Overview of CYC
Abstract
The CYC project is the first serious attempt to build a base of human consensus knowledge -- to encode common sense. The CYC system is intended to provide a "deep" layer of understanding that can eventually be used by other programs to make them more flexible and less "brittle".
This paper discusses of the motivations behind the encoding of common sense, provides an overview of the CYC system and touches on some of its applications.
Introduction
Often it is enough is to solve problems in a very specific domain. Expert systems are often called upon when this type of intelligence is desired. Knowledge based systems have enjoyed commercial success and have shown promise as candidates for practical intelligent systems of the near future [1].
Traditional knowledge based systems -- and programs in general -- have an undesirable performance characteristic that can be described as brittleness. The term -- coined by Doug Lenat -- refers to software, which when given well conditioned data, will produce acceptable results, but when confronted by some unanticipated situation, is likely to reach the wrong conclusion. This brittleness arises because knowledge based systems lack the common sense that all humans possess. Lenat laments that
"It is all too easy to find examples of such brittle behavior: a skin disease diagnosis system being told about a rusty old car, and concluding it has measles; a car loan authorization system approving a loan from someone whose ``years at the same job'' exceeded their age; a digitalis dosage system that doesn't complain when someone accidentally types a patient's age and weight in reverse order (even though this 49 pound, 102 year old patient was admitted to the h...
... middle of paper ...
...ady self-aware. If you ask it what it is, it knows that it is a computer. If you ask who we are, it knows that we are users. It knows that it is running on a certain machine at a certain place in a certain time. It knows who is talking to it. It knows that a conversation or a run of an application is happening. It has the same kind of time sense that you and I do." [4]
Bibliography
[1] Guha, R.V and Lenat, D.B. Pittman K., Pratt, D. Shepherd M. Cyc: Toward Programs With Common Sense. MCC Technical report archive 1990.
[2] www.cyc.com
[3] Guha, R.V and Lenat, D.B. Pittman K., Pratt, D. Shepherd M. CYC: a midterm report. Communications of the ACM. August 1990/ Vol 33. No 8.
[4] Garfinkel, S. 2001 Double Take. Wired Magazine. January 1997/5.01.
[5] Guha, R.V and Lenat, D.B. Ideas for Applying CYC. MCC Technical report archive 1991.
[6] www.mcc.com
Due to the unique nature of the intelligence field, error of judgments can (and has) had catastrophic consequences. These errors are a result of complex decision making processes involved in the generation of intelligence products, affected by not only training and expertise, but by cognitive factors, particularly bias. The aim of this paper is to identify two different models of decision making (bounded rationality and intuitive decision making), the biases found in both models that affect the final intelligence product, and how these biases can be mitigated in order to avoid intelligence failures or minimise their impact.
Since such arbitrary factors influence and individual’s reasoning it should not be thought that the human judgement is not trustworthy. Therefore, a human’s rationality lies within the trustworthiness of the judgement and since it cannot be trusted humans are irrational. While the thought that humans ought to be able to recognize this shortcoming exists, it does not matter because the individual functions on a dysfunctional system. This idea is known as the shoddy software hypothesis, which focuses on the idea that the average mind works with certain heuristics, but not with others. The mind functions in a certain way to produce rational decisions, however it is extremely limited with this hypothesis. This is because the mind can only function rationally in certain conditions rather than all conditions showing that humans are fallible and irrational. Humans make errors systematically and repeatedly because of the shoddy software. This shoddy software is like working on a broken computer, a functioning computer can take one form of input and produce the right answer, however as the
If a machine passes the test, then it is clear that for many ordinary people it would be a sufficient reason to say that that is a thinking machine. And, in fact, since it is able to conversate with a human and to actually fool him and convince him that the machine is human, this would seem t...
The case based reasoning system proposed here mimics the human decision making process by learning from previous experience and using the knowledge to solve current problem. This system will utilize previous adverse episodes and their solutions to prevent reoccurrences, and also to detect the oc...
As our world expands through the growing abilities and applications of computers in our everyday lives, it seems that the role of the computer has been reversed. Before we knew that the computer only understood what we programmed it to understand; however, now the majority of our society is learning more from computers than they are able to input into it. Dumm (1986 p.69)
CYC is a very large, multi-contextual knowledge base and inference engine. The development of CYC was started at the Microelectronics and Computer Technology Corporation(MCC) during the early 1980s and continued at Cycorp, Inc. On January 1, 1995 at Austin, Texas. Doug Lenat, the former head of the CYC Project at MCC and the president of Cycorp at present, has lead the development of the CYC project from the beginning. The goal of the Cyc project is to break the software brittleness bottleneck once and for all by constructing a foundation of basic common sense knowledge system and semantic substratum of terms, rules, and relations that will enable a variety of knowledge-intensive products and services. T...
