ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

ARTIFICIAL INTELLIGENCE

Lecture Note 1: INTRODUCTION

by

Nur Azzah Abu Bakar

School of Computing, UUM CAS

CIRCULATION OF THIS NOTE IS RESTRICTED TO STUDENTS OF STIN1013 INTRODUCTION TO ARTIFICIAL INTELLIGENCE, SCHOOL OF COMPUTING, UNIVERSITI UTARA MALAYSIA.

IT SHOULD NOT BE CITED IN ANY EXTERNAL DOCUMENT.

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 1

1.1. ARTIFICIAL INTELLIGENCE DEFINITIONS Some people like to view AI as a scientific endeavour; others view it as a very sophisticated engineering enterprise. Both are valid. AI is not only about the construction of systems (programs, hardware and the necessary situating and integrating into the world) that are useful and sufficiently adaptable to deserve the tag ‘intelligent’; it is also about the study of intelligence as an abstract concept and human intelligence in particular. Perhaps the first and most obvious question to address is: “What is Artificial Intelligence?” Unfortunately, there is no accepted definition among those working in this field; in fact, people’s ideas differ rather substantially, as can be seen from the following list of selected definitions (Table 1.1). The one given by Marvin Minsky is probably the most often quoted and universally accepted definition. Table 1.1: AI definitions

Author Definition

Charniack & McDemott (1985)

The study of mental faculties through the use of computational models. The ultimate goal of AI is to build a person.

Winston (1992) The study of the computation that made it possible to perceive, to reason and to act.

Kurzwell (1990) The art of creating machines that performs functions that require intelligence when performed by people.

Luger (2009) The branch of Computer Science that is concerned with the automation of intelligent behaviour.

Minsky (1968) The science of making machines do things that would require intelligence if done by humans.

Rich & Knight, (1994)

The study of how to make computers do things which, at the moment, people are better.

Schalkoff (1990) A field of study that seeks to explain and emulate intelligent behaviour in terms of computational processes.

Shapiro (2010)

A field of computer science and engineering concerned with computational understanding of what is known as intelligent behaviour, and with the creation of artefacts that exhibit such behaviour.

Definitions of AI vary along two dimensions; one is concerned with thought processes and reasoning whereas the other address behaviour. Thus, AI can be viewed from two different perspectives as follows: a. Cognitive Science-Psychology

This perspective views AI as a process-oriented field that studies how humans gained knowledge and became knowledgeable and later they used their knowledge to solve problems or make decisions. The goal is to understand the process by which humans

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 2

exhibit intelligent behaviour, and to replicate this on computers. Definitions 1 and 2 in Table 1.1 reflect this view as they emphasize on thought processes and reasoning. b. Engineering This perspective views AI as an output-oriented field which emphasizes on creating various applications which can demonstrate intelligent behaviour of humans. This view naturally emphasises techniques and technology and fosters research into the whole process of the specification, design and implementation of smart and/or self-adapting systems. The goal is to build machines which exhibit intelligent behaviour; it is of no consequence whether or not the techniques used bear any similarity to those employed by the human mind. Definitions 3 to 8 in Table 1.1 reflect this view. Russell and Norvig (2010) added to these views criteria to measure success of the machines (or AI systems). In each view, the system is measured against human performance and the ideal concept of intelligence (or rationality). As a result, AI definitions can further be viewed in four dimensions, i.e. system that think like human, system that thinks rationally, systems that act like human and systems that act rationally (see Figure 1.1).

Human performance Rationality

Co

gnit

ive

Systems that think like human Systems that think rationally

Engi

nee

ring

Systems that act like human Systems that act rationally

Figure 1.1: Dimensions of AI definitions from Russell & Norvig viewpoint

When we say “systems that think ...” one might ask this question: “Can a system (machine) think?” I leave this to you to find the answer to the question; detail instruction is given in the Exercise section at the end of this note. Regardless of the views, one important concept being highlighted in all the definitions; be it explicitly or implicitly, is the concept of intelligence. These definitions, however, are not overly enlightening as they leave unanswered to the question: “What is intelligence?” Without proper understanding of the term, you might still unclear what

