Symbolic Artificial Intelligence

Symbolic artificial intelligence, a collection of methods based on human-readable symbolic representations, logic, and search, was the dominant paradigm of AI research from the mid-1950s until the mid-1990s, according to Wikipedia. The approach produced tools including logic programming, production rules, semantic nets, and frames, and applications such as knowledge-based systems, expert systems, automated theorem provers, ontologies, the semantic web, and automated planning and scheduling. Two boom cycles — the first following early successes like the Logic Theorist and Arthur Samuel's checkers program, and a second from 1969 to 1986 driven by expert systems such as XCON at DEC — were each followed by disappointment and funding cuts, including the first AI Winter.

Symbolic artificial intelligence | Part of a series https://en.wikipedia.org/wiki/Category:Artificial intelligence on | | Artificial intelligence AI https://en.wikipedia.org/wiki/Artificial intelligence | |---| | Glossary https://en.wikipedia.org/wiki/Glossary of artificial intelligence | In artificial intelligence https://en.wikipedia.org/wiki/Artificial intelligence AI , symbolic artificial intelligence also known as classical artificial intelligence or logic-based artificial intelligence \ 1\ cite note-:1-1 <sup \ 2\ cite note-2 </sup is a collection of methods based on high-level symbolic https://en.wikipedia.org/wiki/Physical symbol systems hypothesis human-readable https://en.wikipedia.org/wiki/Human-readable representations of problems, logic https://en.wikipedia.org/wiki/Formal logic , and search https://en.wikipedia.org/wiki/Search algorithm .<sup \ 1\ cite note-:1-1 </sup Symbolic AI used tools such as logic programming https://en.wikipedia.org/wiki/Logic programming , production rules <https://en.wikipedia.org/wiki/Production computer science , semantic nets https://en.wikipedia.org/wiki/Semantic nets and frames <https://en.wikipedia.org/wiki/Frame artificial intelligence . It developed applications such as knowledge-based systems https://en.wikipedia.org/wiki/Knowledge-based systems in particular, expert systems https://en.wikipedia.org/wiki/Expert systems , symbolic mathematics https://en.wikipedia.org/wiki/Symbolic mathematics , automated theorem provers https://en.wikipedia.org/wiki/Automated theorem provers , ontologies https://en.wikipedia.org/wiki/Ontologies , the semantic web https://en.wikipedia.org/wiki/Semantic web , and automated planning and scheduling https://en.wikipedia.org/wiki/Automated planning and scheduling systems. The symbolic AI paradigm led to important ideas in search https://en.wikipedia.org/wiki/Artificial intelligence Search and optimisation , symbolic programming https://en.wikipedia.org/wiki/Symbolic programming languages, agents https://en.wikipedia.org/wiki/Intelligent agent , multi-agent systems https://en.wikipedia.org/wiki/Multi-agent systems , the semantic web, and the strengths and limitations of formal knowledge and reasoning systems https://en.wikipedia.org/wiki/Automated reasoning . Symbolic AI was the dominant paradigm https://en.wikipedia.org/wiki/Paradigm of AI research from the mid-1950s until the mid-1990s.<sup \ 3\ cite note-FOOTNOTEKolata1982-3 </sup Researchers in the 1960s and the 1970s were convinced that symbolic approaches would eventually succeed in creating a machine with artificial general intelligence https://en.wikipedia.org/wiki/Artificial general intelligence and considered this the ultimate goal of their field.<sup \ 4\ cite note-4 </sup An early boom, with early successes such as the Logic Theorist https://en.wikipedia.org/wiki/Logic Theorist and Samuel <https://en.wikipedia.org/wiki/Arthur Samuel computer scientist 's Checkers Playing Program <https://en.wikipedia.org/wiki/Arthur Samuel computer scientist , led to unrealistic expectations and promises and was followed by the first AI Winter https://en.wikipedia.org/wiki/AI winter as funding dried up. \ 5\ cite note-FOOTNOTEKautz2022107–109-5 <sup \ 6\ cite note-FOOTNOTERussellNorvig202119-6 </sup A second boom 1969–1986 occurred with the rise of expert systems, their promise of capturing corporate expertise, and an enthusiastic corporate embrace. \ 7\ cite note-FOOTNOTERussellNorvig202122–23-7 <sup \ 8\ cite note-FOOTNOTEKautz2022109–110-8 </sup That boom, and some early successes, e.g., with XCON https://en.wikipedia.org/wiki/XCON at DEC https://en.wikipedia.org/wiki/Digital Equipment Corporation , was followed again by later disappointment.<sup \ 8\ cite note-FOOTNOTEKautz2022109–110-8 </sup Problems with difficulties in knowledge acquisition, maintaining large knowledge bases, and brittleness in handling out-of-domain problems arose. Another, second, AI Winter 1988–2011 followed.<sup \ 9\ cite note-FOOTNOTEKautz2022110-9 </sup Subsequently, AI researchers focused on addressing underlying problems in handling uncertainty and in knowledge acquisition.<sup \ 10\ cite note-FOOTNOTEKautz2022110–111-10 </sup Uncertainty was addressed with formal methods such as hidden Markov models https://en.wikipedia.org/wiki/Hidden Markov model , Bayesian reasoning https://en.wikipedia.org/wiki/Bayesian reasoning , and statistical relational learning https://en.wikipedia.org/wiki/Statistical relational learning . \ 11\ cite note-FOOTNOTERussellNorvig202125-11 <sup \ 12\ cite note-FOOTNOTEKautz2022111-12 </sup Symbolic machine learning addressed the knowledge acquisition problem with contributions including Version Space https://en.wikipedia.org/wiki/Version space learning , Valiant https://en.wikipedia.org/wiki/Leslie Valiant 's PAC learning https://en.wikipedia.org/wiki/Probably approximately correct learning , Quinlan https://en.wikipedia.org/wiki/Ross Quinlan 's ID3 https://en.wikipedia.org/wiki/ID3 algorithm decision-tree https://en.wikipedia.org/wiki/Decision-tree learning, case-based learning https://en.wikipedia.org/wiki/Case-based reasoning , and inductive logic programming https://en.wikipedia.org/wiki/Inductive logic programming to learn relations. \ 13\ cite note-FOOTNOTEKautz2020110–111-13 Neural networks https://en.wikipedia.org/wiki/Artificial neural network , a subsymbolic<sup clarification needed https://en.wikipedia.org/wiki/Wikipedia:Please clarify </sup approach, had been pursued from early days and reemerged strongly in 2012. Early examples are Rosenblatt https://en.wikipedia.org/wiki/Frank Rosenblatt 's perceptron https://en.wikipedia.org/wiki/Perceptron learning work, the backpropagation https://en.wikipedia.org/wiki/Backpropagation work of Rumelhart, Hinton and Williams,<sup \ 14\ cite note-14 </sup and work in convolutional neural networks https://en.wikipedia.org/wiki/Convolutional neural network by LeCun et al. in 1989.<sup \ 15\ cite note-15 </sup However, neural networks were not viewed as successful until about 2012: "Until Big Data became commonplace, the general consensus in the Al community was that the so-called neural-network approach was hopeless. Systems just didn't work that well, compared to other methods. ... A revolution came in 2012, when a number of people, including a team of researchers working with Hinton, worked out a way to use the power of GPUs https://en.wikipedia.org/wiki/GPUs to enormously increase the power of neural networks."<sup \ 16\ cite note-FOOTNOTEMarcusDavis2019-16 </sup Over the next several years, deep learning https://en.wikipedia.org/wiki/Deep learning had spectacular success in handling vision, speech recognition https://en.wikipedia.org/wiki/Speech recognition , speech synthesis, image generation, and machine translation, though symbolic approaches continue to be useful in a few domains such as computer algebra systems https://en.wikipedia.org/wiki/Computer algebra system and proof assistants https://en.wikipedia.org/wiki/Proof assistant . History edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=1 A short history of symbolic AI to the present day follows below. Time periods and titles are drawn from Henry Kautz's 2020 AAAI Robert S. Engelmore Memorial Lecture<sup \ 17\ cite note-FOOTNOTEKautz2020-17 </sup and the longer Wikipedia article on the History of AI https://en.wikipedia.org/wiki/History of AI , with dates and titles differing slightly for increased clarity. The first AI summer: irrational exuberance, 1948–1966 edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=2 Success at early attempts in AI occurred in three main areas: artificial neural networks, knowledge representation, and heuristic search, contributing to high expectations. This section summarizes Kautz's reprise of early AI history. Approaches inspired by human or animal cognition or behavior edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=3 Cybernetic approaches attempted to replicate the feedback loops between animals and their environments. A robotic turtle, with sensors, motors for driving and steering, and seven vacuum tubes for control, based on a preprogrammed neural net, was built as early as 1948. This work can be seen as an early precursor to later work in neural networks, reinforcement learning, and situated robotics. \ 18\ cite note-FOOTNOTEKautz2022106-18 An important early symbolic AI program was the Logic theorist https://en.wikipedia.org/wiki/Logic theorist , written by Allen Newell https://en.wikipedia.org/wiki/Allen Newell , Herbert Simon https://en.wikipedia.org/wiki/Herbert A. Simon and Cliff Shaw https://en.wikipedia.org/wiki/Cliff Shaw in 1955–56, as it was able to prove 38 elementary theorems from Whitehead and Russell's Principia Mathematica https://en.wikipedia.org/wiki/Principia Mathematica . Newell, Simon, and Shaw later generalized this work to create a domain-independent problem solver, GPS https://en.wikipedia.org/wiki/General Problem Solver General Problem Solver . GPS solved problems represented with formal operators via state-space search using means-ends analysis https://en.wikipedia.org/wiki/Means-ends analysis . \ 19\ cite note-FOOTNOTENewellSimon1972-19 During the 1960s, symbolic approaches achieved great success at simulating intelligent behavior in structured environments such as game-playing, symbolic mathematics, and theorem-proving. AI research was concentrated in four institutions in the 1960s: Carnegie Mellon University https://en.wikipedia.org/wiki/Carnegie Mellon University , Stanford https://en.wikipedia.org/wiki/Stanford , MIT https://en.wikipedia.org/wiki/MIT and later University of Edinburgh https://en.wikipedia.org/wiki/University of Edinburgh . Each one developed its own style of research. Earlier approaches based on cybernetics https://en.wikipedia.org/wiki/Cybernetics or artificial neural networks https://en.wikipedia.org/wiki/Artificial neural network were abandoned or pushed into the background. Herbert Simon https://en.wikipedia.org/wiki/Herbert A. Simon and Allen Newell https://en.wikipedia.org/wiki/Allen Newell studied human problem-solving skills and attempted to formalize them, and their work laid the foundations of the field of artificial intelligence, as well as cognitive science https://en.wikipedia.org/wiki/Cognitive science , operations research https://en.wikipedia.org/wiki/Operations research and management science https://en.wikipedia.org/wiki/Management science . Their research team used the results of psychological https://en.wikipedia.org/wiki/Psychology experiments to develop programs that simulated the techniques that people used to solve problems. \ 20\ < cite note-FOOTNOTEMcCorduck2004139–179, 245–250, 322–323 EPAM -20 <sup \ 21\ cite note-FOOTNOTECrevier1993145–149-21 </sup This tradition, centered at Carnegie Mellon University would eventually culminate in the development of the Soar <https://en.wikipedia.org/wiki/Soar cognitive architecture architecture in the middle 1980s. \ 22\ cite note-FOOTNOTEMcCorduck2004450–451-22 \ 23\ cite note-FOOTNOTECrevier1993258–263-23 Heuristic search edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=4 In addition to the highly specialized domain-specific kinds of knowledge that we will see later used in expert systems, early symbolic AI researchers discovered another more general application of knowledge. These were called heuristics, rules of thumb that guide a search in promising directions: "How can non-enumerative search be practical when the underlying problem is exponentially hard? The approach advocated by Simon and Newell is to employ heuristics <https://en.wikipedia.org/wiki/Heuristic computer science : fast algorithms that may fail on some inputs or output suboptimal solutions."<sup \ 24\ cite note-FOOTNOTEKautz2022108-24 </sup Another important advance was to find a way to apply these heuristics that guarantees a solution will be found, if there is one, not withstanding the occasional fallibility of heuristics: "The A algorithm https://en.wikipedia.org/wiki/A search algorithm provided a general frame for complete and optimal heuristically guided search. A is used as a subroutine within practically every AI algorithm today but is still no magic bullet; its guarantee of completeness is bought at the cost of worst-case exponential time. \ 24\ cite note-FOOTNOTEKautz2022108-24 Early work on knowledge representation and reasoning edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=5 Early work covered both applications of formal reasoning emphasizing first-order logic https://en.wikipedia.org/wiki/First-order logic , along with attempts to handle common-sense reasoning https://en.wikipedia.org/wiki/Commonsense reasoning in a less formal manner. Modeling formal reasoning with logic: the "neats" edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=6 Unlike Simon and Newell, John McCarthy <https://en.wikipedia.org/wiki/John McCarthy computer scientist felt that machines did not need to simulate the exact mechanisms of human thought, but could instead try to find the essence of abstract reasoning and problem-solving with logic,<sup \ 25\ < cite note-FOOTNOTERussellNorvig2021p. 9 logicist AI , p. 19 McCarthy's work -25 </sup regardless of whether people used the same algorithms.<sup \ a\ cite note-29 </sup His laboratory at Stanford https://en.wikipedia.org/wiki/Stanford University SAIL https://en.wikipedia.org/wiki/Stanford Artificial Intelligence Laboratory focused on using formal logic https://en.wikipedia.org/wiki/Logic to solve a wide variety of problems, including knowledge representation https://en.wikipedia.org/wiki/Knowledge representation , planning and learning https://en.wikipedia.org/wiki/Machine learning .