complete general knowledge base and commonsense reasoning engine." The site provides access to open source software, documentation, and the knowledge base. It is provided for free use by Cycorp.
Cycorp, Inc.: "Cycorp was founded in 1994 to research, develop, and commercialize Artificial Intelligence. Cycorp's vision is to create the world's first true artificial intelligence, having both common sense and the ability to reason with it."
WordNet: A lexical database for English. "WordNet is an online lexical reference system whose design is inspired by current psycholinguistic theories of human lexical memory." WordNet is a project of George A. Miller and the Princeton Cognitive Science Laboratory.

The frame problem

Some Philosophical Problems from the Standpoint of Artificial Intelligence: A classic paper written in 1969 by John McCarthy and Patrick J. Hayes. The concept of a "frame problem" was introduced in this paper.
Frame Problem: Short 1998 article by Eric Lormand, with emphasis on the philosophical aspects.

Recommended references: Magazine/journal articles

Artificial Intelligence: Hype or Reality
Adrian A. Hopgood
Computer, May 2003, pp. 24-28: The serious study of artificial intelligence is conservatively assumed to have begun over 50 years ago, and can be defined as "the science of mimicking human mental faculties in a computer." Intelligent behavior can be described on a scale from simple reaction to sophisticated expertise. But ironically, behavior in the middle of the scale, such as common sense reasoning, has proven the most difficult to emulate.
21st-Century AI: Proud, Not Smug
Tim Menzies
Intelligent Systems, May/June 2003, pp. 18-24: Artificial intelligence (as an engineering discipline) is maturing. Examples of successful AI applications include data mining and natural language processing. But the "easy" work may be over. Some of the fertile areas of research for the near future include case-based reasoning, large knowledge bases, agent architectures, improved Bayes nets, and genetic algorithms.
AI and Agents: State of the Art
Eduardo Alonso
AI Magazine, Fall 2002, pp. 25-29: Many observers feel that AI has failed to deliver on its original goal of building intelligent systems of general competence. In order to meet this goal, AI systems need to be autonomous and flexible, capable of operating in unpredictable, dynamic, and usually social domains. These requirements seem to call for using "agents" to implement AI systems.
Lord of the Robots
David Talbot
Technology Review, April 2002, pp. 78-82: An interview with Rodney Brooks, director of MIT's Artificial Intelligence Lab, provides an overview of the current state of AI research, especially in connection with robotics.
A. I. Reboots
Michael Hiltzik
Technology Review, March 2002, pp. 46-55: Research and development in artificial intelligence is still an active area. But, at least for the near term, the goals have changed from emulating human capabilities and thought processes to practical applications -- such as "smart" databases and application software -- that assist rather than replace humans.
Creativity at the Metalevel
Bruce G. Buchanan
AI Magazine, Fall 2001, pp. 13-28: Human creativity has been studied extensively by psychologists and philosophers, and a number of similar models of the creative process have been proposed. All seem to involve an iterative process of generating plausible solutions and testing them. All these have certain commonalities, such as methods of solution generation and collections of verification criteria. Higher level creativity may arise at the "metalevel" by systematically varying such methods and criteria.
[Article in PDF format]
Seeing Clearly and Moving Forward
Robert Laddaga; Mark L. Swinson; Paul Robertson
Intelligent Systems, November/December 2000, pp. 46-49: Work will continue to be important in areas such as data mining, learning, knowledge representation, planning and scheduling, natural language processing, expert systems, and deductive and inductive reasoning. But as "ubiquitous computing" becomes prevalent, computers will need the ability perceive and interact with the world of humans and their artifacts. This makes the study of vision, speech, and language understanding especially important for incorporation in embedded computers and robotic systems.
What Does the Future Hold?
Howard Shrobe
AI Magazine, Winter 2000, pp. 41-57: In the past, AI system implementations tended to consist of black boxes that were presented with some sort of problem description and produced some sort of answer. But human-like performance seems to require an ability to sense and interact with a real-world environment. Applications involving embedded systems and "ubiquitous computing" lead in the same direction. The implication is that AI systems must meet three new requirements: autonomy, robustness, and the ability to make sense out of multiple kinds of input.
[Article in PDF format]
AI's Greatest Trends and Controversies
Marti A. Hearst; Haym Hirsch
Intelligent Systems, January/February 2000, pp. 8-17: 24 of the best-known authorities on AI and related fields offer their own personal perspectives on the first half century of AI research.
The Importance of Importance
David Waltz
AI Magazine, Fall 1999, pp. 18-35: AI began by working with what were assumed to be basic, abstract human skills, such as logic, reasoning, and knowledge representation. But this approach neglected another innate human skill: being able to quickly perceive what is "important" in a problem. This is a perceptual ability, which is difficult to formalize. The further advance of AI will depend on better understanding of this and other similar faculties such as action, language, and creativity. To do this, AI will need to assimilate the insights of neighboring disciplines like neuroscience, psychology, and linguistics.
[Article in PDF format]
AI Growing Up
James F. Allen
AI Magazine, Winter 1998, pp. 13-23: The definition of "AI" has not been especially clear, so a more precise definition is proposed: "AI is the science of making machines do tasks that humans can do or try to do." Past approaches to AI have been misguided in various ways, such as working on problems that were too abstract or focusing on just one kind of technique. More progress might be made by addressing real but circumscribed problems using a suite of sophisticated techniques.
[Article in PDF format]
What Are Intelligence? And Why?
Randall Davis
AI Magazine, Spring 1998, pp. 91-110: The title of this article makes the point that "intelligence" is many things, both conceptually and in implementation. Intelligence can be considered from several points of view, such as the logical and the psychological. As a product of evolution, different components of intelligence have come about for a variety of purposes. Behavior that can be called "intelligent" is not limited to humans. AI can be regarded as the study of the design space of possible forms of intelligence. As a science, AI may be more like biology, which tries to understand diverse phenomena, than like mathematics, which tries to extrapolate from a limited set of general principles. One kind of reasoning that needs to be studied more closely is visual or perceptual rather than logical in nature.
[Article in PDF format]
Chess is Too Easy
Selmer Bringsjord
Technology Review, March/April 1998, pp. 23-28: One of the chief features that seems to be missing from efforts to provide computers with "intelligence" is a capability for original or creative behavior. The author has developed the software of a "storytelling machine" that can write original but very short stories. It is concluded, however, that only a human can have sufficient understanding of human thoughts and emotions to write a story comparable in quality to the work of the best human authors.