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 3

AI is about; the following sections aim to foster your understanding by defining natural intelligence and comparing it with artificial intelligence. 1.1.1. Natural intelligence Natural intelligence being emphasized here refers to intelligence possessed by humans. Some of our common sense of the term includes ability to learn or understand, to remember, to think through a problem in order to find a solution, and to reason from first principles to solve problems for which no ‘cookbook’ answers exits. Like love, it is an abstract concept. People express their love, e.g. by showing ‘how much they care’ behaviour. People express their intelligence in the same way, but this time their behaviour shows their ability to do things such as to a. reason b. solve problems c. identify objects or images d. understand speech e. understand the relationships between objects or images f. make prediction g. plan etc. Natural intelligence exists in, and being used by all humans, and span across various levels. A two year old child uses his intelligence to, e.g. utter some words, understand basic instructions, recognize objects and deal with limited and simple tasks. As time goes by, he will be able to develop more complex sentences and follow more complex instructions. Similarly, a forensic expert uses his intelligence to solve a tragic murder case by reasoning with the collected evidences and his experience, or a pathologist who decides on patient’s health condition by intelligently interprets a laboratory test result. Of course intelligence is a vehicle in these examples, but their levels vary according to the complexity of the task being dealt with. The more complex the task is the higher level of intelligence is required. More explanation on the type of task and the level of intelligence is discussed in Section 1.3. 1.1.2. Natural versus artificial intelligence AI comes along with computational methods to automate human’s natural intelligence such as described in previous section. Therefore, AI attempts to do the same tasks the humans do, but in artificial way (as opposed to humans who do the tasks in natural way). In other words, AI mimics how humans naturally apply their intelligence in dealing with various tasks ranging from as simple as understanding a word to as difficult as solving a highly complex problems. Section 1.4.1 introduces you to various subfields of AI including knowledge representation and inference, planning and search, vision, natural language processing, expert system, data mining, neural network, agent, fuzzy logic, robotic, genetic algorithm and many others. Each is meant to create systems that imitate different ability of humans, or systems that are inspired by humans’ natural or biological process.

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 4

For example, the aim of natural language processing is to create a system that is able to understand or generate strings of words (mimics the ability of a two year old child in previous example). On the other hand, expert system aims to create a system that is able to reason with the knowledge in its knowledge base and the information at hand to make decision (mimics the ability of a forensic expert or pathologist in previous example). AI imitates these abilities into an NLP system or expert system; when natural intelligence is imitated into a machine, it is AI! A good justification for this claim lies on the fact that intelligence in machines is programmed by humans and therefore it is artificial. You can now create your own definition of AI based on your current state of understanding of the field. Table 1.2 shows some dimensions to compare between natural intelligence and artificial intelligence. Table 1.2: A comparison between natural and artificial intelligence

Natural intelligence Artificial intelligence

Not consistent Consistent

Cannot be duplicated Can be duplicated

High cost Low cost (long-term)

Hard to be documented Easy to be documented

Creative Not creative (work as programmed)

Wider focus Limited or narrow focus

1.2. ARTIFICIAL INTELLIGENCE PROGRAM Another slippery question is “What is AI program?” The obvious and hence unsatisfying answer is “any program which exhibits intelligent behaviour”; correct, yet it gives no insights into how people go about building intelligent programs, or what they consist of. The overall structure of AI program is as illustrated in Figure 1.2. It tells us that in building an intelligent program, one need to encode the relevant information and knowledge in a form suitable for computer processing (this is referred to as internal representation), processes the information and then retranslated it back to a form understandable by humans.

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 5

Figure 1.2: Structure of an AI program

Of course the same structure applies to any computer program, but the nature of the internal representation and how it is processed in AI program are different; this has set AI programs apart from any conventional programs. As intelligent behaviour requires knowledge, the builder of any AI program must consider two major issues: a. knowledge representation b. knowledge application (i.e. reasoning) It is important to highlight here that representing and applying knowledge are two different things. There are some people who have a great store of knowledge, yet they seem to lack sufficient common sense, or indeed intelligence to put this knowledge to use. Truly intelligent people not only know a great deal, but they also know how to reason with this knowledge to derive new information. Equally important is their ability to direct their reasoning fruitfully thus enabling the speedy analysis of a situation, or solution of a problem. Further discussion on how one can represent knowledge in a computer, and how and when to use the knowledge to make inference is discussed further in Lecture Note 2. 1.2.1. Artificial Intelligence program versus conventional program Any AI program has the following two characteristics which conventional programs typically do not: a. there is some explicit representation of knowledge in the program b. the basis for programming is to manipulate symbols rather than numbers. Conventional programs do indeed embody considerable amounts of knowledge; however, it is not made explicit. Instead, the knowledge remains deeply buried in the code. Two unfortunate consequences of this are that the knowledge is difficult to modify and cannot be used to explain the program’s reasoning. The knowledge used is not readily accessible to the programmers, for examination and modification.