<sup \ 29\ cite note-FOOTNOTEMcCorduck2004251–259-30 </sup Logic was also the focus of the work at the University of Edinburgh https://en.wikipedia.org/wiki/University of Edinburgh and elsewhere in Europe which led to the development of the programming language Prolog https://en.wikipedia.org/wiki/Prolog and the science of logic programming. \ 30\ cite note-FOOTNOTECrevier1993193–196-31 \ 31\ cite note-FOOTNOTEHowe1994-32 Modeling implicit common-sense knowledge with frames and scripts: the "scruffies" edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=7 Researchers at MIT https://en.wikipedia.org/wiki/MIT such as Marvin Minsky https://en.wikipedia.org/wiki/Marvin Minsky and Seymour Papert https://en.wikipedia.org/wiki/Seymour Papert \ 32\ cite note-FOOTNOTEMcCorduck2004259–305-33 \ 33\ cite note-FOOTNOTECrevier199383–102, 163–176-34 <sup \ 6\ cite note-FOOTNOTERussellNorvig202119-6 </sup found that solving difficult problems in vision https://en.wikipedia.org/wiki/Computer vision and natural language processing https://en.wikipedia.org/wiki/Natural language processing required ad hoc solutions—they argued that no simple and general principle like logic https://en.wikipedia.org/wiki/Logic would capture all the aspects of intelligent behavior. Roger Schank https://en.wikipedia.org/wiki/Roger Schank described their "anti-logic" approaches as " scruffy https://en.wikipedia.org/wiki/Neats vs. scruffies " as opposed to the " neat https://en.wikipedia.org/wiki/Neats vs. scruffies " paradigms at CMU https://en.wikipedia.org/wiki/Carnegie Mellon University and Stanford . \ 34\ cite note-FOOTNOTEMcCorduck2004421–424, 486–489-35 \ 35\ cite note-FOOTNOTECrevier1993168-36 Commonsense knowledge bases https://en.wikipedia.org/wiki/Commonsense knowledge bases such as Doug Lenat https://en.wikipedia.org/wiki/Doug Lenat 's Cyc https://en.wikipedia.org/wiki/Cyc are an example of "scruffy" AI, since they must be built by hand, one complicated concept at a time. \ 36\ cite note-FOOTNOTEMcCorduck2004489-37 \ 37\ cite note-FOOTNOTECrevier1993239–243-38 \ 38\ cite note-FOOTNOTERussellNorvig2021316, 340-39 The first AI winter: crushed dreams, 1967–1977 edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=8 The first AI winter was a shock: During the first AI summer, many people thought that machine intelligence could be achieved in just a few years. The Defense Advance Research Projects Agency DARPA launched programs to support AI research to use AI to solve problems of national security; in particular, to automate the translation of Russian to English for intelligence operations and to create autonomous tanks for the battlefield. Researchers had begun to realize that achieving AI was going to be much harder than was supposed a decade earlier, but a combination of hubris and disingenuousness led many university and think-tank researchers to accept funding with promises of deliverables that they should have known they could not fulfill. By the mid-1960s neither useful natural language translation systems nor autonomous tanks had been created, and a dramatic backlash set in. New DARPA leadership canceled existing AI funding programs. ... Outside of the United States, the most fertile ground for AI research was the United Kingdom. The AI winter in the United Kingdom was spurred on not so much by disappointed military leaders as by rival academics who viewed AI researchers as charlatans and a drain on research funding. A professor of applied mathematics, Sir James Lighthill, was commissioned by Parliament to evaluate the state of AI research in the nation https://en.wikipedia.org/wiki/Lighthill report . The report stated that all of the problems being worked on in AI would be better handled by researchers from other disciplines—such as applied mathematics. The report also claimed that AI successes on toy problems could never scale to real-world applications due to combinatorial explosion. \ 39\ cite note-FOOTNOTEKautz2022109-40 The second AI summer: knowledge is power, 1978–1987 edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=9 Knowledge-based systems edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=10 As limitations with weak, domain-independent methods became more and more apparent,<sup \ 40\ cite note-FOOTNOTERussellNorvig202122-41 </sup researchers from all three traditions began to build knowledge https://en.wikipedia.org/wiki/Knowledge representation into AI applications. \ 41\ cite note-FOOTNOTEMcCorduck2004266–276, 298–300, 314, 421-42 <sup \ 7\ cite note-FOOTNOTERussellNorvig202122–23-7 </sup The knowledge revolution was driven by the realization that knowledge underlies high-performance, domain-specific AI applications. Edward Feigenbaum https://en.wikipedia.org/wiki/Edward Feigenbaum said: - "In the knowledge lies the power."<sup \ 42\ cite note-Feignebaum Interview-43 </sup to describe that high performance in a specific domain requires both general and highly domain-specific knowledge. Ed Feigenbaum and Doug Lenat called this The Knowledge Principle: 1 The Knowledge Principle: if a program is to perform a complex task well, it must know a great deal about the world in which it operates. 2 A plausible extension of that principle, called the Breadth Hypothesis: there are two additional abilities necessary for intelligent behavior in unexpected situations: falling back on increasingly general knowledge, and analogizing to specific but far-flung knowledge. \ 43\ cite note-Knowledge Principle-44 Success with expert systems edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=11 This "knowledge revolution" led to the development and deployment of expert systems https://en.wikipedia.org/wiki/Expert system introduced by Edward Feigenbaum https://en.wikipedia.org/wiki/Edward Feigenbaum , the first commercially successful form of AI software. \ 44\ cite note-FOOTNOTERussellNorvig202122–24-45 \ 45\ cite note-FOOTNOTEMcCorduck2004327–335, 434–435-46 \ 46\ cite note-FOOTNOTECrevier1993145–62, 197–203-47 Key expert systems were: - DENDRAL https://en.wikipedia.org/wiki/DENDRAL , which found the structure of organic molecules from their chemical formula and mass spectrometer readings. - MYCIN https://en.wikipedia.org/wiki/MYCIN , which diagnosed bacteremia – and suggested further lab tests, when necessary – by interpreting lab results, patient history, and doctor observations. "With about 450 rules, MYCIN was able to perform as well as some experts, and considerably better than junior doctors."<sup \ 47\ cite note-FOOTNOTERussellNorvig202123-48 </sup - INTERNIST https://en.wikipedia.org/wiki/Internist-I and CADUCEUS <https://en.wikipedia.org/wiki/CADUCEUS expert system which tackled internal medicine diagnosis. Internist attempted to capture the expertise of the chairman of internal medicine at the University of Pittsburgh School of Medicine https://en.wikipedia.org/wiki/University of Pittsburgh School of Medicine while CADUCEUS could eventually diagnose up to 1000 different diseases. - GUIDON, which showed how a knowledge base built for expert problem solving could be repurposed for teaching.<sup \ 48\ cite note-FOOTNOTEClancey1987-49 </sup - XCON https://en.wikipedia.org/wiki/XCON , to configure VAX computers, a then laborious process that could take up to 90 days. XCON reduced the time to about 90 minutes.<sup \ 9\ cite note-FOOTNOTEKautz2022110-9 </sup DENDRAL https://en.wikipedia.org/wiki/DENDRAL is considered the first expert system that relied on knowledge-intensive problem-solving. It is described below, by Ed Feigenbaum https://en.wikipedia.org/wiki/Ed Feigenbaum , from a Communications of the ACM https://en.wikipedia.org/wiki/Communications of the ACM interview, Interview with Ed Feigenbaum https://cacm.acm.org/magazines/2010/6/92472-an-interview-with-ed-feigenbaum/fulltext%7CAn : One of the people at Stanford interested in computer-based models of mind was Joshua Lederberg https://en.wikipedia.org/wiki/Joshua Lederberg , the 1958 Nobel Prize winner in genetics. When I told him I wanted an induction "sandbox", he said, "I have just the one for you." His lab was doing mass spectrometry of amino acids. The question was: how do you go from looking at the spectrum of an amino acid to the chemical structure of the amino acid? That's how we started the DENDRAL Project: I was good at heuristic search methods, and he had an algorithm that was good at generating the chemical problem space. We did not have a grandiose vision. We worked bottom up. Our chemist was Carl Djerassi https://en.wikipedia.org/wiki/Carl Djerassi , inventor of the chemical behind the birth control pill, and also one of the world's most respected mass spectrometrists. Carl and his postdocs were world-class experts in mass spectrometry. We began to add to their knowledge, inventing knowledge of engineering as we went along. These experiments amounted to titrating DENDRAL more and more knowledge. The more you did that, the smarter the program became. We had very good results. The generalization was: in the knowledge lies the power. That was the big idea. In my career that is the huge, "Ah ha ," and it wasn't the way AI was being done previously. Sounds simple, but it's probably AI's most powerful generalization. \ 42\ cite note-Feignebaum Interview-43 The other expert systems mentioned above came after DENDRAL. MYCIN exemplifies the classic expert system architecture of a knowledge-base of rules coupled to a symbolic reasoning mechanism, including the use of certainty factors to handle uncertainty. GUIDON shows how an explicit knowledge base can be repurposed for a second application, tutoring, and is an example of an intelligent tutoring system https://en.wikipedia.org/wiki/Intelligent tutoring system , a particular kind of knowledge-based application. Clancey showed that it was not sufficient simply to use MYCIN https://en.wikipedia.org/wiki/MYCIN 's rules for instruction, but that he also needed to add rules for dialogue management and student modeling.<sup \ 48\ cite note-FOOTNOTEClancey1987-49 </sup XCON is significant because of the millions of dollars it saved DEC https://en.wikipedia.org/wiki/Digital Equipment Corporation , which triggered the expert system boom where most all major corporations in the US had expert systems groups, to capture corporate expertise, preserve it, and automate it: By 1988, DEC's AI group had 40 expert systems deployed, with more on the way. DuPont had 100 in use and 500 in development. Nearly every major U.S. corporation had its own Al group and was either using or investigating expert systems. \ 47\ cite note-FOOTNOTERussellNorvig202123-48 Chess expert knowledge was encoded in Deep Blue <https://en.wikipedia.org/wiki/Deep Blue chess computer . In 1996, this allowed IBM https://en.wikipedia.org/wiki/IBM 's Deep Blue <https://en.wikipedia.org/wiki/Deep Blue chess computer , with the help of symbolic AI, to win in a game of chess against the world champion at that time, Garry Kasparov https://en.wikipedia.org/wiki/Garry Kasparov . \ 49\ cite note-50 Architecture of knowledge-based and expert systems edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=12 A key component of the system architecture for all expert systems is the knowledge base, which stores facts and rules for problem-solving.<sup \ 50\ cite note-FOOTNOTEHayes-RothMurrayAdelman2015-51 </sup The simplest approach for an expert system knowledge base is simply a collection or network of production rules <https://en.wikipedia.org/wiki/Production system computer science . Production rules connect symbols in a relationship similar to an If-Then statement. The expert system processes the rules to make deductions and to determine what additional information it needs, i.e. what questions to ask, using human-readable symbols. For example, OPS5 https://en.wikipedia.org/wiki/OPS5 , CLIPS https://en.wikipedia.org/wiki/CLIPS and their successors Jess <https://en.wikipedia.org/wiki/Jess programming language and Drools https://en.wikipedia.org/wiki/Drools operate in this fashion. Expert systems can operate in either a forward chaining https://en.wikipedia.org/wiki/Forward chaining – from evidence to conclusions – or backward chaining https://en.wikipedia.org/wiki/Backward chaining – from goals to needed data and prerequisites – manner. More advanced knowledge-based systems, such as Soar <https://en.wikipedia.org/wiki/Soar cognitive architecture can also perform meta-level reasoning, that is reasoning about their own reasoning in terms of deciding how to solve problems and monitoring the success of problem-solving strategies. Blackboard systems https://en.wikipedia.org/wiki/Blackboard system are a second kind of knowledge-based https://en.wikipedia.org/wiki/Knowledge-based system or expert system https://en.wikipedia.org/wiki/Expert system architecture. They model a community of experts incrementally contributing, where they can, to solve a problem. The problem is represented in multiple levels of abstraction or alternate views. The experts knowledge sources volunteer their services whenever they recognize they can contribute. Potential problem-solving actions are represented on an agenda that is updated as the problem situation changes. A controller decides how useful each contribution is, and who should make the next problem-solving action. One example, the BB1 blackboard architecture<sup \ 51\ cite note-BB1-52 </sup was originally inspired by studies of how humans plan to perform multiple tasks in a trip.