Imagine asking your computer to do something in the same way you would ask a friend to do it. Without having to memorize special commands that only it could understand. For computer scientists this has been an ambitious goal; that can further simplify computers. Artificial Intelligence, a system that can mimic human intelligence by performing task that usually only a human can do, usually has to use a form of natural language processing. Natural language processing, a sub-field of computer science and artificial intelligence, concerns the successfully interaction between a computer and a human. Currently one of the best examples of A.I.(Artificial Intelligence) is IBM 's Watson. A machine that gained popularity after appearing on the show
Finally almost a decade after the Dartmouth Conference, Centers for artificial intelligence research began to form at Carnegie Mellon and MIT. Further advancements were made in the field. The General Problem Solver (GPS) was developed based on the Wiener's feedback principle. The GPS was capable of solving a greater range of common sense problems.
Mayfield, James; Finin, Tim; Narayanaswamy, Rajkumar; Shah, Chetan; MacCartney, William; Goolsbey, Keith, "The Cycic Friends Network: getting Cyc agents to reason together", University Of Maryland - Baltimore County, http://www.cs.umbc.edu/~cikm/iia/submitted/viewing/mayfield
In order to see how artificial intelligence plays a role on today’s society, I believe it is important to dispel any misconceptions about what artificial intelligence is. Artificial intelligence has been defined many different ways, but the commonality between all of them is that artificial intelligence theory and development of computer systems that are able to perform tasks that would normally require a human intelligence such as decision making, visual recognition, or speech recognition. However, human intelligence is a very ambiguous term. I believe there are three main attributes an artificial intelligence system has that makes it representative of human intelligence (Source 1). The first is problem solving, the ability to look ahead several steps in the decision making process and being able to choose the best solution (Source 1). The second is the representation of knowledge (Source 1). While knowledge is usually gained through experience or education, intelligent agents could very well possibly have a different form of knowledge. Access to the internet, the la...
Whether or not computers will be conscious or spiritual in the future, the fact remains that they are increasingly able to accomplish tasks which were thought only achievable by humans, such as playing chess, or even reading printed text aloud to assist the blind. Stanford computer scientist John McCarthy believes that the only reason computers cannot do some tasks as well as humans is that we do not have an actual understanding of how we solve the problems ourselves. He writes that, “whenever people do better than computers on some task or computers use a lot of computation to do as well as people, this demonstrates that the program designers lack understanding of the intellectual mechanisms required to do the task efficiently” (McCarthy 2004). This lack of understanding of underlying mechanisms is apparent in the situation when experts in a field cannot fully explain how they accomplish a complex task, such as playing the saxophone or swinging a golf club, which they may only fully understand on a subconscious level.
Moran, L. A., Horton, H. R., Scrimgeour, K. G., & Perry, M. D. (2012) Principles of
The traditional notion that seeks to compare human minds, with all its intricacies and biochemical functions, to that of artificially programmed digital computers, is self-defeating and it should be discredited in dialogs regarding the theory of artificial intelligence. This traditional notion is akin to comparing, in crude terms, cars and aeroplanes or ice cream and cream cheese. Human mental states are caused by various behaviours of elements in the brain, and these behaviours in are adjudged by the biochemical composition of our brains, which are responsible for our thoughts and functions. When we discuss mental states of systems it is important to distinguish between human brains and that of any natural or artificial organisms which is said to have central processing systems (i.e. brains of chimpanzees, microchips etc.). Although various similarities may exist between those systems in terms of functions and behaviourism, the intrinsic intentionality within those systems differ extensively. Although it may not be possible to prove that whether or not mental states exist at all in systems other than our own, in this paper I will strive to present arguments that a machine that computes and responds to inputs does indeed have a state of mind, but one that does not necessarily result in a form of mentality. This paper will discuss how the states and intentionality of digital computers are different from the states of human brains and yet they are indeed states of a mind resulting from various functions in their central processing systems.
Weizenbaum, Joseph. Computer Power and Human Reason: From Judgment to Calculation. New York: W. H. Freeman, 1976.
unified because reasoning and problem solving may involve several areas simultaneously. A robot circuitrepair syste m, for instance, needs to reason about circuits in terms of electrical connectivity and physical layout, and about time both for circuit timing analysis and estimating labor costs. The sentences describing time therefore must be capable of being combined w ith those describing spatial layout, and must work equally well for nanoseconds and minutes, and for angstroms and meters. After we present the general ontology, we will apply it to write sentences describing the domain of grocery shopping. A brief reverie on the subject of shopping brings to mind a vast array of topics in need of representation: locations, movement, physical objects, shapes, sizes, grasping, releasing, colors, categories of objects, anchovies, amounts of stuff, nutrition, cooking, nonstick frying pans, taste, time, money, direct debit cards, arithmetic, economics, and so on. The domain is more than adequate to exercise our ontology, and leaves plenty of scope for the reader to do some creative knowledge representation of his or her own. 228 Chapter 8. Building a Knowledge Base Our discussion of the