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 6

When writing AI programs the primary activity is in defining, characterising, and manipulating symbols (would ordinarily be in the form of strings of letters or words). However, in no way can it be said that all programs based on symbolic computation are AI programs; and also, nor is it the case that AI programs may not do number crunching. Some applications, such as vision programs, do indeed require considerable amounts of numerical computation in addition to symbolic programming. It is also not the case that AI programs are better than conventional programs. AI programming techniques have been developed in response to specific needs of creating intelligent computers. There are still zillions of tasks which are most effectively performed using conventional techniques. It is simply a matter of selecting the right tool for the right job. From my viewpoint, AI and conventional techniques are two entities that complement each other.

1.3. ARTIFICIAL INTELLIGENCE TASKS Humans perform various tasks every day, which differ in terms of their difficulty or complexity and also the level of intelligence required. Rich and Knight (1991) classified these tasks into three categories, i.e. mundane, formal and expert tasks, which are the targets of work in AI. 1.3.1. Mundane task Mundane tasks can be simply said as every day task or task which most humans are able to perform successfully in meeting their daily needs. This include perception, both vision and speech; natural language understanding, generation and translation; commonsense reasoning and robot control. To make it simple, consider some of these examples: recognizing objects, understanding written or spoken words, deciding how to get to work in the morning and moving around freely within a limited space; an example with the two year old child in previous section also falls under this category. Despite the fact that these tasks are assumed as easy to be done by humans, they proved to be the most difficult tasks to be automated for the reason that they are much unstructured in nature. Perceptual tasks, for example, are difficult because they involve analog signals which are typically very noisy. This is exacerbated by the fact that usually a large number of things must be perceived at once. Natural language understanding, on the other hand, requires a great deal of knowledge about the language itself including its grammar and vocabulary, as well as the topics so that unstated assumptions can be recognized. 1.3.2. Formal task Much of the early work in AI focused on formal tasks such as game playing (e.g. checker and chess) and theorem proving. Among the famous work include Samuel’s checkers-

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 7

playing program and the Logic Theorist. Game playing and theorem proving share the property that people who do them well are considered to be displaying intelligence. 1.3.3. Expert task Expert task is a specialized task in which carefully acquired expertise is necessary. Examples of the task include engineering design, scientific discovery, medical diagnosis and financial analysis. Early assumption been made in AI was these skills were harder and less amenable to be automated than the mundane tasks as they are learned later; everybody gained perceptual, linguistic and commonsense skills first before they acquire engineering, medicine and finance skills. This naive assumption has directed the work in AI to focus more on the mundane tasks, but later it proved that it was not right. Expert skills indeed require knowledge that many of us do not have, but they often require much less knowledge than do the mundane skills and that the knowledge is usually easier to represent and deal with inside the programs. As a result, the problem area where AI has first been flourished as a practical discipline (as opposed to a pure research) was the one that require specialized expertise, in other words, expert systems.

1.4. FOUNDATIONS OF AI AI merges Computer engineering with various disciplines including Philosophy, Mathematics, Psychology and Linguistic. This has made AI as a special field; while the disciplines provide strong theoretical foundation for AI, Computer engineering provides the tools to make AI a reality. The ideas and theories from these disciplines have much influence on the work in AI. For example, the idea of means-end analysis in Philosophy influenced Newell and Simon to work on their General Problem Solver (GPS) program (Newell & Simon, 1972). Mathematics has much contributed to AI through formal theories of logic, probability, decision making and computation. Linguistics influenced the early work in knowledge representation; both modern linguistics and AI intersect in a hybrid field called computational intelligence (another name for NLP). 1.4.1. Subfields of Artificial Intelligence The disciplines mentioned above are the major and early contributors to AI. In addition, there are also other disciplines that form what is known as AI today, as depicted in Figure 1.2 (Turban et al., 2006). The circles labelled as A to P are the root disciplines for AI. On top of the tree are ‘fruits’ that represent various subfields of AI (bear in mind that it is not a comprehensive list of AI subfields). A very brief description of each subfield is provided as follows. You will be introduced to some of these subfields in later weeks of this semester, after we accomplish the first two major subfields being discussed below.

