The invention and the continuous research on intelligent machines have greatly impacted on the way man carries out his day to day activities. This research paper examines the topic “Artificial Intelligence” in its broader perspective. It gives the definitions and brief background information on the topic of study and then extends to detailed examination of the topic as given in various literature materials. Artificial Intelligence is considered as a great step in technological innovations and this paper analyses its importance and the areas where it finds applications. A conclusion of this work affirms that, indeed Artificial Intelligence has greatly enhanced the way of life and changed the face of the society.
The invention of computers and the rapid changes in technology has taken life to a whole new level. Nowadays, intelligent machines -- which can perform logical operations -- have been invented; thanks to artificial intelligence. This has really made work easier, especially the invention of robots which look and act like human beings. This research paper focuses on Artificial Intelligence (AI) and its impacts on the society.
Definitions Artificial Intelligence
In basic terms, Artificial Intelligence is defined as the technology of making intelligent machines. An intelligent machine is one that performs logical operations just as human beings (McCorduck, 2004). AI also tends to make computers understand the behavior and intelligence of human; however, it does not confine to biologically observable methods (McCarthy, 1957).
Artificial intelligence can also be defined as a computer that tries to achieve or simulates true human intelligence (Bowles, 2010).
These are computers and programs with extensive information in a specialized field (Bowles, 2010).
A robot is a computer that looks like a human being, and at the same time acts as one (Bowles, 2010).
This is the copying of some real life aspects and displaying the experience on a computer (Bowles, 2010).
This is the use of computers in creating simulated environments with illusions of reality (Bowles, 2010).
The idea, hope and dream of having intelligent machines came with the invention of the modern computers in the 20th century. The term artificial intelligence came into existence in 1956 when John McCarthy proposed it. The first artificial intelligence software program, the Logic Theorist, was demonstrated by Herbert Simon and others. Thereafter, programmers began developing softwares which could brilliantly perform reasoning tasks. There were high hopes of achieving real human intelligence on the computer; however, this failed. Natural intelligence and the variations in language skills proved difficult to imitate. Programmers found it very difficult to duplicate human intelligence and therefore limited themselves to computer programs with extensive information (expert systems) in specialized fields. Such computers included those that could perform medical diagnosis. They included MYCIN and Internist. The most advanced artificial intelligence was realized in 1997 when IBM's Deep Blue managed to out-do the world’s top chess player (Bowles, 2010).
The aim of this research paper is to analyze the impacts of Artificial Intelligence on the society.
The paper covers the basics of AI and gradually advances to the slightly complex disciplines of AI. This is achieved through extensive literature review. It then focuses on the applications of AI, which is the major theme.
In this section, I have keenly analyzed Artificial Intelligence as a major discipline and extended the analysis to the subjects covered under AI.
Branches and Disciplines of Artificial Intelligence
Artificial intelligence is broad based and cuts across various disciplines. It involves various schemes in knowledge representation, intelligent search techniques, uncertainty resolving techniques, and machine learning schemes, among others (Russell and Norvig, 2003). The realization of Artificial Intelligence needs a combination of various disciplines also known as the parent disciplines as indicated in the figure below. The figure summarizes both the applications and the subjects of Artificial intelligence.
Source: Artificial Intelligence. Retrieved at http://www.learnartificialneuralnetworks.com/ai.html
The subjects within the broader AI are discussed hereunder.
Knowledge Representation, Reasoning and Common Sense
Reasoning involves evaluation of various initial states, and reaching a pre-defined state which meets the required goal. The more the transition states, the lower the reasoning efficiency. For a reasoning system to have higher efficiency, then, the intermediary states must be minimized as possible. The states must therefore be complete and well organized. Such states require minimum search so than an appropriate knowledge can be identified at any given problem. Knowledge must therefore be well organized in any system. Artificial Intelligence employs the use of various techniques of knowledge representation. Such techniques include predicate logic, production rules, filler and slots, and semantic nets among others. For a given knowledge representational scheme to be selected, both the application nature and the users must be considered. In this area, Artificial Intelligence is furthest from reaching the human intelligence (Kurtzweil, 2005). However, non-monotonic reasoning systems have been developed. The action theories have also been developed; however, more ideas and facts are still needed.
Planning is a very significant area in Artificial Intelligence. Even though planning and reasoning share various issues, reasoning is all about testing whether a given goal satisfies the set of given data and knowledge. Planning on the other hand, is concerned with the evaluation of the methods which can result into the achievement of a successful goal. Planning can be automated. An automated planning is greatly applied in robotics and in navigational problems (McCorduck, 2004).
A planning program begins with the known facts (general facts) about a given situation and the statement of goals. From the given information, strategies for achieving the goal are then generated. The strategies, in most cases, are sequence of actions.
Acquisition of knowledge is hard for human beings and requires proper training. In a similar way, it is hard for machines. Knowledge acquisition involves the generation of pieces of knowledge out of a given knowledge base, the setting of dynamic data structures, learning from the environment, and refining the knowledge. Currently, research is underway in machine learning through automated knowledge acquisition (McCorduck, 2004).