Recommended references: Books

Ray Kurzweil -- The Age of Spiritual Machines: When Computers Exceed Human Intelligence
Penguin Books, 1999: Kurzweil is a successful inventor (in areas like speech recognition, optical character recognition, music synthesis) with a good track record in prediction. His predictions in this book are, still, rather daring. After a brief look at the history of computer evolution and the question of whether computers can ever be "conscious", he launches into some audacious predictions for the development of computers and robots in the next three decades and the remainder of the 21st century.
[Book home page]
John L. Casti -- The Cambridge Quintet: A Work of Scientific Speculation
Perseus Books, 1998: Casti examines the philosophical and practical prospects of artificial intelligence using the device of an imagined dinner party in 1949 involving C. P. Snow, Erwin Schrödinger, J. B. S. Haldane, Alan Turing, and Ludwig Wittgenstein. The central conflict is between Turing and Wittgenstein over the question of whether a machine can "think".
David G. Stork, ed. -- HAL's Legacy: 2001's Computer as Dream and Reality
MIT Press, 1997: Using the movie 2001's computer HAL as a point of reference, this books presents 16 essays by experts in artificial intelligence that discuss both the technology and human significance of computer technology. It is now clear that HAL was an overly optimistic projection of what would be possible in the given time frame. The important question is when, or whether, a computer like HAL will be feasible.
Margaret A. Boden, ed. -- The Philosophy of Artificial Intelligence
Oxford University Press, 1990: An excellent collection of papers by some of the seminal thinkers and critics concerned with artificial intelligence, such as Warren McCulloch, Alan Turing, John Searle, Allen Newell, Herbert Simon, Daniel Dennett, Patrick Hayes, Hubert Dreyfus, and Paul Churchland. Topics include the nature of intelligence, the frame problem, and connectionism.
George Johnson -- Mahinery of the Mind: Inside the New Science of Artificial Intelligence
Microsoft Press, 1986: Johnson's book is a time capsule from the "heroic age" of artificial intelligence, just before the "AI winter". It provides a history, for general readers and without technical details, of the people who created the discipline known as AI, the systems they were famous for, and the main ideas of the field.

Home