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 8

Figure 1.2: The AI tree

a. Knowledge representation and inference This is the heart of AI. Any sort of intelligence that one can imagine will certainly require the existence and application of knowledge. We humans know an awful lot, but we do not know very well how that knowledge is stored in our brains. Neither do we understand much about the mechanisms of accessing the knowledge and making inferences from it. This very large and significant subfield concern itself with these issues.

b. Planning and search This subfield is also known as problem solving. Solving problems invariably concerns itself with creating some plan of action to solve the problem. Creating this plan invariably requires some sort of implicit or explicit searching through all the possible approaches to solving the problem. This is another core AI subfield which is relevant to most of the rest of AI. c. Vision This subfield concerns with this question: “how do we see objects?” or more importantly “how do we recognise what it is that we see?”

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 9

d. Natural language processing The goal of Natural language processing (NLP) is to build computers which humans can converse with in their native languages, rather than some highly formal computer language. Most work has concentrated on English, but there are a number of other languages under study including Malay, Arabic, Japanese, Chinese, French and Italian. e. Expert systems Expert systems have been defined to be computer programs which can perform some task which is difficult enough to require genuine human expertise (as opposed to simply requiring some intelligence). The range of applications to date is extremely wide, from medical diagnosis to mineral prospecting to nuclear power plant monitoring system. Expert systems development makes use of knowledge acquisition techniques a lot, i.e. to acquire knowledge from human experts in a particular domain.

f. Fuzzy logic Fuzzy logic (FL) is based on fuzzy set theory which classifies elements or groups of items into sets with varying degrees. It is unlike mathematical sets which classify elements into sets or not into sets. Let’s take a sparrow and a bat as examples. In mathematical sets, a sparrow belongs to a bird set but a bat does not and that’s it. In fuzzy set, a bat can belong to a bird set (as it has wings, one of the attributes of bird) but only to a certain degrees, probably 20% or 0.2; a sparrow, however, is definitely belongs to a bird set and therefore, the membership degree is 1. Nowadays, there are many household appliances with FL built in feature that makes their use easier. FL has been used in, e.g. rice cookers, to give them ability to make proper adjustment to cooking time and temperature. Other devices that use FL to function include washing machine, refrigerators, computers, toilet, traffic light controller, subway cars etc. g. Robotics Robotics aims to build robots which can perform motor tasks like going through a room or picking up an object from an assembly line. This subfield is, to some extent, away from mainstream AI and overlaps considerably with physics, applied maths, mechanical engineering etc. By the way, successful robots also need to apply technology from many other areas of AI including planning and search, vision as well as knowledge representation and inference.

h. Intelligent tutoring system This subfield is about the use of AI in education and concerns about how computers can be used to teach. Well, it is not just how to use computers in teaching; this is what a conventional program called computer-aided instruction (CAI) have been doing; it is how to do more than that so as to allow the learners to gain maximum benefits from it. Intelligent tutoring system (ITS) or intelligent computer-aided instruction (ICAI) focuses

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 10

on tutoring or coaching the learners according to their own pace. Issues related to this field include building intelligent interfaces to existing computer systems, building models of learners which can be used to expedite learning etc.

i. Machine learning Learning is an expanding area and overlaps with almost all other areas of AI. In robotics, the system learns rules of behaviour from experience in some environment; in natural language, a system learns syntactic rules from example sentences; in vision, a system learns to recognise some objects given some example of images; in expert systems, a system learns from examples cases. Most of the work in AI has been on learning from examples; this is what is called inductive learning. Real machine learning applications typically require hundreds to thousands of examples in order for interesting knowledge to be learned. Machine learning is crucial in AI; a system cannot be claimed as intelligent if it does not have an ability to learn. j. Neural networks This field is about a learning algorithm/technique which is inspired by the hypothesis that mental activity consists primarily of electrochemical activity in networks of brain cells called neurons. In neural network (NN), a single neuron is represented in a form of a simple mathematical model; many neurons are connected together to form a network. Other names for this field include connectionism, parallel distributed processing and neural computation.