This is a collection of tools and computing techniques. Such tools and techniques are shared by various disciplines in artificial intelligence. They include: the genetic algorithms, the inductive logic programming, the fuzzy logic, the belief calculus, and the artificial neural nets (Nilsson, 2002). The application of these tools depends on the application domain; thus, they can be used jointly or independently.
In basic terms, soft computing is defined as the emerging computing approaches which together parallel the reasoning-ability of the mind so as to learn and reason in an uncertainty and imprecision environment (Nilsson, 2002).
Imprecision and Uncertainty Management
mprecision is the incompleteness of data. It results from poor authenticity and inappropriate data sources. On the other hand, uncertainty is the knowledge incompleteness. AI approaches presents various techniques which are devised to handle data under the above two conditions.
Logical Artificial Intelligence and logic programming
Here, the system is programmed such that it understands the facts about a given situation where it is required to act. These facts are represented in a logical language. The program then decides its course of action by inferring that some actions are more appropriate in achieving specific goals. Extensive research is currently underway in Logic Programming (McCorduck, 2004).
The programs of Artificial Intelligence normally consider various possibilities such as the moves in a game or conclusions from the theorem-proving programs. The search problems in Artificial Intelligence are not deterministic. Also, the order in which the search space elements are visited is highly dependent on the given data sets (Nilsson, 2002).
Various programs have been programmed such that they compare the observations made with a given predetermined pattern. Complex patterns demands that more complex methods are used.
Mathematical logic languages are used in representing the world’s facts.
Learning from experience
Programs only learn the behaviors and factors resulting from their formalism. These learning systems have limited abilities in representing information.
Epistemology and Ontology
Epistemology studies the different kinds of knowledge necessary for providing solutions to world problems while Ontology studies the different kinds of things that are in existence. Artificial Intelligence programs handle various objects and their basic properties.
This is the copying of some aspects of real life and representing the experience on a computer (Bowles, 2010). A simulated environment with illusions of reality is created by a process called virtual reality. Virtual reality and computer simulation have applications in various fields like in games and in advanced motion controls. Simulations can also be used to solve real problems and to analyze real systems.
Pilots use simulation for flight training through the help of such programs as Microsoft Flight Simulator. Drivers also use such programs when learning how to operate city buses. The simulation involves driving in town, virtual passengers, animated wipers, artificial intelligent traffic, etc (Bowles, 2010). The program helps in simulating real driving behaviors which help in training drivers.
Applications of Artificial Intelligence
Artificial intelligence has a broad application in our everyday life. Some of the areas where AI is greatly applied include robotics, expert systems, game playing, image recognition, theorem proving, and natural language processing, among others (Nilsson, 2002). These applications are discussed hereunder.
Artificial intelligence has taken robotics to a higher level; however, the human-like robots are yet to be developed. It is proving extremely difficult to create machines which think and move in a similar manner as humans. Robots have existed for decades; however, they are just programmed mechanical arms whose role is to move components from one place to the other in a production process. Work has been extremely made easy by the robots since they out-do humans in speed, accuracy and energy; just as computers. The robots can handle large volumes of work faster and for longer durations (RedOrbit.com., 2009).
A research study on accuracy and speed of robots against that of human in a heavy industry, gave the following results (McCarthy, 1990).
It is obvious from the results that robots are more accurate and faster that human in some tasks.
The major area where humans beat computers and robots is reasoning and decision making. Such machines lack the ability to reason on their own and make appropriate decisions.
Scientists have however moved steps ahead in giving computers some limited ability to make decisions. Currently, there is an intensive study on how the human brain works in hopes of developing robots that can act as humans.
These are computer programs which mimic experts in data evaluation. They are applied in various fields as highlighted hereunder.
In hospitals, such programs are used to diagnose medical conditions of patients based on the vital signs like blood pressure, temperature, heart rate, etc (Bowles, 2010).
In manufacturing, the expert systems are used to cut the operating costs. Such systems are trained on how to recognize when a production machine needs a preventive maintenance.
In financial services industry, computer programs are extensively used. The programs have a neural network technology which evaluates the information of a credit applicant. Such information include age, income, occupation, and past credit records. After the evaluation, the program determines whether the individual is fit for credit or not; however, the program does not decide on the relevant information for the analysis, and cannot determine on its own the constituents of a bad or good risk.
Computer Vision and Image Understanding
Artificial Intelligence techniques and tools are very necessary in vision systems especially the high level vision. In such an application, a machine recognizes the object from its image. This is achieved through pattern classification. Pattern classification is also achieved through the supervised learning algorithms. Interpretation of the patterns is made possible through knowledge-based computation.
Natural Language and Speech Understanding
Artificial Intelligence techniques are very important in understanding of the natural language and speech. Speech analysis involves the separation of the spoken words and determination of the features of the separated word. The word is then identified through the application of the techniques of pattern classification. The method currently employed in word classification is the artificial neural networks.
