CSNB234 Artificial Intelligence CSNB234 Artificial Intelligence AI in our everyday lives Phone Assistants Games 2 CSNB234 Artificial Intelligence AI in our everyday lives Spam Blockers Google Translate 3 CSNB234 Artificial Intelligence Why Learn AI? Basic Level • To better understand the systems and tools that you interact with on a daily basis Simple Medical Diagnosis Career Guidance Animal Identification Advanced Level • Create AI applications, like the Google Self Driving Car, or IBM’s Watson Google Self-Driving Bike And SelfDriving Car: rely on their sensors and software to drive themselves IBM’s Watson: a technology platform that uses natural language processing & machine learning to reveal insights from large amounts of unstructured data 4 CSNB234 Artificial Intelligence AI preparing you … Software Engineer Hardware Engineer • Create shopping list recommendation engines • Analyzing and processing big data • Developing electronic parking assistants or home assistant robots Computer Graphics Cyber Security • Creation of more intelligent interactive characters, who can adapt on user's input to determine its type of gameplay, its mood • Detect threats, including those are yet to be discovered, by identifying characteristics within families of threats 5 CSNB234 Artificial Intelligence Outline • Artificial Intelligence (AI) Background and History • AI Languages • Overview of AI Application Areas 6 CSNB234 Artificial Intelligence 7 CSNB234 Artificial Intelligence Can a machine think? • Can be answered by the following “tests” for machine (i.e. the program/software) • The Alan Turing Test – Alan Turing (father of AI) • Revised Turing Test – ELIZA (By Joseph Weizenbaum of MIT) 8 CSNB234 Artificial Intelligence Various AI Definitions HUMAN PERFORMANCE THOUGHT PROCESSES AND REASONING “The exciting new effort to make computers think ... machines with minds, in the full and literal sense” (Haugeland, 1985) “The automation of activities that we associate with human thinking, activities such as decision-making, problem solving, learning ...” (Bellman, 1978) Systems that think like humans BEHAVIOR “The art of creating machines that perform functions that require intelligence when performed by people” (Kurzweil, 1990) “The study of how to make computers do things at which, at the moment, people are better” (Rich and Knight, 1991) Systems that act like humans RATIONALITY “The study of mental faculties through the use of computational models” (Charniak and McDermott, 1985) “The study of the computations that make it possible to perceive, reason, and act” (Winston, 1992) Systems that think rationally “A field of study that seeks to explain and emulate intelligent behavior in terms of computational processes” (Schalkoff, 1990) “The branch of computer science that is concerned with the automation of intelligent behavior” (Luger and Stubblefield, 1993) Systems that act rationally 9 CSNB234 Artificial Intelligence AI Definition • AI can be defined as a part of computer science that concerned with the designing of intelligent computer systems, that is, systems that exhibit characteristics we associate with intelligence in human behavior. • The goal of AI is to develop computers that can think, see, hear, walk, talk and feel. 10 CSNB234 Artificial Intelligence AI Definition • What computer can do better than people? – Numerical computation: Fast & accurate – Information storage: Voluminous amounts – Repetitive operations : Not getting bored (??) • However, these are mechanical mindless activities, and thus cannot be regarded as ‘intelligent’ tasks 11 CSNB234 Artificial Intelligence What people can do better than computers? • Activities that involve intelligence include: Understanding Common sense reasoning Natural language processing and generation Planning & Design Learning (e.g. from mistakes, by analogy, by experience or examples) – Emotions – – – – – 12 CSNB234 Artificial Intelligence What is “intelligence”? It has the ability – To respond to situation very flexibly – To make sense out of ambiguous messages – To recognize the relative importance of different elements of a situation Intelligent Agent Environment Sensors • • Actuators • Intelligent agent interacts with environment Agent receives state of environment through its sensor & make decision that can carry out by actuator (based on sensor data) The important part: AI able to maps sensor to actuator through control policy/rules 13 CSNB234 Artificial Intelligence Human Intelligence vs AI Human Intelligence • Natural intelligence • Intuition, common sense, judgment, creativity, beliefs, etc. • Ability to demonstrate their intelligence by communicating effectively • Probable reasoning & critical thinking Artificial Intelligence • Intelligences posses by machines • Ability to simulate human behavior & cognitive processes • Capture & preserve human expertise • Flexibly response – ability to comprehend large amounts of data quickly 14 CSNB234 Artificial Intelligence IQ • The most widely accepted psychometric test is the Intelligence Quotient, or IQ test 15 CSNB234 Artificial Intelligence Alan Turing • Father of Artificial Intelligence is Alan Turing (1912-1954) 16 CSNB234 Artificial Intelligence Turing Test • This test was invented by Alan Turing (1912-1954) • It was first described in his 1950 article Computing machinery and intelligence (Mind, Vol. 59, No. 236, pp. 433-460) • An interrogator is connected to one person and one machine via a terminal, and therefore can't see his counterparts. • The test is to find out which of the two candidates is the machine, and which is human only by asking them questions. 