k. Genetic algorithm In contrast to NN, this learning algorithm makes use of a metaphor based on evolution, specifically those that follow the survival of the fittest principle of Charles Darwin. Genetic algorithm (GA) is a search algorithm (in particular, a heuristic search algorithm), most commonly referred to as an optimization technique as it finds the best solution (among the many solutions) to a given problem. In GA, solutions to a problem are represented as a population of strings called chromosome (usually represented as strings of 0s and 1s, e.g. 110100110). A value (called fitness function) is assigned to the chromosomes so that they can be evaluated; only the fittest will be selected to evolve next. GA belongs to a larger class of evolutionary computing (EA). l. Data mining The term is being used intertwined with knowledge discovery in databases (KDD). In actual, data mining is one particular step of KDD processes. Data mining and KDD concern with the development of methods and techniques for making sense of the data from large databases, i.e. getting the data into more compact form (e.g. a short report or a predictive model etc) so that if gives us basis for future decision making and planning. Data mining is particularly useful when a large volume of digital data exists as it can expedite the data analysis. Classical approach to data analysis relies fundamentally

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 11

on analysts becoming intimately familiar with the data. As data volumes grow dramatically (many databases now are in terabytes size), the manual analysis is becoming impractical. This field itself is a confluence of many disciplines including machine learning, pattern recognition and statistics (Tan, Steinbach & Kumar, 2006).

m. Intelligent agent An intelligent agent (or simply an agent) is a program that acts (i.e. gathers information or performs services) without your immediate presence and on some regular schedule. This program is used extensively on the Web as, for example, the web browsers, news retrieval mechanisms or shopping assistants. Some are also used as tools to track Web behaviour; they "watch" as your surf the Net and record how often you visit certain sites. Later, they can be used to automatically download your favourite sites, let you know when your favourite site has been updated, and even tailor specific pages to suit your preferences. Besides this software agent, there is also a physical agent (i.e. robots) which operates in the physical world. This agent (e.g. a gripper to pick up things, a paint sprayer, a welding gun etc) is usually used to perform a variety of manufacturing tasks. Many also exist in a form of mobile robots that can move around, e.g. a “delivery boy” that take some object at some location in the building, navigate its way about the building and make the delivery. Mobile robots are also used to do routine tasks in hazardous environments such as the surface of Mars, the dangerous bits of a nuclear power station, or near fire (Cawsey, 1997). n. Swarm systems This newly emerged field within AI is inspired by nature such as ant colonies, birds flocking, animals herding and fish schooling. Two popular swarm-inspired methods are ant colony optimization (ACO) and particle swarm optimization (PSO). ACO is inspired by the behaviour of a swarm of ants whereas PSO is inspired by the social behaviour of a flock of birds or a school of fish. Let’s consider a swarm of ant, when their route is blocked they will find the shortest new route, showing their robustness. These ants can be added or removed without compromising the total system due to its distributed nature. This kind of adaptation of a system is reliable as single parts can break down without impairing the whole system. This makes a complex system easy to handle because of the simplicity of their individual parts.

1.5. HISTORY OF AI In general, the history of AI can be divided into phases, or era, to reflect what had actually happened to the field during each stipulated period of time. 1943 – 1956 The birth of AI.

1956 – late 1960s The rise of AI.

Late 1960s – early 1970s First AI winter.

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 12

Early 1970s – mid 1980s

The era of expert systems.

Late 1980s – present The return of neural networks and emergence of new approaches (base claims on rigorous theorems or hard experimental evidences rather than intuition; show relevance to real-world application rather than toy problems).

I leave this to you to further explore this field since the early days of its inception. Detail instruction is given in Exercise.

EXERCISE 1. Define Artificial Intelligence in your own words. 2. List the skills and knowledge required to successfully do the following tasks: reading

a book; crossing the road; ordering a pizza; arranging a trip to Langkawi. What type of tasks are they? Why do you think so?

3. Research on the Turing Test and report how it is conducted (you might discover different versions of the test). What sort of intelligence being tested in the test? What are the objections? How Turing responds to these objections?

4. Every year the Loebner Prize is awarded to the program that comes closest to passing a version of the Turing Test. Research and report on the latest winner of the prize.