<sup \ 52\ cite note-OPM-53 </sup An innovation of BB1 was to apply the same blackboard model to solving its control problem, i.e., its controller performed meta-level reasoning with knowledge sources that monitored how well a plan or the problem-solving was proceeding and could switch from one strategy to another as conditions – such as goals or times – changed. BB1 has been applied in multiple domains: construction site planning, intelligent tutoring systems, and real-time patient monitoring. The second AI winter, 1988–1993 edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=13 At the height of the AI boom, companies such as Symbolics https://en.wikipedia.org/wiki/Symbolics , LMI https://en.wikipedia.org/wiki/Lisp Machines , and Texas Instruments https://en.wikipedia.org/wiki/Texas Instruments were selling LISP machines https://en.wikipedia.org/wiki/LISP machine specifically targeted to accelerate the development of AI applications and research. In addition, several artificial intelligence companies, such as Teknowledge and Inference Corporation https://en.wikipedia.org/wiki/Inference Corporation , were selling expert system shells, training, and consulting to corporations. Unfortunately, the AI boom did not last and Kautz best describes the second AI winter that followed: Many reasons can be offered for the arrival of the second AI winter. The hardware companies failed when much more cost-effective general Unix workstations from Sun https://en.wikipedia.org/wiki/Sun Microsystems together with good compilers for LISP and Prolog came onto the market. Many commercial deployments of expert systems were discontinued when they proved too costly to maintain. Medical expert systems never caught on for several reasons: the difficulty in keeping them up to date; the challenge for medical professionals to learn how to use a bewildering variety of different expert systems for different medical conditions; and perhaps most crucially, the reluctance of doctors to trust a computer-made diagnosis over their gut instinct, even for specific domains where the expert systems could outperform an average doctor. Venture capital money deserted AI practically overnight. The world AI conference IJCAI hosted an enormous and lavish trade show and thousands of nonacademic attendees in 1987 in Vancouver https://en.wikipedia.org/wiki/Vancouver ; the main AI conference the following year, AAAI 1988 in St. Paul https://en.wikipedia.org/wiki/St. Paul, Minnesota , was a small and strictly academic affair. \ 9\ cite note-FOOTNOTEKautz2022110-9 Adding in more rigorous foundations, 1993–2011 edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=14 Uncertain reasoning edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=15 Both statistical approaches and extensions to logic were tried. One statistical approach, hidden Markov models https://en.wikipedia.org/wiki/Hidden Markov model , had already been popularized in the 1980s for speech recognition work.<sup \ 11\ cite note-FOOTNOTERussellNorvig202125-11 </sup Subsequently, in 1988, Judea Pearl https://en.wikipedia.org/wiki/Judea Pearl popularized the use of Bayesian Networks https://en.wikipedia.org/wiki/Bayesian Networks as a sound but efficient way of handling uncertain reasoning with his publication of the book Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference.<sup \ 53\ cite note-FOOTNOTEPearl1988-54 </sup and Bayesian approaches were applied successfully in expert systems.<sup \ 54\ cite note-FOOTNOTESpiegelhalterDawidLauritzenCowell1993-55 </sup Even later, in the 1990s, statistical relational learning, an approach that combines probability with logical formulas, allowed probability to be combined with first-order logic, e.g., with either Markov Logic Networks https://en.wikipedia.org/wiki/Markov logic network or Probabilistic Soft Logic https://en.wikipedia.org/wiki/Probabilistic Soft Logic . Other, non-probabilistic extensions to first-order logic to support were also tried. For example, non-monotonic reasoning https://en.wikipedia.org/wiki/Non-monotonic reasoning could be used with truth maintenance systems https://en.wikipedia.org/wiki/Reason maintenance . A truth maintenance system https://en.wikipedia.org/wiki/Truth maintenance system tracked assumptions and justifications for all inferences. It allowed inferences to be withdrawn when assumptions were found out to be incorrect or a contradiction was derived. Explanations could be provided for an inference by explaining which rules were applied https://en.wikipedia.org/wiki/Explainable artificial intelligence to create it and then continuing through underlying inferences and rules all the way back to root assumptions. \ 55\ cite note-FOOTNOTERussellNorvig2021335–337-56 Lotfi Zadeh https://en.wikipedia.org/wiki/Lotfi Zadeh had introduced a different kind of extension to handle the representation of vagueness. For example, in deciding how "heavy" or "tall" a man is, there is frequently no clear "yes" or "no" answer, and a predicate for heavy or tall would instead return values between 0 and 1. Those values represented to what degree the predicates were true. His fuzzy logic https://en.wikipedia.org/wiki/Fuzzy logic further provided a means for propagating combinations of these values through logical formulas. \ 56\ cite note-FOOTNOTERussellNorvig2021459-57 Machine learning edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=16 Symbolic machine learning approaches were investigated to address the knowledge acquisition https://en.wikipedia.org/wiki/Knowledge acquisition bottleneck. One of the earliest is Meta-DENDRAL https://en.wikipedia.org/wiki/Dendral Meta-Dendral . Meta-DENDRAL used a generate-and-test technique to generate plausible rule hypotheses to test against spectra. Domain and task knowledge reduced the number of candidates tested to a manageable size. Feigenbaum https://en.wikipedia.org/wiki/Ed Feigenbaum described Meta-DENDRAL as ...the culmination of my dream of the early to mid-1960s having to do with theory formation. The conception was that you had a problem solver like DENDRAL that took some inputs and produced an output. In doing so, it used layers of knowledge to steer and prune the search. That knowledge got in there because we interviewed people. But how did the people get the knowledge? By looking at thousands of spectra. So we wanted a program that would look at thousands of spectra and infer the knowledge of mass spectrometry that DENDRAL could use to solve individual hypothesis formation problems. We did it. We were even able to publish new knowledge of mass spectrometry in the Journal of the American Chemical Society https://en.wikipedia.org/wiki/Journal of the American Chemical Society , giving credit only in a footnote that a program, Meta-DENDRAL, actually did it. We were able to do something that had been a dream: to have a computer program come up with a new and publishable piece of science. \ 42\ cite note-Feignebaum Interview-43 In contrast to the knowledge-intensive approach of Meta-DENDRAL, Ross Quinlan https://en.wikipedia.org/wiki/Ross Quinlan invented a domain-independent approach to statistical classification, decision tree learning https://en.wikipedia.org/wiki/Decision tree learning , starting first with ID3 https://en.wikipedia.org/wiki/ID3 algorithm <sup \ 57\ cite note-58 </sup and then later extending its capabilities to C4.5 https://en.wikipedia.org/wiki/C4.5 .<sup \ 58\ cite note-59 </sup The decision trees created are glass box https://en.wikipedia.org/wiki/Glass box , interpretable classifiers, with human-interpretable classification rules. Advances were made in understanding machine learning theory, too. Tom Mitchell https://en.wikipedia.org/wiki/Tom M. Mitchell introduced version space learning https://en.wikipedia.org/wiki/Version space learning which describes learning as a search through a space of hypotheses, with upper, more general, and lower, more specific, boundaries encompassing all viable hypotheses consistent with the examples seen so far.<sup \ 59\ cite note-60 </sup More formally, Valiant https://en.wikipedia.org/wiki/Leslie Valiant introduced Probably Approximately Correct Learning https://en.wikipedia.org/wiki/Probably approximately correct learning PAC Learning , a framework for the mathematical analysis of machine learning. \ 60\ cite note-61 Symbolic machine learning encompassed more than learning by example. E.g., John Anderson <https://en.wikipedia.org/wiki/John Robert Anderson psychologist provided a cognitive model https://en.wikipedia.org/wiki/Cognitive model of human learning where skill practice results in a compilation of rules from a declarative format to a procedural format with his ACT-R https://en.wikipedia.org/wiki/ACT-R cognitive architecture https://en.wikipedia.org/wiki/Cognitive architecture . For example, a student might learn to apply "Supplementary angles are two angles whose measures sum 180 degrees" as several different procedural rules. E.g., one rule might say that if X and Y are supplementary and you know X, then Y will be 180 - X. He called his approach "knowledge compilation". ACT-R https://en.wikipedia.org/wiki/ACT-R has been used successfully to model aspects of human cognition, such as learning and retention. ACT-R is also used in intelligent tutoring systems https://en.wikipedia.org/wiki/Intelligent tutoring systems , called cognitive tutors https://en.wikipedia.org/wiki/Cognitive tutors , to successfully teach geometry, computer programming, and algebra to school children. \ 61\ cite note-62 Inductive logic programming was another approach to learning that allowed logic programs to be synthesized from input-output examples. E.g., Ehud Shapiro https://en.wikipedia.org/wiki/Ehud Shapiro 's MIS Model Inference System could synthesize Prolog programs from examples. \ 62\ cite note-63 John R. Koza https://en.wikipedia.org/wiki/John R. Koza applied genetic algorithms https://en.wikipedia.org/wiki/Genetic algorithms to program synthesis https://en.wikipedia.org/wiki/Program synthesis to create genetic programming https://en.wikipedia.org/wiki/Genetic programming , which he used to synthesize LISP programs. Finally, Zohar Manna https://en.wikipedia.org/wiki/Zohar Manna and Richard Waldinger https://en.wikipedia.org/wiki/Richard Waldinger provided a more general approach to program synthesis https://en.wikipedia.org/wiki/Program synthesis that synthesizes a functional program https://en.wikipedia.org/wiki/Functional programming in the course of proving its specifications to be correct. \ 63\ cite note-64 As an alternative to logic, Roger Schank https://en.wikipedia.org/wiki/Roger Schank introduced case-based reasoning CBR . The CBR approach outlined in his book, Dynamic Memory,<sup \ 64\ cite note-Schank-65 </sup focuses first on remembering key problem-solving cases for future use and generalizing them where appropriate. When faced with a new problem, CBR retrieves the most similar previous case and adapts it to the specifics of the current problem.<sup \ 65\ cite note-66 </sup Another alternative to logic, genetic algorithms https://en.wikipedia.org/wiki/Genetic algorithms and genetic programming https://en.wikipedia.org/wiki/Genetic programming are based on an evolutionary model of learning, where sets of rules are encoded into populations, the rules govern the behavior of individuals, and selection of the fittest prunes out sets of unsuitable rules over many generations. \ 66\ cite note-67 Symbolic machine learning was applied to learning concepts, rules, heuristics, and problem-solving. Approaches, other than those above, include: 1. Learning from instruction or advice—i.e., taking human instruction, posed as advice, and determining how to operationalize it in specific situations. For example, in a game of Hearts, learning exactly how to play a hand to "avoid taking points."<sup \ 67\ cite note-68 </sup 2. Learning from exemplars—improving performance by accepting subject-matter expert SME feedback during training. When problem-solving fails, querying the expert to either learn a new exemplar for problem-solving or to learn a new explanation as to exactly why one exemplar is more relevant than another. For example, the program Protos learned to diagnose tinnitus cases by interacting with an audiologist.<sup \ 68\ cite note-69 </sup 3. Learning by analogy—constructing problem solutions based on similar problems seen in the past, and then modifying their solutions to fit a new situation or domain.<sup \ 69\ cite note-70 </sup <sup \ 70\ cite note-71 </sup 4. Apprentice learning systems—learning novel solutions to problems by observing human problem-solving. Domain knowledge explains why novel solutions are correct and how the solution can be generalized. LEAP learned how to design VLSI circuits by observing human designers.<sup \ 71\ cite note-72 </sup 5. Learning by discovery—i.e., creating tasks to carry out experiments and then learning from the results. Doug Lenat https://en.wikipedia.org/wiki/Douglas Lenat 's Eurisko https://en.wikipedia.org/wiki/Eurisko , for example, learned heuristics to beat human players at the Traveller <https://en.wikipedia.org/wiki/Traveller role-playing game role-playing game for two years in a row.<sup \ 72\ cite note-73 </sup 6. Learning macro-operators—i.e., searching for useful macro-operators to be learned from sequences of basic problem-solving actions. Good macro-operators simplify problem-solving by allowing problems to be solved at a more abstract level.<sup \ 73\ cite note-74 </sup Deep learning and neuro-symbolic AI 2011–now edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=17 With the rise of deep learning, the symbolic AI approach has been compared to deep learning as complementary "...with parallels having been drawn many times by AI researchers between Kahneman's https://en.wikipedia.org/wiki/Daniel Kahneman research on human reasoning and decision making – reflected in his book Thinking, Fast and Slow https://en.wikipedia.org/wiki/Thinking, Fast and Slow – and the so-called "AI systems 1 and 2", which would in principle be modelled by deep learning and symbolic reasoning, respectively." In this view, symbolic reasoning is more apt for deliberative reasoning, planning, and explanation while deep learning is more apt for fast pattern recognition in perceptual applications with noisy data. \ 74\ cite note-Rossi-75 \ 75\ cite note-Selman-76 Neuro-symbolic AI: integrating neural and symbolic approaches edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=18 Neuro-symbolic AI attempts to integrate neural and symbolic architectures in a manner that addresses strengths and weaknesses of each, in a complementary fashion, in order to support robust AI capable of reasoning, learning, and cognitive modeling. As argued by Valiant https://en.wikipedia.org/wiki/Leslie Valiant <sup \ 76\ cite note-FOOTNOTEValiant2008-77 </sup and many others,<sup \ 77\ cite note-FOOTNOTEGarcezBesoldDe RaedtFöldiák2015-78 </sup the effective construction of rich computational cognitive models https://en.wikipedia.org/wiki/Cognitive model demands the combination of sound symbolic reasoning and efficient machine learning models. Gary Marcus https://en.wikipedia.org/wiki/Gary Marcus , similarly, argues that: "We cannot construct rich cognitive models in an adequate, automated way without the triumvirate of hybrid architecture, rich prior knowledge, and sophisticated techniques for reasoning.",<sup \ 78\ cite note-FOOTNOTEMarcus202044-79 </sup and in particular: "To build a robust, knowledge-driven approach to AI we must have the machinery of symbol-manipulation in our toolkit. Too much of useful knowledge is abstract to make do without tools that represent and manipulate abstraction, and to date, the only machinery that we know of that can manipulate such abstract knowledge reliably is the apparatus of symbol manipulation." \ 79\ cite note-FOOTNOTEMarcus202017-80 Henry Kautz https://en.wikipedia.org/wiki/Henry Kautz , \ 17\ cite note-FOOTNOTEKautz2020-17 Francesca Rossi https://en.wikipedia.org/wiki/Francesca Rossi ,<sup \ 80\ cite note-FOOTNOTERossi2022-81 </sup and Bart Selman https://en.wikipedia.org/wiki/Bart Selman <sup \ 81\ cite note-FOOTNOTESelman2022-82 </sup have also argued for a synthesis. Their arguments are based on a need to address the two kinds of thinking discussed in Daniel Kahneman https://en.wikipedia.org/wiki/Daniel Kahneman 's book, Thinking, Fast and Slow https://en.wikipedia.org/wiki/Thinking, Fast and Slow . Kahneman describes human thinking as having two components, System 1 and System 2 https://en.wikipedia.org/wiki/Thinking, Fast and Slow Two systems . System 1 is fast, automatic, intuitive and unconscious. System 2 is slower, step-by-step, and explicit. System 1 is the kind used for pattern recognition while System 2 is far better suited for planning, deduction, and deliberative thinking. In this view, deep learning best models the first kind of thinking while symbolic reasoning best models the second kind and both are needed. Garcez https://en.wikipedia.org/wiki/Artur Garcez and Lamb describe research in this area as being ongoing for at least the past twenty years,<sup \ 82\ cite note-FOOTNOTEGarcezLamb20202-83 </sup dating from their 2002 book on neurosymbolic learning systems.