Artificial intelligence, also known as machine intelligence, is defined as intelligence exhibited by anything manufactured (i.e. ) by humans or other sentient beings or systems (should such things ever exist on Earth or elsewhere). It is usually hypothetically applied to general-purpose computers. The term is also used to refer to the field of scientific investigation into the plausibility of and approaches to creating such systems.
The second is much harder, raising questions of consciousness and self, mind (including the unconscious mind) and the question of what components are involved in the only type of intelligence it is universally agreed we have available to study: that of human beings. Intelligent behavior in humans is complex and difficult to study or understand. Study of animals and artificial systems that are not just models of what exists already are also considered widely pertinent.
Several distinct types of artificial intelligence have been elucidated below. Also, the subject divisions, history, proponents and opponents and applications of research in the subject are described. Finally, references to fictional and non-fictional descriptions of AI are provided.
One popular and early definition of artificial intelligence research, put forth by John McCarthy at the Dartmouth Conference in 1955 is "making a machine behave in ways that would be called intelligent if a human were so behaving." However this definition seems to ignore the possibility of strong AI (see below). Another definition of artificial intelligence is intelligence arising from an artificial device. Most definitions could be categorized as concerning either systems that think like humans, systems that act like humans, systems that think rationally or systems that act rationally.
To date, much of the work in this field has been done with computer simulations of intelligence based on predefined sets of rules. Very little progress has been made in strong AI. Depending on how one defines one's goals, a moderate amount of progress has been made in weak AI.
Several philosophers, notably John Searle and Hubert Dreyfus, have argued on philosophical grounds against the feasibility of building human-like consciousness or intelligence in a disembodied machine. Searle is most known for his Chinese room argument, which claims to demonstrate that even a machine that passed the Turing test would not necessarily be conscious in the human sense. Dreyfus, in his book What Computers Still Can't Do: A Critique of Artificial Reason, has argued that consciousness cannot be captured by rule- or logic-based systems or by systems that are not attached to a physical body, but leaves open the possibility that a robotic system using neural networks or similar mechanisms might achieve artificial intelligence.
Other philosophers hold opposing views. Many see no problem with Weak AI but there is much support for Strong AI too. Daniel C. Dennett argues in Consciousness Explained that if there is no magic spark or soul, then Man is just a machine, and he asks why the Man-machine should have a privileged position over all other possible machines when it comes to intelligence.
Some philosophers hold that if Weak AI is accepted as possible then so must Strong AI. The Weak AI position, that intelligence might be apparent but would not be real, is debunked in many ways, but one accessible example can be found in Simon Blackburn's introduction to philosophy, Think. Blackburn points out that you might appear intelligent but there is no way of telling if that intelligence is real (whatever that means in this context): We have to take it on trust or faith.
Supporters of Strong AI claim that the anti-AI argument boils down in the end to some combination of
arrogance as a privileged position is claimed, a magic spark is introduced (by God, for instance)
defining intelligence as that of which machines are incapable.
An argument supporting Strong AI which those who deny its possibility must necessarily attack:
Approaches to artificial intelligence that do not focus on linguistic intelligence include robotics and collective intelligence approaches, which focus on active manipulation of an environment, or consensus decision making, and draw from biology and political science when seeking models of how "intelligent" behavior is organized.
Artificial intelligence theory also draws from animal studies, in particular with insects, which are easier to emulate as robots (see artificial life), as well as animals with more complex cognition, including apes, who resemble humans in many ways but have less developed capacities for planning and cognition. AI researchers argue that animals, which are simpler than humans, ought to be considerably easier to mimic. But satisfactory computational models for animal intelligence are not available.
Seminal papers advancing the concept of machine intelligence include A Logical Calculus of the Ideas Immanent in Nervous Activity (1943), by Warren McCulloch and Walter Pitts, and On Computing Machinery and Intelligence (1950), by Alan Turing, and Man-Computer Symbiosis by J.C.R. Licklider. See cybernetics and Turing test for further discussion.