5. Examine the AI literature to discover whether the following tasks can currently be solved by computers. If any of the following is still infeasible, try to find out what the difficulties are and predict when, if ever, you think they will be overcome.

a. Playing table-tennis (ping pong). b. Driving in the centre of Cairo, Egypt. c. Driving in Victorville, California. d. Buying a week’s worth of groceries at the market. e. Buying a week’s worth of groceries on the Web. f. Playing a decent game of bridge at a competitive level. g. Discovering and proving new mathematical problems. h. Writing an intentionally funny story. i. Giving competent legal advice in a specialized area of law. j. Translating spoken English into Swedish in real time. k. Performing a complex surgical operation.

6. There have been many AI contests being held in the area of robotic, information retrieval, machine translation, speech recognition etc. Research on five of these contests and report the progress made over the years.

7. Expand section 1.5 by adding the details to each of the given period of time of AI history. You might discover some important events, places, persons, programs etc while searching for the information; include these in your description.

ARTIFICIAL INTELLIGENCE Lecture Note 1: INTRODUCTION

Nur Azzah Abu Bakar, School of Computing, UUM CAS 13

REFERENCES Cawsey, A. (1998). The Essence of Artificial Intelligence. Pearson Education Limited: England. Charniack, E. & McDemott, D. (1985). Introduction to Artificial Intelligence. Reading: Addison-

Wesley. Kurzweil, R. (1990). The Age of Intelligent Machines. MIT Press. Luger, G. F. (2009). Artificial Intelligence: Structure and Strategies for Complex Problem Solving

(6th

edition). Pearson Education Inc.: Boston. Minsky, M. (1968). Semantic Information Processing. Cambridge, Massachussets: MIT Press. Negnevitsky, M. (2011). Artificial Intelligence: A Guide to Intelligent Systems. Pearson Education

Limited: England. Rich, E. & Knight, K. (1991). Artificial Intelligence (2nd edition). McGraw-Hill: Singapore. Russell, S. J. & Norvig, P. (2010). Artificial Intelligence: A Modern Approach (3rd edition).

Pearson: New Jersey. Tan, P., Steinbach, M. & Kumar, V. (2006). Introduction to Data Mining. Pearson Education Inc.:

Boston. Turban, E., Aronson, J. E., Liang, T. & Sharda, R. (2006). Decision Support and Business

Intelligence System (8th edition). Pearson: New Jersey.

USEFUL LINKS 1. http://www.aaai.org/home.html for various links to AI resources. You will find links to

various AI topics, conferences, journal etc. 2. http://www.cs.unm.edu/~luger/ai-final/ for some online stuff from Luger F. Subblefield. 3. http://loebner.net/Prizef/TuringArticle.html for Computing Machinery and Intelligence

article, featuring Turing Test. 4. http://crl.ucsd.edu/~saygin/papers/MMTT.pdf for reviews on Turing Test after 50 years. 5. https://www.ai-class.com/ for enrolment into online AI class run by Sebastian Thrun and

Peter Norvig, both are experts in the field [note: enrolment is now closed, check it out regularly if interested].

6. http://www.cs.berkeley.edu/~russell/aima1e/chapter02.pdf for online 1995’s Russell and Norvig AIMA book: Chapter 2 – Intelligent Agent.

7. http://www.irobot.com for details about Roomba, an intelligent vacuum cleaner robot. 8. http://www-robotics.usc.edu/~maja/teaching/cs584/.../thrun-stanley05.pdf for details

about Stanley, a robot car that won 2007’s DARPA Grand Challenge in Nevada. 9. http://www.ehow.com/video_4767093_artificial-intelligence-used_.html for more info on

where AI is used. 10. http://www.youtube.com provides interesting videos for learning AI. Search using keywords

such as “asimo”, “roomba”, “chatterbot”, “deepblue” etc to see (and find more links), for example, how ASIMO avoids obstacles while moving around real people, how Roomba does the house cleaning then goes to recharge itself, how the airline bots respond to your natural language enquiries with regard to flight schedule and others (also visit http://alice.pandorabots.com, http://www.jabberwacky.com/, http://www.chatbots.org/ etc).

11. http://www.cs.rochester.edu/~brown/242/assts/termprojs/games.pdf for AI in games. 12. http://www.youtube.com/watch?feature=player_detailpage&v=J_Q5X0nTmrA for an

interactive video on Fuzzy Logic.

Note: Resources on AI are widely available on the internet, and are not limited to the above list.