<sup \ 83\ cite note-FOOTNOTEGarcezBrodaGabbayGabbay2002-84 </sup A series of workshops on neuro-symbolic reasoning has been held every year since 2005. \ 84\ cite note-85 In their 2015 paper, Neural-Symbolic Learning and Reasoning: Contributions and Challenges, Garcez et al. argue that: The integration of the symbolic and connectionist paradigms of AI has been pursued by a relatively small research community over the last two decades and has yielded several significant results. Over the last decade, neural symbolic systems have been shown capable of overcoming the so-called propositional fixation of neural networks, as McCarthy 1988 put it in response to Smolensky 1988 ; see also Hinton, 1990 . Neural networks were shown capable of representing modal and temporal logics d'Avila Garcez and Lamb, 2006 and fragments of first-order logic Bader, Hitzler, Hölldobler, 2008; d'Avila Garcez, Lamb, Gabbay, 2009 . Further, neural-symbolic systems have been applied to a number of problems in the areas of bioinformatics, control engineering, software verification and adaptation, visual intelligence, ontology learning, and computer games. \ 77\ cite note-FOOTNOTEGarcezBesoldDe RaedtFöldiák2015-78 Approaches for integration are varied. Henry Kautz https://en.wikipedia.org/wiki/Henry Kautz 's taxonomy of neuro-symbolic architectures, along with some examples, follows: - Symbolic Neural symbolic—is the current approach of many neural models in natural language processing, where words or subword tokens are both the ultimate input and output of large language models. Examples include BERT <https://en.wikipedia.org/wiki/BERT language model , RoBERTa, and GPT-3 https://en.wikipedia.org/wiki/GPT-3 . - Symbolic Neural —is exemplified by AlphaGo https://en.wikipedia.org/wiki/AlphaGo , where symbolic techniques are used to call neural techniques. In this case the symbolic approach is Monte Carlo tree search https://en.wikipedia.org/wiki/Monte Carlo tree search and the neural techniques learn how to evaluate game positions. - Neural|Symbolic—uses a neural architecture to interpret perceptual data as symbols and relationships that are then reasoned about symbolically. - Neural:Symbolic → Neural—relies on symbolic reasoning to generate or label training data that is subsequently learned by a deep learning model, e.g., to train a neural model for symbolic computation by using a Macsyma https://en.wikipedia.org/wiki/Macsyma -like symbolic mathematics system to create or label examples. - Neural {Symbolic}—uses a neural net that is generated from symbolic rules. An example is the Neural Theorem Prover,<sup \ 85\ cite note-86 </sup which constructs a neural network from an AND–OR https://en.wikipedia.org/wiki/And–or tree proof tree generated from knowledge base rules and terms. Logic Tensor Networks<sup \ 86\ cite note-87 </sup also fall into this category. - Neural Symbolic —allows a neural model to directly call a symbolic reasoning engine, e.g., to perform an action or evaluate a state. Many key research questions remain, such as: - What is the best way to integrate neural and symbolic architectures?<sup \ 87\ cite note-:0-88 </sup - How should symbolic structures be represented within neural networks and extracted from them? - How should common-sense knowledge be learned and reasoned about? - How can abstract knowledge that is hard to encode logically be handled? Techniques and contributions edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=19 This section provides an overview of techniques and contributions in an overall context leading to many other, more detailed articles in Wikipedia. Sections on Machine Learning Machine learning and Uncertain Reasoning Uncertain reasoning are covered earlier in the history section A short history . AI programming languages edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=20 The key AI programming language in the US during the last symbolic AI boom period was LISP <https://en.wikipedia.org/wiki/LISP programming language . LISP <https://en.wikipedia.org/wiki/LISP programming language is the second oldest programming language after FORTRAN https://en.wikipedia.org/wiki/FORTRAN and was created in 1958 by John McCarthy <https://en.wikipedia.org/wiki/John McCarthy computer scientist . LISP provided the first read–eval–print loop https://en.wikipedia.org/wiki/Read–eval–print loop to support rapid program development. Compiled functions could be freely mixed with interpreted functions. Program tracing, stepping, and breakpoints were also provided, along with the ability to change values or functions and continue from breakpoints or errors. It had the first self-hosting compiler <https://en.wikipedia.org/wiki/Self-hosting compilers , meaning that the compiler itself was originally written in LISP and then ran interpretively to compile the compiler code. Other key innovations pioneered by LISP that have spread to other programming languages include: Programs were themselves data structures that other programs could operate on, allowing the easy definition of higher-level languages. In contrast to the US, in Europe the key AI programming language during that same period was Prolog https://en.wikipedia.org/wiki/Prolog . Prolog provided a built-in store of facts and clauses that could be queried by a read–eval–print loop https://en.wikipedia.org/wiki/Read–eval–print loop . The store could act as a knowledge base and the clauses could act as rules or a restricted form of logic. As a subset of first-order logic Prolog was based on Horn clauses https://en.wikipedia.org/wiki/Horn clauses with a closed-world assumption https://en.wikipedia.org/wiki/Closed-world assumption —any facts not known were considered false—and a unique name assumption https://en.wikipedia.org/wiki/Unique name assumption for primitive terms—e.g., the identifier barack obama was considered to refer to exactly one object. Backtracking https://en.wikipedia.org/wiki/Backtracking and unification <https://en.wikipedia.org/wiki/Unification computer science are built-in to Prolog. Alain Colmerauer https://en.wikipedia.org/wiki/Alain Colmerauer and Philippe Roussel are credited as the inventors of Prolog. Prolog is a form of logic programming, which was invented by Robert Kowalski https://en.wikipedia.org/wiki/Robert Kowalski . Its history was also influenced by Carl Hewitt https://en.wikipedia.org/wiki/Carl Hewitt 's PLANNER https://en.wikipedia.org/wiki/PLANNER , an assertional database with pattern-directed invocation of methods. For more detail see the section on the origins of Prolog in the PLANNER article <https://en.wikipedia.org/wiki/Planner programming language The genesis of Prolog . Prolog is also a kind of declarative programming https://en.wikipedia.org/wiki/Declarative programming . The logic clauses that describe programs are directly interpreted to run the programs specified. No explicit series of actions is required, as is the case with imperative programming https://en.wikipedia.org/wiki/Imperative programming languages. Japan championed Prolog for its Fifth Generation Project https://en.wikipedia.org/wiki/Fifth Generation Project , intending to build special hardware for high performance. Similarly, LISP machines https://en.wikipedia.org/wiki/LISP machines were built to run LISP, but as the second AI boom turned to bust these companies could not compete with new workstations that could now run LISP or Prolog natively at comparable speeds. See the history section A short history for more detail. Smalltalk https://en.wikipedia.org/wiki/Smalltalk was another influential AI programming language. For example, it introduced metaclasses https://en.wikipedia.org/wiki/Metaclasses and, along with Flavors <https://en.wikipedia.org/wiki/Flavors programming language and CommonLoops https://en.wikipedia.org/wiki/CommonLoops , influenced the Common Lisp Object System https://en.wikipedia.org/wiki/Common Lisp Object System , or CLOS https://en.wikipedia.org/wiki/CLOS , that is now part of Common Lisp https://en.wikipedia.org/wiki/Common Lisp , the current standard Lisp dialect. CLOS https://en.wikipedia.org/wiki/CLOS is a Lisp-based object-oriented system that allows multiple inheritance https://en.wikipedia.org/wiki/Multiple inheritance , in addition to incremental extensions to both classes and metaclasses, thus providing a run-time meta-object protocol https://en.wikipedia.org/wiki/Meta-object protocol . \ 88\ cite note-meta-object protocol-89 For other AI programming languages see this list of programming languages for artificial intelligence https://en.wikipedia.org/wiki/List of programming languages for artificial intelligence . Currently, Python <https://en.wikipedia.org/wiki/Python programming language , a multi-paradigm programming language https://en.wikipedia.org/wiki/Multi-paradigm programming language , is the most popular programming language, partly due to its extensive package library that supports data science https://en.wikipedia.org/wiki/Data science , natural language processing, and deep learning. Python <https://en.wikipedia.org/wiki/Python programming language includes a read-eval-print loop, functional elements such as higher-order functions https://en.wikipedia.org/wiki/Higher-order function , and object-oriented programming https://en.wikipedia.org/wiki/Object-oriented programming that includes metaclasses. Search edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=21 Search arises in many kinds of problem solving, including planning https://en.wikipedia.org/wiki/Automated planning , constraint satisfaction https://en.wikipedia.org/wiki/Constraint satisfaction , and playing games such as checkers https://en.wikipedia.org/wiki/Checkers , chess https://en.wikipedia.org/wiki/Chess , and go <https://en.wikipedia.org/wiki/Go game . The best known AI-search tree search algorithms are breadth-first search https://en.wikipedia.org/wiki/Breadth-first search , depth-first search https://en.wikipedia.org/wiki/Depth-first search , A https://en.wikipedia.org/wiki/A search algorithm , and Monte Carlo Search https://en.wikipedia.org/wiki/Monte Carlo tree search . Key search algorithms for Boolean satisfiability https://en.wikipedia.org/wiki/Boolean satisfiability are WalkSAT https://en.wikipedia.org/wiki/WalkSAT , conflict-driven clause learning https://en.wikipedia.org/wiki/Conflict-driven clause learning , and the DPLL algorithm https://en.wikipedia.org/wiki/DPLL algorithm . For adversarial search when playing games, alpha-beta pruning https://en.wikipedia.org/wiki/Alpha-beta pruning , branch and bound https://en.wikipedia.org/wiki/Branch and bound , and minimax https://en.wikipedia.org/wiki/Minimax were early contributions. Knowledge representation and reasoning edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=22 Multiple different approaches to represent knowledge and then reason with those representations have been investigated. Below is a quick overview of approaches to knowledge representation and automated reasoning. Knowledge representation edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=23 Semantic networks https://en.wikipedia.org/wiki/Semantic networks , conceptual graphs https://en.wikipedia.org/wiki/Conceptual graphs , frames <https://en.wikipedia.org/wiki/Frame artificial intelligence , and logic https://en.wikipedia.org/wiki/Formal logic are all approaches to modeling knowledge such as domain knowledge, problem-solving knowledge, and the semantic meaning of language. Ontologies https://en.wikipedia.org/wiki/Ontologies model key concepts and their relationships in a domain. Example ontologies are YAGO <https://en.wikipedia.org/wiki/YAGO database , WordNet https://en.wikipedia.org/wiki/WordNet , and DOLCE https://en.wikipedia.org/wiki/Upper ontology DOLCE . DOLCE https://en.wikipedia.org/wiki/Upper ontology DOLCE is an example of an upper ontology https://en.wikipedia.org/wiki/Upper ontology that can be used for any domain while WordNet is a lexical resource that can also be viewed as an ontology https://en.wikipedia.org/wiki/Upper ontology WordNet . YAGO incorporates WordNet as part of its ontology, to align facts extracted from Wikipedia https://en.wikipedia.org/wiki/Wikipedia with WordNet synsets https://en.wikipedia.org/wiki/Synsets . The Disease Ontology https://en.wikipedia.org/wiki/Disease Ontology is an example of a medical ontology currently being used. Description logic https://en.wikipedia.org/wiki/Description logic is a logic for automated classification of ontologies and for detecting inconsistent classification data. OWL https://en.wikipedia.org/wiki/Web Ontology Language is a language used to represent ontologies with description logic https://en.wikipedia.org/wiki/Description logic . Protégé <https://en.wikipedia.org/wiki/Protégé software is an ontology editor that can read in OWL https://en.wikipedia.org/wiki/Web Ontology Language ontologies and then check consistency with deductive classifiers https://en.wikipedia.org/wiki/Deductive classifier such as such as HermiT. \ 89\ cite note-HermiT-90 First-order logic is more general than description logic. The automated theorem provers discussed below can prove theorems in first-order logic. Horn clause https://en.wikipedia.org/wiki/Horn clause logic is more restricted than first-order logic and is used in logic programming languages such as Prolog. Extensions to first-order logic include temporal logic https://en.wikipedia.org/wiki/Temporal logic , to handle time; epistemic logic https://en.wikipedia.org/wiki/Epistemic logic , to reason about agent knowledge; modal logic https://en.wikipedia.org/wiki/Modal logic , to handle possibility and necessity; and probabilistic logics https://en.wikipedia.org/wiki/Probabilistic logic to handle logic and probability together. Automatic theorem proving edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=24 Examples of automated theorem provers for first-order logic are: Prover9 https://en.wikipedia.org/wiki/Prover9 can be used in conjunction with the Mace4 https://en.wikipedia.org/wiki/Mace4 model checker https://en.wikipedia.org/wiki/Model checking . ACL2 https://en.wikipedia.org/wiki/ACL2 is a theorem prover that can handle proofs by induction and is a descendant of the Boyer-Moore Theorem Prover, also known as Nqthm https://en.wikipedia.org/wiki/Nqthm . Reasoning in knowledge-based systems edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=25 Knowledge-based systems have an explicit knowledge base https://en.wikipedia.org/wiki/Knowledge base , typically of rules, to enhance reusability across domains by separating procedural code and domain knowledge. A separate inference engine https://en.wikipedia.org/wiki/Inference engine processes rules and adds, deletes, or modifies a knowledge store. Forward chaining https://en.wikipedia.org/wiki/Forward chaining inference engines are the most common, and are seen in CLIPS https://en.wikipedia.org/wiki/CLIPS and OPS5 https://en.wikipedia.org/wiki/OPS5 . Backward chaining https://en.wikipedia.org/wiki/Backward chaining occurs in Prolog, where a more limited logical representation is used, Horn Clauses https://en.wikipedia.org/wiki/Horn clause . Pattern-matching, specifically unification <https://en.wikipedia.org/wiki/Unification computer science , is used in Prolog. A more flexible kind of problem-solving occurs when reasoning about what to do next occurs, rather than simply choosing one of the available actions. This kind of meta-level reasoning is used in Soar and in the BB1 blackboard architecture. Cognitive architectures such as ACT-R may have additional capabilities, such as the ability to compile frequently used knowledge into higher-level chunks <https://en.wikipedia.org/wiki/Chunking psychology . Commonsense reasoning edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=26 Marvin Minsky https://en.wikipedia.org/wiki/Marvin Minsky first proposed frames as a way of interpreting common visual situations, such as an office, and Roger Schank extended this idea to scripts https://en.wikipedia.org/wiki/Script theory for common routines, such as dining out. Cyc https://en.wikipedia.org/wiki/Cyc has attempted to capture useful common-sense knowledge and has "micro-theories" to handle particular kinds of domain-specific reasoning. Qualitative simulation, such as Benjamin Kuipers https://en.wikipedia.org/wiki/Benjamin Kuipers 's QSIM,<sup \ 90\ cite note-QSIM-91 </sup approximates human reasoning about naive physics, such as what happens when we heat a liquid in a pot on the stove. We expect it to heat and possibly boil over, even though we may not know its temperature, its boiling point, or other details, such as atmospheric pressure. Similarly, Allen <https://en.wikipedia.org/wiki/James F. Allen computer scientist 's temporal interval algebra https://en.wikipedia.org/wiki/Allen's interval algebra is a simplification of reasoning about time and Region Connection Calculus https://en.wikipedia.org/wiki/Region Connection Calculus is a simplification of reasoning about spatial relationships. Both can be solved with constraint solvers https://en.wikipedia.org/wiki/Constraint programming . Constraints and constraint-based reasoning edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=27 Constraint solvers perform a more limited kind of inference than first-order logic. They can simplify sets of spatiotemporal constraints, such as those for RCC https://en.wikipedia.org/wiki/Region connection calculus or Temporal Algebra https://en.wikipedia.org/wiki/Allen's interval algebra , along with solving other kinds of puzzle problems, such as Wordle https://en.wikipedia.org/wiki/Wordle , Sudoku https://en.wikipedia.org/wiki/Sudoku , cryptarithmetic problems https://en.wikipedia.org/wiki/Verbal arithmetic , and so on. Constraint logic programming https://en.wikipedia.org/wiki/Constraint logic programming can be used to solve scheduling problems, for example with constraint handling rules https://en.wikipedia.org/wiki/Constraint handling rules CHR . Automated planning edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=28 The General Problem Solver https://en.wikipedia.org/wiki/General Problem Solver GPS cast planning as problem-solving used means-ends analysis https://en.wikipedia.org/wiki/Means-ends analysis to create plans. STRIPS https://en.wikipedia.org/wiki/Stanford Research Institute Problem Solver took a different approach, viewing planning as theorem proving. Graphplan https://en.wikipedia.org/wiki/Graphplan takes a least-commitment approach to planning, rather than sequentially choosing actions from an initial state, working forwards, or a goal state if working backwards. Satplan https://en.wikipedia.org/wiki/Satplan is an approach to planning where a planning problem is reduced to a Boolean satisfiability problem https://en.wikipedia.org/wiki/Boolean satisfiability problem . Natural language processing edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=29 Natural language processing focuses on treating language as data to perform tasks such as identifying topics without necessarily understanding the intended meaning. Natural language understanding, in contrast, constructs a meaning representation and uses that for further processing, such as answering questions. Parsing https://en.wikipedia.org/wiki/Parsing , tokenizing <https://en.wikipedia.org/wiki/Tokenization lexical analysis , spelling correction https://en.wikipedia.org/wiki/Spell checker , part-of-speech tagging https://en.wikipedia.org/wiki/Part-of-speech tagging , noun and verb phrase chunking https://en.wikipedia.org/wiki/Shallow parsing are all aspects of natural language processing long handled by symbolic AI, but since improved by deep learning approaches. In symbolic AI, discourse representation theory https://en.wikipedia.org/wiki/Discourse representation theory and first-order logic have been used to represent sentence meanings. Latent semantic analysis https://en.wikipedia.org/wiki/Latent semantic analysis LSA and explicit semantic analysis https://en.wikipedia.org/wiki/Explicit semantic analysis also provided vector representations of documents. In the latter case, vector components are interpretable as concepts named by Wikipedia articles. New deep learning approaches based on Transformer models <https://en.wikipedia.org/wiki/Transformer machine learning model have now eclipsed these earlier symbolic AI approaches and attained state-of-the-art performance in natural language processing . However, Transformer models are opaque and do not yet produce human-interpretable semantic representations for sentences and documents. Instead, they produce task-specific vectors where the meaning of the vector components is opaque. Agents and multi-agent systems edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=30 Agents https://en.wikipedia.org/wiki/Software agent are autonomous systems embedded in an environment they perceive and act upon in some sense. Russell and Norvig's standard textbook on artificial intelligence is organized to reflect agent architectures of increasing sophistication.<sup \ 91\ cite note-FOOTNOTERussellNorvig2021-92 </sup The sophistication of agents varies from simple reactive agents, to those with a model of the world and automated planning https://en.wikipedia.org/wiki/Automated planning capabilities, possibly a BDI agent https://en.wikipedia.org/wiki/Belief–desire–intention software model , i.e., one with beliefs, desires, and intentions – or alternatively a reinforcement learning https://en.wikipedia.org/wiki/Reinforcement learning model learned over time to choose actions – up to a combination of alternative architectures, such as a neuro-symbolic architecture<sup \ 87\ cite note-:0-88 </sup that includes deep learning for perception. \ 92\ cite note-93 In contrast, a multi-agent system https://en.wikipedia.org/wiki/Multi-agent system consists of multiple agents that communicate amongst themselves with some inter-agent communication language such as Knowledge Query and Manipulation Language https://en.wikipedia.org/wiki/Knowledge Query and Manipulation Language KQML . The agents need not all have the same internal architecture. Advantages of multi-agent systems include the ability to divide work among the agents and to increase fault tolerance when agents are lost. Research problems include how agents reach consensus https://en.wikipedia.org/wiki/Consensus dynamics , distributed problem solving https://en.wikipedia.org/wiki/Cooperative distributed problem solving , multi-agent learning https://en.wikipedia.org/wiki/Multi-agent learning , multi-agent planning https://en.wikipedia.org/wiki/Multi-agent planning , and distributed constraint optimization https://en.wikipedia.org/wiki/Distributed constraint optimization . Controversies edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=31 Controversies arose from early on in symbolic AI, both within the field—e.g., between logicists the pro-logic "neats" https://en.wikipedia.org/wiki/Neats and scruffies and non-logicists the anti-logic "scruffies" https://en.wikipedia.org/wiki/Neats and scruffies —and between those who embraced AI but rejected symbolic approaches—primarily connectionists https://en.wikipedia.org/wiki/Connectionism —and those outside the field. Critiques from outside of the field were primarily from philosophers, on intellectual grounds, but also from funding agencies, especially during the two AI winters. The Frame Problem: knowledge representation challenges for first-order logic edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=32 Limitations were discovered in using simple first-order logic to reason about dynamic domains. Problems were discovered both with regards to enumerating the preconditions for an action to succeed and in providing axioms for what did not change after an action was performed. McCarthy and Hayes introduced the Frame Problem https://en.wikipedia.org/wiki/Frame problem in 1969 in the paper, "Some Philosophical Problems from the Standpoint of Artificial Intelligence."<sup \ 93\ cite note-FOOTNOTEMcCarthyHayes1969-94 </sup A simple example occurs in "proving that one person could get into conversation with another", as an axiom asserting "if a person has a telephone he still has it after looking up a number in the telephone book" would be required for the deduction to succeed. Similar axioms would be required for other domain actions to specify what did not change. A similar problem, called the Qualification Problem https://en.wikipedia.org/wiki/Qualification problem , occurs in trying to enumerate the preconditions for an action to succeed. An infinite number of pathological conditions can be imagined, e.g., a banana in a tailpipe could prevent a car from operating correctly. McCarthy's approach to fix the frame problem was circumscription <https://en.wikipedia.org/wiki/Circumscription logic , a kind of non-monotonic logic https://en.wikipedia.org/wiki/Non-monotonic logic where deductions could be made from actions that need only specify what would change while not having to explicitly specify everything that would not change. Other non-monotonic logics https://en.wikipedia.org/wiki/Non-monotonic logic provided truth maintenance systems https://en.wikipedia.org/wiki/Reason maintenance that revised beliefs leading to contradictions. Other ways of handling more open-ended domains included probabilistic reasoning https://en.wikipedia.org/wiki/Probabilistic logic systems and machine learning to learn new concepts and rules. McCarthy's Advice Taker https://en.wikipedia.org/wiki/Advice taker can be viewed as an inspiration here, as it could incorporate new knowledge provided by a human in the form of assertions or rules. For example, experimental symbolic machine learning systems explored the ability to take high-level natural language advice and to interpret it into domain-specific actionable rules. Similar to the problems in handling dynamic domains, common-sense reasoning is also difficult to capture in formal reasoning. Examples of common-sense reasoning include implicit reasoning about how people think or general knowledge of day-to-day events, objects, and living creatures. This kind of knowledge is taken for granted and not viewed as noteworthy. Common-sense reasoning is an open area of research and challenging both for symbolic systems e.g., Cyc has attempted to capture key parts of this knowledge over more than a decade and neural systems e.g., self-driving cars https://en.wikipedia.org/wiki/Self-driving car that do not know not to drive into cones or not to hit pedestrians walking a bicycle . McCarthy viewed his Advice Taker https://en.wikipedia.org/wiki/Advice taker as having common-sense, but his definition of common-sense was different than the one above.