To understand a natural language, there must be both semantic and syntactic interpretations of all the words in a given sentence and all the sentences. Syntactic steps helps in the analysis of grammar while the semantic analysis is used to determine the sentence meaning.
Robots have been made to understand the speech in the natural language. Such robots are of vast importance since they execute all the tasks verbally instructed to them. Commercially, natural language and speech understanding has been employed. In this application, a phonetic typewriter is used. The typewriter is capable of printing the words that are pronounced by the human operator.
Artificial Intelligence techniques also help in scheduling. A scheduling problem requires one to plan on the time schedule for a given set of events in order to improve the efficiency. Flow Shops are used to determine the optimal scheduling with regards to the job size and the machine size. The most preferred solution for scheduling problems depends on the nature of the given conditions; however, in most cases, sub-optimal solution is better. Scheduling solutions are achieved through the application of artificial neural nets in conjunction with genetic algorithms. Heuristic search is also employed.
Artificial Intelligence is greatly applied in process control. In such control, the known process models and the control objectives required are used in designing the controller. If the plant’s dynamics is not known completely, then, the existing controller design techniques cease to be relevant. In such a situation, the appropriate control system is the Rule-based control. Rule-based control is achieved by sets of production rules instinctively set by the expert personnel. The premise (antecedent) part is searched with respect to the plant parameters’ dynamic response. If the premise part of the rule matches the response of the plant, then, the rule is selected and finally fired. If more rules are in a position of being fired, then the controller employs a given set of strategies to resolve the conflict. If in the resulting situation, the plant response fails to match with any rule, then a solution is found through the application of fuzzy logic. This logic has the ability to match the premise part with the dynamic response part. Various industrial plants employ the use of fuzzy control. An example of such application includes the nuclear reactor power control.
Other than the controller, process control also involves the design of the plant estimator or the process estimator. The role of the estimator is to follow the actual plant’s response when both have a common input and are jointly excited. Resent research have identified the techniques of artificial neural network and fuzzy logic as the current plant estimation tools.
Artificial Intelligence has a wide range of applications in man’s everyday life. Thousands of the applications are deeply rooted in every industry. Various tools have been created to aid in solving the complex computer science problems, thanks to AI. In banking and finance, Artificial Intelligence systems have been used in organizing operations, investing in stocks, and in managing properties. Institutions, especially the financial ones, have used AI techniques in detecting charges and claims, paving way for the human investigation. In medicine, AI systems have been used in clinics to organize various schedules like staff rotation, bed schedules, and to provide important medical examination and information. Decision support systems in clinics use the artificial neural networks in medical diagnosis. Robots have become part and parcel of every heavy industry. The robots normally handle those jobs that are dangerous, or considered so, to humans. They have been very efficient in repetitive jobs as they eliminate the mistakes which humans normally make due to concentration lapse. In automated online services, artificial intelligence is greatly employed. Also, similar techniques are used in the answering machines at the call centers. Such machines have programs that recognize speech. This gives the computers the ability to handle the first levels of customer support. In transportation, the automatic gearbox of some automobiles normally has the fuzzy logic controllers. Other fields like telecommunications, music, aviation, games, among others, greatly apply Artificial Intelligence techniques. Indeed AI has greatly enhanced man’s way of life.
Bowles, M. D. (2010). Introduction to Computer Literacy. Bridgepoint Education inc., San Diago
Dennett, D. (1998). Brainchildren: Essays on Designing Minds. MIT Press.
McCarthy, J. (1957). “Programs with Common Sense,” In Mechanisation of Thought Processes, Proceedings of the Symposium of the National Physics Laboratory, pages 77-84, London, U.K. Retrieved at http://www-formal.stanford.edu/jmc/mcc59.html
McCarthy, J. (1990). Formalizing Common Sense: Papers by John McCarthy. Ablex Publishing Corporation.
Kurtzweil, R. (2005). The singularity is near: when humans transcend biology. New York: Viking.
McCorduck, P. (2004). Machines who think: A personal inquiry into the history and prospects of artificial intelligence. Natick, MA: A K Peters.
NeuroAI. (2007). Artificial Intelligence. Retrieved at http://www.learnartificialneuralnetworks.com/ai.html
Nilsson, N. J. (2002). Artificial intelligence: A new synthesis. San Francisco, CA: Kaufmann.
RedOrbit.com. (2009). Robot scientists mark advancement for artificial intelligence. Retrieved at http://www.redorbit.com/news/technology/1665464/robot_scientists_mark_advancement_for_artificial_intelligence/
Richmond, T. (2003). “Logic and artificial intelligence” In Edward N. Zalta, editor, The Stanford Encyclopedia of Philosophy. Retrieved at http://plato.stanford.edu/entries/logic-ai/
Russell, S. J. and Norvig, P. (2003). Artificial Intelligence: A Modern Approach. 2nd ed. Upper Saddle River, New Jersey: Prentice Hall. Retrieved at http://aima.cs.berkeley.edu/