There were also early papers which denied the possibility of machine intelligence on logical or philosophical grounds such as Minds, Machines and Gödel; (1961) by John Lucas[1].
With the development of practical techniques based on AI research, advocates of AI have argued that opponents of AI have repeatedly changed their position on tasks such as computer chess or speech recognition that were previously regarded as "intelligent" in order to deny the accomplishments of AI. They point out that this moving of the goalposts effectively defines "intelligence" as "whatever humans can do that machines cannot".
John von Neumann (quoted by E.T. Jaynes) anticipated this in 1948 by saying, in response to a comment at a lecture that it was impossible for a machine to think: "You insist that there is something a machine cannot do. If you will tell me precisely what it is that a machine cannot do, then I can always make a machine which will do just that!". Von Neumann was presumably alluding to the Church-Turing thesis which states that any effective procedure can be simulated by a (generalized) computer.
In 1969 McCarthy and Hayes started the discussion about the frame problem with their essay, "Some Philosophical Problems from the Standpoint of Artificial Intelligence".
Historically, there are two broad styles of AI research - the "neats" and "scruffies". "Neat", classical or symbolic AI research, in general, involves symbolic manipulation of abstract concepts, and is the methodology used in most expert systems. Parallel to this are the "scruffy", or "connectionist", approaches, of which neural networks are the best-known example, which try to "evolve" intelligence through building systems and then improving them through some automatic process rather than systematically designing something to complete the task. Both approaches appeared very early in AI history. Throughout the 1960s and 1970s scruffy approaches were pushed to the background, but interest was regained in the 1980s when the limitations of the "neat" approaches of the time became clearer. However, it has become clear that contemporary methods using both broad approaches have severe limitations.
Artificial intelligence research was very heavily funded in the 1980s by the Defense Advanced Research Projects Agency in the United States and by the Fifth Generation Computer project in Japan. The failure of the work funded at the time to produce immediate results, despite the grandiose promises of some AI practitioners, led to correspondingly large cutbacks in funding by government agencies in the late 1980s, leading to a general downturn in activity in the field known as AI winter. Over the following decade, many AI researchers moved into related areas with more modest goals such as machine learning, robotics, and computer vision, though research in pure AI continued at reduced levels.
Computer algebra systems, such as Mathematica and Macsyma, are commonplace.
Machine vision systems are used in many industrial applications.
The vision of artificial intelligence replacing human professional judgment has arisen many times in the history of the field, in science fiction and today in some specialized areas where "expert systems" are used to augment or to replace professional judgment in some areas of engineering and of medicine.
Over time, debates have tended to focus less and less on "possibility" and more on "desirability", as emphasized in the "Cosmist" (versus "Terran") debates initiated by Hugo de Garis and Kevin Warwick. A Cosmist, according to de Garis, is actually seeking to build more intelligent successors to the human species. The emergence of this debate suggests that desirability questions may also have influenced some of the early thinkers "against".
Some issues that bring up interesting ethical questions are:
ELIZA, a program which pretends to be a psychoterapist, developed circa 1970.
PAM (Plan Applier Mechanism) - a story understanding system developed by John Wilensky in 1978.
SAM (Script applier mechanism) - a story understanding system, dveloped in 1975.
SHRDLU - an early natural language understanding computer program developed in 1968-1970.
Creatures, a computer game with breeding, evolving creatures coded from the genetic level upwards using a sophisticated biochemistry and neural network brains.
EURISKO - a language for solving problems which consists of heuristics, including heuristics describing how to use and change its heuristics. Developed in 1978 by Douglas Lenat.
To some computer scientists, the phrase artificial intelligence has acquired somewhat of a bad name due to the large discrepancy between what has been achieved so far in the field and some more usual notions of intelligence. This problem has been aggrevated by various irresponsible popular science writers and media personalities such as Source | Copyright