<sup \ 94\ cite note-FOOTNOTEMcCarthy1959-95 </sup He defined a program as having common sense " if it automatically deduces for itself a sufficiently wide class of immediate consequences of anything it is told and what it already knows ." Connectionist AI: philosophical challenges and sociological conflicts edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=33 Connectionist approaches include earlier work on neural networks https://en.wikipedia.org/wiki/Artificial neural network ,<sup \ 95\ cite note-FOOTNOTENilsson19987-96 </sup such as perceptrons https://en.wikipedia.org/wiki/Perceptron ; work in the mid to late 80s, such as Danny Hillis https://en.wikipedia.org/wiki/Danny Hillis 's Connection Machine https://en.wikipedia.org/wiki/Connection Machine and Yann LeCun https://en.wikipedia.org/wiki/Yann LeCun 's advances in convolutional neural networks https://en.wikipedia.org/wiki/Convolutional neural network ; to today's more advanced approaches, such as Transformers <https://en.wikipedia.org/wiki/Transformer machine learning model , GANs https://en.wikipedia.org/wiki/Generative adversarial network , and other work in deep learning. Three philosophical positions<sup \ 96\ cite note-FOOTNOTEOlazaran1993411–416-97 </sup have been outlined among connectionists: 1. Implementationism—where connectionist architectures implement the capabilities for symbolic processing, 2. Radical connectionism—where symbolic processing is rejected totally, and connectionist architectures underlie intelligence and are fully sufficient to explain it, 3. Moderate connectionism—where symbolic processing and connectionist architectures are viewed as complementary and both are required for intelligence. Olazaran, in his sociological history of the controversies within the neural network community, described the moderate connectionism view as essentially compatible with current research in neuro-symbolic hybrids: The third and last position I would like to examine here is what I call the moderate connectionist view, a more eclectic view of the current debate between connectionism https://en.wikipedia.org/wiki/Connectionism and symbolic AI. One of the researchers who has elaborated this position most explicitly is Andy Clark https://en.wikipedia.org/wiki/Andy Clark , a philosopher from the School of Cognitive and Computing Sciences of the University of Sussex Brighton, England . Clark defended hybrid partly symbolic, partly connectionist systems. He claimed that at least two kinds of theories are needed in order to study and model cognition. On the one hand, for some information-processing tasks such as pattern recognition connectionism has advantages over symbolic models. But on the other hand, for other cognitive processes such as serial, deductive reasoning, and generative symbol manipulation processes the symbolic paradigm offers adequate models, and not only "approximations" contrary to what radical connectionists would claim . \ 97\ cite note-FOOTNOTEOlazaran1993415–416-98 Gary Marcus https://en.wikipedia.org/wiki/Gary Marcus has claimed that the animus in the deep learning community against symbolic approaches now may be more sociological than philosophical: To think that we can simply abandon symbol-manipulation is to suspend disbelief. And yet, for the most part, that's how most current AI proceeds. Hinton https://en.wikipedia.org/wiki/Geoffrey Hinton and many others have tried hard to banish symbols altogether. The deep learning hope—seemingly grounded not so much in science, but in a sort of historical grudge—is that intelligent behavior will emerge purely from the confluence of massive data and deep learning. Where classical computers and software solve tasks by defining sets of symbol-manipulating rules dedicated to particular jobs, such as editing a line in a word processor or performing a calculation in a spreadsheet, neural networks https://en.wikipedia.org/wiki/Artificial neural network typically try to solve tasks by statistical approximation and learning from examples. According to Marcus, Geoffrey Hinton https://en.wikipedia.org/wiki/Geoffrey Hinton and his colleagues have been vehemently "anti-symbolic": When deep learning reemerged in 2012, it was with a kind of take-no-prisoners attitude that has characterized most of the last decade. By 2015, his hostility toward all things symbols had fully crystallized. He gave a talk at an AI workshop at Stanford comparing symbols to aether <https://en.wikipedia.org/wiki/Aether classical element , one of science's greatest mistakes. ... Since then, his anti-symbolic campaign has only increased in intensity. In 2016, Yann LeCun https://en.wikipedia.org/wiki/Yann LeCun , Bengio https://en.wikipedia.org/wiki/Yoshua Bengio , and Hinton wrote a manifesto for deep learning in one of science's most important journals, Nature. It closed with a direct attack on symbol manipulation, calling not for reconciliation but for outright replacement. Later, Hinton told a gathering of European Union leaders that investing any further money in symbol-manipulating approaches was "a huge mistake," likening it to investing in internal combustion engines in the era of electric cars. \ 98\ cite note-FOOTNOTEMarcus202020-99 Part of these disputes may be due to unclear terminology: Turing award winner Judea Pearl https://en.wikipedia.org/wiki/Judea Pearl offers a critique of machine learning which, unfortunately, conflates the terms machine learning and deep learning. Similarly, when Geoffrey Hinton refers to symbolic AI, the connotation of the term tends to be that of expert systems dispossessed of any ability to learn. The use of the terminology is in need of clarification. Machine learning is not confined to association rule https://en.wikipedia.org/wiki/Association rule learning mining, c.f. the body of work on symbolic ML and relational learning the differences to deep learning being the choice of representation, localist logical rather than distributed, and the non-use of gradient-based learning algorithms https://en.wikipedia.org/wiki/Gradient descent . Equally, symbolic AI is not just about production rules <https://en.wikipedia.org/wiki/Production system computer science written by hand. A proper definition of AI concerns knowledge representation and reasoning https://en.wikipedia.org/wiki/Knowledge representation and reasoning , autonomous multi-agent systems https://en.wikipedia.org/wiki/Multi-agent system , planning and argumentation https://en.wikipedia.org/wiki/Argumentation framework , as well as learning. \ 99\ cite note-FOOTNOTEGarcezLamb20208-100 It is worth noting that, from a theoretical perspective, the boundary of advantages between connectionist AI and symbolic AI may not be as clear-cut as it appears. For instance, Heng Zhang and his colleagues have proved that mainstream knowledge representation formalisms are recursively isomorphic, provided they are universal or have equivalent expressive power.<sup \ 100\ cite note-101 </sup This finding implies that there is no fundamental distinction between using symbolic or connectionist knowledge representation formalisms for the realization of artificial general intelligence https://en.wikipedia.org/wiki/Artificial general intelligence AGI . Moreover, the existence of recursive isomorphisms suggests that different technical approaches can draw insights from one another. From this perspective, it seems unnecessary to overemphasize the advantages of any single technical school; instead, mutual learning and integration may offer the most promising path toward the realization of AGI. Situated robotics: the world as a model edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=34 Another critique of symbolic AI is the embodied cognition https://en.wikipedia.org/wiki/Embodied cognition approach: The embodied cognition https://en.wikipedia.org/wiki/Embodied cognition approach claims that it makes no sense to consider the brain separately: cognition takes place within a body, which is embedded in an environment. We need to study the system as a whole; the brain's functioning exploits regularities in its environment, including the rest of its body. Under the embodied cognition approach, robotics, vision, and other sensors become central, not peripheral. \ 101\ cite note-FOOTNOTERussellNorvig2021982-102 Rodney Brooks https://en.wikipedia.org/wiki/Rodney Brooks invented behavior-based robotics https://en.wikipedia.org/wiki/Behavior-based robotics , one approach to embodied cognition. Nouvelle AI https://en.wikipedia.org/wiki/Nouvelle AI , another name for this approach, is viewed as an alternative to both symbolic AI and connectionist AI. His approach rejected representations, either symbolic or distributed, as not only unnecessary, but as detrimental. Instead, he created the subsumption architecture https://en.wikipedia.org/wiki/Subsumption architecture , a layered architecture for embodied agents. Each layer achieves a different purpose and must function in the real world. For example, the first robot he describes in Intelligence Without Representation , has three layers. The bottom layer interprets sonar sensors to avoid objects. The middle layer causes the robot to wander around when there are no obstacles. The top layer causes the robot to go to more distant places for further exploration. Each layer can temporarily inhibit or suppress a lower-level layer. He criticized AI researchers for defining AI problems for their systems, when: "There is no clean division between perception abstraction and reasoning in the real world."<sup \ 102\ cite note-FOOTNOTEBrooks1991143-103 </sup He called his robots "Creatures" and each layer was "composed of a fixed-topology network of simple finite state machines."<sup \ 103\ cite note-FOOTNOTEBrooks1991151-104 </sup In the Nouvelle AI approach, "First, it is vitally important to test the Creatures we build in the real world; i.e., in the same world that we humans inhabit. It is disastrous to fall into the temptation of testing them in a simplified world first, even with the best intentions of later transferring activity to an unsimplified world."<sup \ 104\ cite note-FOOTNOTEBrooks1991150-105 </sup His emphasis on real-world testing was in contrast to "Early work in AI concentrated on games, geometrical problems, symbolic algebra, theorem proving, and other formal systems"<sup \ 105\ cite note-FOOTNOTEBrooks1991142-106 </sup and the use of the blocks world https://en.wikipedia.org/wiki/Blocks world in symbolic AI systems such as SHRDLU https://en.wikipedia.org/wiki/SHRDLU . Current views edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=35 Each approach—symbolic, connectionist, and behavior-based—has advantages, but has been criticized by the other approaches. Symbolic AI has been criticized as disembodied, liable to the qualification problem, and poor in handling the perceptual problems where deep learning excels. In turn, connectionist AI https://en.wikipedia.org/wiki/Connectionism has been criticized as poorly suited for deliberative step-by-step problem solving, incorporating knowledge, and handling planning. Finally, Nouvelle AI excels in reactive and real-world robotics domains but has been criticized for difficulties in incorporating learning and knowledge. Hybrid AIs https://en.wikipedia.org/wiki/Hybrid intelligent system incorporating one or more of these approaches are currently viewed as the path forward. \ 17\ cite note-FOOTNOTEKautz2020-17 \ 80\ cite note-FOOTNOTERossi2022-81 <sup \ 81\ cite note-FOOTNOTESelman2022-82 </sup Russell and Norvig conclude that: Overall, Dreyfus https://en.wikipedia.org/wiki/Hubert Dreyfus saw areas where AI did not have complete answers and said that Al is therefore impossible; we now see many of these same areas undergoing continued research and development leading to increased capability, not impossibility. \ 101\ cite note-FOOTNOTERussellNorvig2021982-102 See also edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=36 Notes edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=37 1. ↑ cite ref-29 McCarthy once said: "This is AI, so we don't care if it's psychologically real".<sup \ 3\ cite note-FOOTNOTEKolata1982-3 </sup McCarthy reiterated his position in 2006 at the AI@50 https://en.wikipedia.org/wiki/AI@50 conference where he said "Artificial intelligence is not, by definition, simulation of human intelligence".<sup \ 26\ cite note-FOOTNOTEMaker2006-26 </sup Pamela McCorduck https://en.wikipedia.org/wiki/Pamela McCorduck writes that there are "two major branches of artificial intelligence: one aimed at producing intelligent behavior regardless of how it was accomplished, and the other aimed at modeling intelligent processes found in nature, particularly human ones.",<sup \ 27\ cite note-FOOTNOTEMcCorduck2004100–101-27 </sup Stuart Russell https://en.wikipedia.org/wiki/Stuart J. Russell and Peter Norvig https://en.wikipedia.org/wiki/Peter Norvig wrote "Aeronautical engineering texts do not define the goal of their field as making 'machines that fly so exactly like pigeons that they can fool even other pigeons.'"<sup \ 28\ cite note-FOOTNOTERussellNorvig20212-28 </sup Citations edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=38 1. 1 cite ref-:1 1-0 2 cite ref-:1 1-1 Garnelo, Marta; Shanahan, Murray 2019-10-01 . "Reconciling deep learning with symbolic artificial intelligence: representing objects and relations" https://doi.org/10.1016%2Fj.cobeha.2018.12.010 . Current Opinion in Behavioral Sciences . 29 : 17–23. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1016/j.cobeha.2018.12.010 https://doi.org/10.1016%2Fj.cobeha.2018.12.010 . hdl <https://en.wikipedia.org/wiki/Hdl identifier : 10044/1/67796 https://hdl.handle.net/10044%2F1%2F67796 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 72336067 https://api.semanticscholar.org/CorpusID:72336067 . 2. ↑ cite ref-2 Thomason, Richmond February 27, 2024 . "Logic-Based Artificial Intelligence" https://plato.stanford.edu/entries/logic-ai/ . In Zalta, Edward N. https://en.wikipedia.org/wiki/Edward N. Zalta ed. . Stanford Encyclopedia of Philosophy https://en.wikipedia.org/wiki/Stanford Encyclopedia of Philosophy . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 1095-5054 https://search.worldcat.org/issn/1095-5054 . OCLC <https://en.wikipedia.org/wiki/OCLC identifier 429049174 https://search.worldcat.org/oclc/429049174 . 3. 1 cite ref-FOOTNOTEKolata1982 3-0 2 cite ref-FOOTNOTEKolata1982 3-1 Kolata 1982 CITEREFKolata1982 . 4. ↑ cite ref-4 Newell, Allen; Simon, Herbert A. 1976-03-01 . "Computer science as empirical inquiry: symbols and search" https://dl.acm.org/doi/10.1145/360018.360022 . Commun. ACM . 19 3 : 113–126. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1145/360018.360022 https://doi.org/10.1145%2F360018.360022 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 0001-0782 https://search.worldcat.org/issn/0001-0782 . 5. ↑ cite ref-FOOTNOTEKautz2022107–109 5-0 Kautz 2022 CITEREFKautz2022 , pp. 107–109. 6. 1 cite ref-FOOTNOTERussellNorvig202119 6-0 2 cite ref-FOOTNOTERussellNorvig202119 6-1 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 19. 7. 1 cite ref-FOOTNOTERussellNorvig202122–23 7-0 2 cite ref-FOOTNOTERussellNorvig202122–23 7-1 Russell & Norvig 2021 CITEREFRussellNorvig2021 , pp. 22–23. 8. 1 cite ref-FOOTNOTEKautz2022109–110 8-0 2 cite ref-FOOTNOTEKautz2022109–110 8-1 Kautz 2022 CITEREFKautz2022 , pp. 109–110. 9. 1 cite ref-FOOTNOTEKautz2022110 9-0 2 cite ref-FOOTNOTEKautz2022110 9-1 3 cite ref-FOOTNOTEKautz2022110 9-2 Kautz 2022 CITEREFKautz2022 , p. 110. 10. ↑ cite ref-FOOTNOTEKautz2022110–111 10-0 Kautz 2022 CITEREFKautz2022 , pp. 110–111. 11. 1 cite ref-FOOTNOTERussellNorvig202125 11-0 2 cite ref-FOOTNOTERussellNorvig202125 11-1 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 25. 12. ↑ cite ref-FOOTNOTEKautz2022111 12-0 Kautz 2022 CITEREFKautz2022 , p. 111. 13. ↑ cite ref-FOOTNOTEKautz2020110–111 13-0 Kautz 2020 CITEREFKautz2020 , pp. 110–111. 14. ↑ cite ref-14 Rumelhart, David E.; Hinton, Geoffrey E.; Williams, Ronald J. 1986 . "Learning representations by back-propagating errors". Nature . 323 6088 : 533–536. Bibcode <https://en.wikipedia.org/wiki/Bibcode identifier : 1986Natur.323..533R https://ui.adsabs.harvard.edu/abs/1986Natur.323..533R . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1038/323533a0 https://doi.org/10.1038%2F323533a0 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 1476-4687 https://search.worldcat.org/issn/1476-4687 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 205001834 https://api.semanticscholar.org/CorpusID:205001834 . 15. ↑ cite ref-15 LeCun, Y.; Boser, B.; Denker, I.; Henderson, D.; Howard, R.; Hubbard, W.; Tackel, L. 1989 . "Backpropagation Applied to Handwritten Zip Code Recognition". Neural Computation . 1 4 : 541–551. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1162/neco.1989.1.4.541 https://doi.org/10.1162%2Fneco.1989.1.4.541 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 41312633 https://api.semanticscholar.org/CorpusID:41312633 . 16. ↑ cite ref-FOOTNOTEMarcusDavis2019 16-0 Marcus & Davis 2019 CITEREFMarcusDavis2019 . 17. 1 cite ref-FOOTNOTEKautz2020 17-0 2 cite ref-FOOTNOTEKautz2020 17-1 3 cite ref-FOOTNOTEKautz2020 17-2 Kautz 2020 CITEREFKautz2020 . 18. ↑ cite ref-FOOTNOTEKautz2022106 18-0 Kautz 2022 CITEREFKautz2022 , p. 106. 19. ↑ cite ref-FOOTNOTENewellSimon1972 19-0 Newell & Simon 1972 CITEREFNewellSimon1972 . 20. ↑ < cite ref-FOOTNOTEMcCorduck2004139–179, 245–250, 322–323 EPAM 20-0 & McCorduck 2004 CITEREFMcCorduck2004 , pp. 139–179, 245–250, 322–323 EPAM . 21. ↑ cite ref-FOOTNOTECrevier1993145–149 21-0 Crevier 1993 CITEREFCrevier1993 , pp. 145–149. 22. ↑ cite ref-FOOTNOTEMcCorduck2004450–451 22-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 450–451. 23. ↑ cite ref-FOOTNOTECrevier1993258–263 23-0 Crevier 1993 CITEREFCrevier1993 , pp. 258–263. 24. 1 cite ref-FOOTNOTEKautz2022108 24-0 2 cite ref-FOOTNOTEKautz2022108 24-1 Kautz 2022 CITEREFKautz2022 , p. 108. 25. ↑ < cite ref-FOOTNOTERussellNorvig2021p. 9 logicist AI , p. 19 McCarthy's work 25-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 9 logicist AI , p. 19 McCarthy's work . 26. ↑ cite ref-FOOTNOTEMaker2006 26-0 Maker 2006 CITEREFMaker2006 . 27. ↑ cite ref-FOOTNOTEMcCorduck2004100–101 27-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 100–101. 28. ↑ cite ref-FOOTNOTERussellNorvig20212 28-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 2. 29. ↑ cite ref-FOOTNOTEMcCorduck2004251–259 30-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 251–259. 30. ↑ cite ref-FOOTNOTECrevier1993193–196 31-0 Crevier 1993 CITEREFCrevier1993 , pp. 193–196. 31. ↑ cite ref-FOOTNOTEHowe1994 32-0 Howe 1994 CITEREFHowe1994 . 32. ↑ cite ref-FOOTNOTEMcCorduck2004259–305 33-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 259–305. 33. ↑ cite ref-FOOTNOTECrevier199383–102, 163–176 34-0 Crevier 1993 CITEREFCrevier1993 , pp. 83–102, 163–176. 34. ↑ cite ref-FOOTNOTEMcCorduck2004421–424, 486–489 35-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 421–424, 486–489. 35. ↑ cite ref-FOOTNOTECrevier1993168 36-0 Crevier 1993 CITEREFCrevier1993 , p. 168. 36. ↑ cite ref-FOOTNOTEMcCorduck2004489 37-0 McCorduck 2004 CITEREFMcCorduck2004 , p. 489. 37. ↑ cite ref-FOOTNOTECrevier1993239–243 38-0 Crevier 1993 CITEREFCrevier1993 , pp. 239–243. 38. ↑ cite ref-FOOTNOTERussellNorvig2021316, 340 39-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 316, 340. 39. ↑ cite ref-FOOTNOTEKautz2022109 40-0 Kautz 2022 CITEREFKautz2022 , p. 109. 40. ↑ cite ref-FOOTNOTERussellNorvig202122 41-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 22. 41. ↑ cite ref-FOOTNOTEMcCorduck2004266–276, 298–300, 314, 421 42-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 266–276, 298–300, 314, 421. 42. 1 cite ref-Feignebaum Interview 43-0 2 cite ref-Feignebaum Interview 43-1 3 cite ref-Feignebaum Interview 43-2 Shustek, Len 2010 . "An interview with Ed Feigenbaum" https://cacm.acm.org/magazines/2010/6/92472-an-interview-with-ed-feigenbaum/fulltext . Communications of the ACM . 53 6 : 41–45. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1145/1743546.1743564 https://doi.org/10.1145%2F1743546.1743564 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 0001-0782 https://search.worldcat.org/issn/0001-0782 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 10239007 https://api.semanticscholar.org/CorpusID:10239007 . Retrieved 2022-08-05. 43. ↑ cite ref-Knowledge Principle 44-0 Lenat, Douglas B; Feigenbaum, Edward A 1988 . "On the thresholds of knowledge". Proceedings of the International Workshop on Artificial Intelligence for Industrial Applications . pp. 291–300. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1109/AIIA.1988.13308 https://doi.org/10.1109%2FAIIA.1988.13308 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 11778085 https://api.semanticscholar.org/CorpusID:11778085 . 44. ↑ cite ref-FOOTNOTERussellNorvig202122–24 45-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , pp. 22–24. 45. ↑ cite ref-FOOTNOTEMcCorduck2004327–335, 434–435 46-0 McCorduck 2004 CITEREFMcCorduck2004 , pp. 327–335, 434–435. 46. ↑ cite ref-FOOTNOTECrevier1993145–62, 197–203 47-0 Crevier 1993 CITEREFCrevier1993 , pp. 145–62, 197–203. 47. 1 cite ref-FOOTNOTERussellNorvig202123 48-0 2 cite ref-FOOTNOTERussellNorvig202123 48-1 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 23. 48. 1 cite ref-FOOTNOTEClancey1987 49-0 2 cite ref-FOOTNOTEClancey1987 49-1 Clancey 1987 CITEREFClancey1987 . 49. ↑ cite ref-50 "The fascination with AI: what is artificial intelligence?" https://www.ionos.com/digitalguide/online-marketing/online-sales/what-is-artificial-intelligence/ . IONOS Digitalguide . Retrieved 2021-12-02. 50. ↑ cite ref-FOOTNOTEHayes-RothMurrayAdelman2015 51-0 Hayes-Roth, Murray & Adelman 2015 CITEREFHayes-RothMurrayAdelman2015 . 51. ↑ cite ref-BB1 52-0 Hayes-Roth, Barbara 1985 . "A blackboard architecture for control". Artificial Intelligence . 26 3 : 251–321. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1016/0004-3702 85 90063-3 https://doi.org/10.1016%2F0004-3702%2885%2990063-3 . 52. ↑ cite ref-OPM 53-0 Hayes-Roth, Barbara 1980 . Human Planning Processes . RAND. 53. ↑ cite ref-FOOTNOTEPearl1988 54-0 Pearl 1988 CITEREFPearl1988 . 54. ↑ cite ref-FOOTNOTESpiegelhalterDawidLauritzenCowell1993 55-0 Spiegelhalter et al. 1993 CITEREFSpiegelhalterDawidLauritzenCowell1993 . 55. ↑ cite ref-FOOTNOTERussellNorvig2021335–337 56-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , pp. 335–337. 56. ↑ cite ref-FOOTNOTERussellNorvig2021459 57-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 459. 57. ↑ cite ref-58 Quinlan, J. Ross. "Chapter 15: Learning Efficient Classification Procedures and their Application to Chess End Games". In Michalski, Carbonell & Mitchell 1983 CITEREFMichalskiCarbonellMitchell1983 . 58. ↑ cite ref-59 Quinlan, J. Ross 1992-10-15 . C4.5: Programs for Machine Learning 1st ed. . San Mateo, Calif: Morgan Kaufmann. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-1-55860-238-0 https://en.wikipedia.org/wiki/Special:BookSources/978-1-55860-238-0 . 59. ↑ cite ref-60 Mitchell, Tom M.; Utgoff, Paul E.; Banerji, Ranan. "Chapter 6: Learning by Experimentation: Acquiring and Refining Problem-Solving Heuristics". In Michalski, Carbonell & Mitchell 1983 CITEREFMichalskiCarbonellMitchell1983 . 60. ↑ cite ref-61 Valiant, L. G. 1984-11-05 . "A theory of the learnable" https://doi.org/10.1145%2F1968.1972 . Communications of the ACM . 27 11 : 1134–1142. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1145/1968.1972 https://doi.org/10.1145%2F1968.1972 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 0001-0782 https://search.worldcat.org/issn/0001-0782 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 12837541 https://api.semanticscholar.org/CorpusID:12837541 . 61. ↑ cite ref-62 Koedinger, K. R.; Anderson, J. R.; Hadley, W. H.; Mark, M. A.; others 1997 . "Intelligent tutoring goes to school in the big city" http://telearn.archives-ouvertes.fr/hal-00197383/ . International Journal of Artificial Intelligence in Education . 8 : 30–43. Retrieved 2012-08-18. 62. ↑ cite ref-63 Shapiro, Ehud Y 1981 . "The Model Inference System". Proceedings of the 7th international joint conference on Artificial intelligence . IJCAI. Vol. 2. p. 1064. 63. ↑ cite ref-64 Manna, Zohar; Waldinger, Richard 1980-01-01 . "A Deductive Approach to Program Synthesis". ACM Trans. Program. Lang. Syst . 2 1 : 90–121. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1145/357084.357090 https://doi.org/10.1145%2F357084.357090 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 14770735 https://api.semanticscholar.org/CorpusID:14770735 . 64. ↑ cite ref-Schank 65-0 Schank, Roger C. 1983-01-28 . Dynamic Memory: A Theory of Reminding and Learning in Computers and People . Cambridge Cambridgeshire : New York: Cambridge University Press. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-0-521-27029-8 https://en.wikipedia.org/wiki/Special:BookSources/978-0-521-27029-8 . 65. ↑ cite ref-66 Hammond, Kristian J. 1989-04-11 . Case-Based Planning: Viewing Planning as a Memory Task . Boston: Academic Press. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-0-12-322060-8 https://en.wikipedia.org/wiki/Special:BookSources/978-0-12-322060-8 . 66. ↑ cite ref-67 Koza, John R. 1992-12-11 . Genetic Programming: On the Programming of Computers by Means of Natural Selection 1st ed. . Cambridge, Mass: A Bradford Book. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-0-262-11170-6 https://en.wikipedia.org/wiki/Special:BookSources/978-0-262-11170-6 . 67. ↑ cite ref-68 Mostow, David Jack. "Chapter 12: Machine Transformation of Advice into a Heuristic Search Procedure". In Michalski, Carbonell & Mitchell 1983 CITEREFMichalskiCarbonellMitchell1983 . 68. ↑ cite ref-69 Bareiss, Ray; Porter, Bruce; Wier, Craig. "Chapter 4: Protos: An Exemplar-Based Learning Apprentice". In Michalski, Carbonell & Mitchell 1986 CITEREFMichalskiCarbonellMitchell1986 , pp. 112–139. 69. ↑ cite ref-70 Carbonell, Jaime. "Chapter 5: Learning by Analogy: Formulating and Generalizing Plans from Past Experience". In Michalski, Carbonell & Mitchell 1983 CITEREFMichalskiCarbonellMitchell1983 , pp. 137–162. 70. ↑ cite ref-71 Carbonell, Jaime. "Chapter 14: Derivational Analogy: A Theory of Reconstructive Problem Solving and Expertise Acquisition". In Michalski, Carbonell & Mitchell 1986 CITEREFMichalskiCarbonellMitchell1986 , pp. 371–392. 71. ↑ cite ref-72 Mitchell, Tom; Mabadevan, Sridbar; Steinberg, Louis. "Chapter 10: LEAP: A Learning Apprentice for VLSI Design". In Kodratoff & Michalski 1990 CITEREFKodratoffMichalski1990 , pp. 271–289. 72. ↑ cite ref-73 Lenat, Douglas. "Chapter 9: The Role of Heuristics in Learning by Discovery: Three Case Studies". In Michalski, Carbonell & Mitchell 1983 CITEREFMichalskiCarbonellMitchell1983 , pp. 243–306. 73. ↑ cite ref-74 Korf, Richard E. 1985 . Learning to Solve Problems by Searching for Macro-Operators . Research Notes in Artificial Intelligence. Pitman Publishing. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 0-273-08690-1 https://en.wikipedia.org/wiki/Special:BookSources/0-273-08690-1 . 74. ↑ cite ref-Rossi 75-0 Rossi, Francesca. "Thinking Fast and Slow in AI" https://aaai-2022.virtualchair.net/plenary 13.html . AAAI. Retrieved 5 July 2022. 75. ↑ cite ref-Selman 76-0 Selman, Bart. "AAAI Presidential Address: The State of AI" https://aaai-2022.virtualchair.net/plenary 2.html . AAAI. Retrieved 5 July 2022. 76. ↑ cite ref-FOOTNOTEValiant2008 77-0 Valiant 2008 CITEREFValiant2008 . 77. 1 cite ref-FOOTNOTEGarcezBesoldDe RaedtFöldiák2015 78-0 2 cite ref-FOOTNOTEGarcezBesoldDe RaedtFöldiák2015 78-1 Garcez et al. 2015 CITEREFGarcezBesoldDe RaedtFöldiák2015 . 78. ↑ cite ref-FOOTNOTEMarcus202044 79-0 Marcus 2020 CITEREFMarcus2020 , p. 44. 79. ↑ cite ref-FOOTNOTEMarcus202017 80-0 Marcus 2020 CITEREFMarcus2020 , p. 17. 80. 1 cite ref-FOOTNOTERossi2022 81-0 2 cite ref-FOOTNOTERossi2022 81-1 Rossi 2022 CITEREFRossi2022 . 81. 1 cite ref-FOOTNOTESelman2022 82-0 2 cite ref-FOOTNOTESelman2022 82-1 Selman 2022 CITEREFSelman2022 . 82. ↑ cite ref-FOOTNOTEGarcezLamb20202 83-0 Garcez & Lamb 2020 CITEREFGarcezLamb2020 , p. 2. 83. ↑ cite ref-FOOTNOTEGarcezBrodaGabbayGabbay2002 84-0 Garcez et al. 2002 CITEREFGarcezBrodaGabbayGabbay2002 . 84. ↑ cite ref-85 "Neurosymbolic AI NeSy Association website" http://www.neural-symbolic.org . www.neural-symbolic.org . Retrieved 2026-08-09. 85. ↑ cite ref-86 Rocktäschel, Tim; Riedel, Sebastian 2016 . "Learning Knowledge Base Inference with Neural Theorem Provers" https://aclanthology.org/W16-1309 . Proceedings of the 5th Workshop on Automated Knowledge Base Construction . San Diego, CA: Association for Computational Linguistics. pp. 45–50. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.18653/v1/W16-1309 https://doi.org/10.18653%2Fv1%2FW16-1309 . Retrieved 2022-08-06. 86. ↑ cite ref-87 Serafini, Luciano; Garcez, Artur d'Avila 2016 , Logic Tensor Networks: Deep Learning and Logical Reasoning from Data and Knowledge , arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 1606.04422 https://arxiv.org/abs/1606.04422 87. 1 cite ref-:0 88-0 2 cite ref-:0 88-1 Garcez, Artur d'Avila; Lamb, Luis C.; Gabbay, Dov M. 2009 . Neural-Symbolic Cognitive Reasoning https://link.springer.com/book/10.1007/978-3-540-73246-4 1st ed. . Berlin-Heidelberg: Springer. Bibcode <https://en.wikipedia.org/wiki/Bibcode identifier : 2009nscr.book.....D https://ui.adsabs.harvard.edu/abs/2009nscr.book.....D . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1007/978-3-540-73246-4 https://doi.org/10.1007%2F978-3-540-73246-4 . ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-3-540-73245-7 https://en.wikipedia.org/wiki/Special:BookSources/978-3-540-73245-7 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 14002173 https://api.semanticscholar.org/CorpusID:14002173 . 88. ↑ cite ref-meta-object protocol 89-0 Kiczales, Gregor; Rivieres, Jim des; Bobrow, Daniel G. 1991-07-30 . The Art of the Metaobject Protocol 1st ed. . Cambridge, Mass: The MIT Press. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-0-262-61074-2 https://en.wikipedia.org/wiki/Special:BookSources/978-0-262-61074-2 . 89. ↑ cite ref-HermiT 90-0 Motik, Boris; Shearer, Rob; Horrocks, Ian 2009-10-28 . "Hypertableau Reasoning for Description Logics" https://doi.org/10.1613%2Fjair.2811 . Journal of Artificial Intelligence Research . 36 : 165–228. arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 1401.3485 https://arxiv.org/abs/1401.3485 . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1613/jair.2811 https://doi.org/10.1613%2Fjair.2811 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 1076-9757 https://search.worldcat.org/issn/1076-9757 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 190609 https://api.semanticscholar.org/CorpusID:190609 . 90. ↑ cite ref-QSIM 91-0 Kuipers, Benjamin 1994 . Qualitative Reasoning: Modeling and Simulation with Incomplete Knowledge . MIT Press. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-0-262-51540-5 https://en.wikipedia.org/wiki/Special:BookSources/978-0-262-51540-5 . 91. ↑ cite ref-FOOTNOTERussellNorvig2021 92-0 Russell & Norvig 2021 CITEREFRussellNorvig2021 . 92. ↑ cite ref-93 Leo de Penning, Artur S. d'Avila Garcez, Luís C. Lamb, John-Jules Ch. Meyer: "A Neural-Symbolic Cognitive Agent for Online Learning and Reasoning." IJCAI 2011: 1653-1658 93. ↑ cite ref-FOOTNOTEMcCarthyHayes1969 94-0 McCarthy & Hayes 1969 CITEREFMcCarthyHayes1969 . 94. ↑ cite ref-FOOTNOTEMcCarthy1959 95-0 McCarthy 1959 CITEREFMcCarthy1959 . 95. ↑ cite ref-FOOTNOTENilsson19987 96-0 Nilsson 1998 CITEREFNilsson1998 , p. 7. 96. ↑ cite ref-FOOTNOTEOlazaran1993411–416 97-0 Olazaran 1993 CITEREFOlazaran1993 , pp. 411–416. 97. ↑ cite ref-FOOTNOTEOlazaran1993415–416 98-0 Olazaran 1993 CITEREFOlazaran1993 , pp. 415–416. 98. ↑ cite ref-FOOTNOTEMarcus202020 99-0 Marcus 2020 CITEREFMarcus2020 , p. 20. 99. ↑ cite ref-FOOTNOTEGarcezLamb20208 100-0 Garcez & Lamb 2020 CITEREFGarcezLamb2020 , p. 8. 100. ↑ cite ref-101 Zhang, Heng; Jiang, Guifei; Quan, Donghui 2025-04-11 . "A Theory of Formalisms for Representing Knowledge" https://ojs.aaai.org/index.php/AAAI/article/view/33674 . Proceedings of the AAAI Conference on Artificial Intelligence . 39 14 : 15257–15264. arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 2412.11855 https://arxiv.org/abs/2412.11855 . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1609/aaai.v39i14.33674 https://doi.org/10.1609%2Faaai.v39i14.33674 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 2374-3468 https://search.worldcat.org/issn/2374-3468 . 101. 1 cite ref-FOOTNOTERussellNorvig2021982 102-0 2 cite ref-FOOTNOTERussellNorvig2021982 102-1 Russell & Norvig 2021 CITEREFRussellNorvig2021 , p. 982. 102. ↑ cite ref-FOOTNOTEBrooks1991143 103-0 Brooks 1991 CITEREFBrooks1991 , p. 143. 103. ↑ cite ref-FOOTNOTEBrooks1991151 104-0 Brooks 1991 CITEREFBrooks1991 , p. 151. 104. ↑ cite ref-FOOTNOTEBrooks1991150 105-0 Brooks 1991 CITEREFBrooks1991 , p. 150. 105. ↑ cite ref-FOOTNOTEBrooks1991142 106-0 Brooks 1991 CITEREFBrooks1991 , p. 142. References edit https://en.wikipedia.org/w/index.php?title=Symbolic artificial intelligence&action=edit§ion=39 - Brooks, Rodney A. 1991 . "Intelligence without representation" https://dx.doi.org/10.1016/0004-3702%2891%2990053-M . Artificial Intelligence . 47 1 : 139–159. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1016/0004-3702 91 90053-M https://doi.org/10.1016%2F0004-3702%2891%2990053-M . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 0004-3702 https://search.worldcat.org/issn/0004-3702 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 207507849 https://api.semanticscholar.org/CorpusID:207507849 . Retrieved 2022-09-13. - Clancey, William 1987 . Knowledge-Based Tutoring: The GUIDON Program MIT Press Series in Artificial Intelligence Hardcover ed. . - Crevier, Daniel https://en.wikipedia.org/wiki/Daniel Crevier 1993 . AI: The Tumultuous Search for Artificial Intelligence . New York, NY: BasicBooks. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 0-465-02997-3 https://en.wikipedia.org/wiki/Special:BookSources/0-465-02997-3 .. - Dreyfus, Hubert L 1981 . "From micro-worlds to knowledge representation: AI at an impasse" http://www.cogsci.ucsd.edu/~coulson/203/dreyfus.pdf PDF . Mind Design . MIT Press, Cambridge, MA: 161–204. - Garcez, Artur S. d'Avila; Broda, Krysia; Gabbay, Dov M.; Gabbay, Augustus de Morgan Professor of Logic Dov M. 2002 . Neural-Symbolic Learning Systems: Foundations and Applications . Springer Science & Business Media. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-1-85233-512-0 https://en.wikipedia.org/wiki/Special:BookSources/978-1-85233-512-0 . - Garcez, Artur; Besold, Tarek; De Raedt, Luc; Földiák, Peter; Hitzler, Pascal; Icard, Thomas; Kühnberger, Kai-Uwe; Lamb, Luís; Miikkulainen, Risto; Silver, Daniel 2015 . Neural-Symbolic Learning and Reasoning: Contributions and Challenges . AAI Spring Symposium - Knowledge Representation and Reasoning: Integrating Symbolic and Neural Approaches. Stanford, CA: AAAI Press. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.13140/2.1.1779.4243 https://doi.org/10.13140%2F2.1.1779.4243 . - Garcez, Artur d'Avila; Gori, Marco; Lamb, Luis C.; Serafini, Luciano; Spranger, Michael; Tran, Son N. 2019 , Neural-Symbolic Computing: An Effective Methodology for Principled Integration of Machine Learning and Reasoning , arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 1905.06088 https://arxiv.org/abs/1905.06088 - Garcez, Artur d'Avila; Lamb, Luis C. 2020 , Neurosymbolic AI: The 3rd Wave , arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 2012.05876 https://arxiv.org/abs/2012.05876 - Haugeland, John https://en.wikipedia.org/wiki/John Haugeland 1985 , Artificial Intelligence: The Very Idea , Cambridge, Mass: MIT Press, ISBN <https://en.wikipedia.org/wiki/ISBN identifier 0-262-08153-9 https://en.wikipedia.org/wiki/Special:BookSources/0-262-08153-9 - Hayes-Roth, Frederick; Murray, William; Adelman, Leonard 2015 . "Expert systems". AccessScience . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1036/1097-8542.248550 https://doi.org/10.1036%2F1097-8542.248550 . - Honavar, Vasant; Uhr, Leonard 1994 . Symbolic Artificial Intelligence, Connectionist Networks & Beyond https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=1072&context=cs techreports Technical report . Iowa State University Digital Repository, Computer Science Technical Reports. 76. p. 6. - Honavar, Vasant 1995 . Symbolic Artificial Intelligence and Numeric Artificial Neural Networks: Towards a Resolution of the Dichotomy . The Springer International Series In Engineering and Computer Science. Springer US. pp. 351–388. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1007/978-0-585-29599-2 11 https://doi.org/10.1007%2F978-0-585-29599-2 11 . - Howe, J. November 1994 . "Artificial Intelligence at Edinburgh University: a Perspective" http://www.inf.ed.ac.uk/about/AIhistory.html . Archived https://web.archive.org/web/20070515072641/http://www.inf.ed.ac.uk/about/AIhistory.html from the original on 15 May 2007. Retrieved 30 August 2007. - Kautz, Henry 2020-02-11 . The Third AI Summer, Henry Kautz, AAAI 2020 Robert S. Engelmore Memorial Award Lecture https://www.youtube.com/watch?v= cQITY0SPiw . Retrieved 2022-07-06. - Kautz, Henry 2022 . "The Third AI Summer: AAAI Robert S. Engelmore Memorial Lecture" https://ojs.aaai.org/index.php/aimagazine/article/view/19122 . AI Magazine . 43 1 : 93–104. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1609/aimag.v43i1.19122 https://doi.org/10.1609%2Faimag.v43i1.19122 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 2371-9621 https://search.worldcat.org/issn/2371-9621 . S2CID <https://en.wikipedia.org/wiki/S2CID identifier 248213051 https://api.semanticscholar.org/CorpusID:248213051 . Retrieved 2022-07-12. - Kodratoff, Yves; Michalski, Ryszard, eds. 1990 . Machine Learning : an Artificial Intelligence Approach . Vol. III. San Mateo, Calif.: Morgan Kaufman. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 0-934613-09-5 https://en.wikipedia.org/wiki/Special:BookSources/0-934613-09-5 . OCLC <https://en.wikipedia.org/wiki/OCLC identifier 893488404 https://search.worldcat.org/oclc/893488404 . - Kolata, G. 1982 . "How can computers get common sense?". Science . 217 4566 : 1237–1238. Bibcode <https://en.wikipedia.org/wiki/Bibcode identifier : 1982Sci...217.1237K https://ui.adsabs.harvard.edu/abs/1982Sci...217.1237K . doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1126/science.217.4566.1237 https://doi.org/10.1126%2Fscience.217.4566.1237 . PMID <https://en.wikipedia.org/wiki/PMID identifier 17837639 https://pubmed.ncbi.nlm.nih.gov/17837639 . - Maker, Meg Houston 2006 . "AI@50: AI Past, Present, Future" https://web.archive.org/web/20070103222615/http://www.engagingexperience.com/2006/07/ai50 ai past pr.html . Dartmouth College. Archived from the original http://www.engagingexperience.com/2006/07/ai50 ai past pr.html on 3 January 2007. Retrieved 16 October 2008. - Marcus, Gary https://en.wikipedia.org/wiki/Gary Marcus ; Davis, Ernest 2019 . Rebooting AI: Building Artificial Intelligence We Can Trust . New York: Pantheon Books https://en.wikipedia.org/wiki/Pantheon Books . ISBN <https://en.wikipedia.org/wiki/ISBN identifier 9781524748258 https://en.wikipedia.org/wiki/Special:BookSources/9781524748258 . OCLC <https://en.wikipedia.org/wiki/OCLC identifier 1083223029 https://search.worldcat.org/oclc/1083223029 . - Marcus, Gary https://en.wikipedia.org/wiki/Gary Marcus 2020 , The Next Decade in AI: Four Steps Towards Robust Artificial Intelligence , arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 2002.06177 https://arxiv.org/abs/2002.06177 - McCarthy, John <https://en.wikipedia.org/wiki/John McCarthy computer scientist 1959 . PROGRAMS WITH COMMON SENSE . Symposium on Mechanization of Thought Processes. NATIONAL PHYSICAL LABORATORY, TEDDINGTON, UK. p. 8. - McCarthy, John <https://en.wikipedia.org/wiki/John McCarthy computer scientist ; Hayes, Patrick https://en.wikipedia.org/wiki/Pat Hayes 1969 . "Some Philosophical Problems From the Standpoint of Artificial Intelligence". Machine Intelligence 4 . B. Meltzer, Donald Michie eds. : 463–502. - McCorduck, Pamela https://en.wikipedia.org/wiki/Pamela McCorduck 2004 , Machines Who Think 2nd ed. , Natick, Massachusetts: A. K. Peters, ISBN <https://en.wikipedia.org/wiki/ISBN identifier 1-5688-1205-1 https://en.wikipedia.org/wiki/Special:BookSources/1-5688-1205-1 - Michalski, Ryszard; Carbonell, Jaime; Mitchell, Tom, eds. 1983 . Machine Learning : an Artificial Intelligence Approach . Vol. I. Palo Alto, Calif.: Tioga Publishing Company. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 0-935382-05-4 https://en.wikipedia.org/wiki/Special:BookSources/0-935382-05-4 . OCLC <https://en.wikipedia.org/wiki/OCLC identifier 9262069 https://search.worldcat.org/oclc/9262069 . - Michalski, Ryszard; Carbonell, Jaime; Mitchell, Tom, eds. 1986 . Machine Learning : an Artificial Intelligence Approach . Vol. II. Los Altos, Calif.: Morgan Kaufman. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 0-934613-00-1 https://en.wikipedia.org/wiki/Special:BookSources/0-934613-00-1 . - Newell, Allen; Simon, Herbert A. 1972 . Human Problem Solving 1st ed. . Englewood Cliffs, New Jersey: Prentice Hall. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 0-13-445403-0 https://en.wikipedia.org/wiki/Special:BookSources/0-13-445403-0 . - Newell, Allen https://en.wikipedia.org/wiki/Allen Newell ; Simon, H. A. https://en.wikipedia.org/wiki/Herbert A. Simon 1976 . "Computer Science as Empirical Inquiry: Symbols and Search" https://doi.org/10.1145%2F360018.360022 . Communications of the ACM . 19 3 : 113–126. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1145/360018.360022 https://doi.org/10.1145%2F360018.360022 . - Nilsson, Nils <https://en.wikipedia.org/wiki/Nils Nilsson researcher 1998 . Artificial Intelligence: A New Synthesis https://archive.org/details/artificialintell0000nils . Morgan Kaufmann. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-1-55860-467-4 https://en.wikipedia.org/wiki/Special:BookSources/978-1-55860-467-4 . Archived https://web.archive.org/web/20200726131654/https://archive.org/details/artificialintell0000nils from the original on 26 July 2020. Retrieved 18 November 2019. - Olazaran, Mikel 1993-01-01 , "A Sociological History of the Neural Network Controversy" https://www.sciencedirect.com/science/article/pii/S0065245808604088 , in Yovits, Marshall C. ed. , Advances in Computers Volume 37 , vol. 37, Elsevier, pp. 335–425, doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1016/S0065-2458 08 60408-8 https://doi.org/10.1016%2FS0065-2458%2808%2960408-8 , ISBN <https://en.wikipedia.org/wiki/ISBN identifier 9780120121373 https://en.wikipedia.org/wiki/Special:BookSources/9780120121373 , retrieved 2023-10-31 - Pearl, J. 1988 . Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference . San Mateo, California: Morgan Kaufmann. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-1-55860-479-7 https://en.wikipedia.org/wiki/Special:BookSources/978-1-55860-479-7 . OCLC <https://en.wikipedia.org/wiki/OCLC identifier 249625842 https://search.worldcat.org/oclc/249625842 . - Russell, Stuart J. https://en.wikipedia.org/wiki/Stuart J. Russell ; Norvig, Peter https://en.wikipedia.org/wiki/Peter Norvig 2021 . Artificial Intelligence: A Modern Approach https://en.wikipedia.org/wiki/Artificial Intelligence: A Modern Approach 4th ed. . Hoboken: Pearson. ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-0-13-461099-3 https://en.wikipedia.org/wiki/Special:BookSources/978-0-13-461099-3 . LCCN <https://en.wikipedia.org/wiki/LCCN identifier 20190474 https://lccn.loc.gov/20190474 . - Rossi, Francesca 2022-07-06 . "AAAI2022: Thinking Fast and Slow in AI AAAI 2022 Invited Talk " https://aaai-2022.virtualchair.net/plenary 13.html . Retrieved 2022-07-06. - Selman, Bart 2022-07-06 . "AAAI2022: Presidential Address: The State of AI" https://aaai-2022.virtualchair.net/plenary 2.html . Retrieved 2022-07-06. - Serafini, Luciano; Garcez, Artur d'Avila 2016-07-07 , Logic Tensor Networks: Deep Learning and Logical Reasoning from Data and Knowledge , arXiv <https://en.wikipedia.org/wiki/ArXiv identifier : 1606.04422 https://arxiv.org/abs/1606.04422 - Spiegelhalter, David J. https://en.wikipedia.org/wiki/David Spiegelhalter ; Dawid, A. Philip; Lauritzen, Steffen https://en.wikipedia.org/wiki/Steffen Lauritzen ; Cowell, Robert G. 1993 . "Bayesian analysis in expert systems". Statistical Science . 8 3 . - Turing, A. M. 1950 . "I.—Computing Machinery and Intelligence" https://doi.org/10.1093/mind/LIX.236.433 . Mind . LIX 236 : 433–460. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1093/mind/LIX.236.433 https://doi.org/10.1093%2Fmind%2FLIX.236.433 . ISSN <https://en.wikipedia.org/wiki/ISSN identifier 0026-4423 https://search.worldcat.org/issn/0026-4423 . Retrieved 2022-09-14. - Valiant, Leslie G 2008 . "Knowledge Infusion: In Pursuit of Robustness in Artificial Intelligence". In Hariharan, R.; Mukund, M.; Vinay, V. eds. . Foundations of Software Technology and Theoretical Computer Science Bangalore . pp. 415–422. - Xifan Yao; Jiajun Zhou; Jiangming Zhang; Claudio R. Boer 2017 . "From Intelligent Manufacturing to Smart Manufacturing for Industry 4.0 Driven by Next Generation Artificial Intelligence and Further on". 2017 5th International Conference on Enterprise Systems ES . IEEE. pp. 311–318. doi <https://en.wikipedia.org/wiki/Doi identifier : 10.1109/es.2017.58 https://doi.org/10.1109%2Fes.2017.58 . ISBN <https://en.wikipedia.org/wiki/ISBN identifier 978-1-5386-0936-1 https://en.wikipedia.org/wiki/Special:BookSources/978-1-5386-0936-1 .