17 CSNB234 Artificial Intelligence Turing Test • If the interrogator cannot make a decision within a certain time (Turing proposed 5 minutes, but the exact amount of time is generally considered irrelevant), – the machine is considered to be intelligent. • If the computer succeeds in fooling the interrogator, i.e. the interrogator cannot distinguish the machine from the human, then – the machine may be assumed to be “intelligent” 18 CSNB234 Artificial Intelligence Turing Test 19 CSNB234 Artificial Intelligence Others AI Theorists • Other AI Theorists: – McDermott, Patrick Winston, Newell, Simon, Rosenblatt – & more (perform an internet search).. • Warren McCulloch (Columbia University) – Human Brain • Claude Shannon (Bell Lab) – Boolean Algebra • Norbert Wiener (MIT) – Mathematician and philosopher • John McCarthy (Dartmouth College) – Computer scientist and cognitive scientist • Marvin Minsky (MIT, graduated in Harvard University) – AI scientist August 9, 1927 – January 24, 2016 20 CSNB234 Artificial Intelligence AI History 1950: Turing Test Result: General problem-solving methods 1956: Dartmouth Conference proposed launch of Joint Research on AI. •John McCarthy, Marvin Minsky, Claude Shannon among the attendees. 1960s: AI established as research field. Focus on knowledge bases started. Areas of interests are chess games, theorem proving and language translation. 1963: Newell & Simon built General Problem Solver (GPS). 1965: DENDRAL developed by Feigenbaum at Stanford University. •Lisp developed by John McCarthy. Result: Knowledgebased expert systems 21 CSNB234 Artificial Intelligence Early Birth of AI Program: Eliza • ELIZA is an early natural language processing computer program created from 1964 to 1966 at the MIT Artificial Intelligence Laboratory by Joseph Weizenbaum 22 CSNB234 Artificial Intelligence AI History 1990: Intelligent agents 1972: PROLOG developed by Alain Colmerauer at University of Marseilles. 1970s: AI commercialization began. MYCIN developed at Stanford University, utilised production rules (diagnosed infectious blood diseases). 1981: Artificial neural networks. ICOT (Institute for New Generation Computer Technology) invented Concurrent Prolog for concurrent programming and parallel execution Result: Software that performs assigned tasks on the users behalf Result: Resembling the interconnected neuronal structures in the human brain Result: Transaction processing and decision support systems using AI. 23 CSNB234 Artificial Intelligence 24 CSNB234 Artificial Intelligence Conventional Systems vs AI Conventional Systems • Procedural • Numerical processing • Algorithmic • Rigid syntax AI Systems • Declarative • Symbolic processing • Heuristic programming • More natural syntax 25 CSNB234 Artificial Intelligence Regular Programming vs AI Programming Regular Programming = Algorithmic • Input: sequence of alphanumeric symbols • Processing: manipulation of the stored symbols by a set of algorithms • Output: sequence of alphanumeric symbols on such a medium as a screen, paper, or disk AI Programming = Heuristics • Input: sight, sound, touch, smell or taste • Processing: knowledge representation and pattern matching, search, logic, problem solving & learning • Output: printed language and synthesized speech, manipulation of physical objects or locomotion i.e., movement in space 26 CSNB234 Artificial Intelligence Symbolic Processing It is a branch of Computer Science that deals with symbolic, non-algorithmic methods of problem solving. Heuristics • It is the branch of Computer Science that deals with ways of representing knowledge using symbols rather than numbers and with rules-of-thumb for processing information. • Developed through intuition, experience & judgment. • Do not represent (our) knowledge of design, rather, they represent guidelines through which a system may be operated. • Often called “Rules of thumb”. • Characteristics Screening Filtering Pruning 27 CSNB234 Artificial Intelligence Heuristic Programming Should not be confused with computer programming. A program is a solution; programming is a procedure for obtaining a solution. A heuristic programming employs a practical method, not guaranteed to be optimal, perfect, logical, or rational, but instead sufficient for reaching an immediate goal. It is important to highlight that Heuristics are the strategies derived from previous experiences with similar problems. 28 CSNB234 Artificial Intelligence Language Levels for AI Problem Solving Symbol Level • Concerns with the particular formalisms used to represent knowledge such as logic or production rules. • Concerns with the structures used to organize knowledge. Knowledge Level • What queries / questions will be asked? • How new knowledge can be added or updated? • What objects and relations are necessary? • Can the system reasons despite of incompleteness of information? 29 CSNB234 Artificial Intelligence AI Systems Development Immature but can be used (tested) Knowledge and expertise slowly building up.. This methodology is called Rapid Prototype 30 CSNB234 Artificial Intelligence Essential Requirements for AI Language Support of Symbolic Computation Flexibility of Control Late Binding & Constraint Propagation Support of Exploratory Programming Methodologies Clear and Welldefined Semantics 31 CSNB234 Artificial Intelligence Essential Requirements for AI Language 1 Support of Symbolic Computation • Implementation of a set of operation on symbolic rather than numeric data. • Predicate calculus is a powerful tool for constructing qualitative descriptions of a domain. 32 CSNB234 Artificial Intelligence Essential Requirements for AI Language 2 Flexibility of Control • Rule-based systems being the most important paradigm for building AI programs. • AI cannot be achieved through step-by-step execution of a fixed sequence of instructions . • Production rules can be fired in virtually any order (i.e. not step-by-step) in response to a given situation. 33 CSNB234 Artificial Intelligence Essential Requirements for AI Language • AI programs seldom respond to standard software approaches such as top-down design, stepwise refinement. • This is due to the nature of AI problems that they could be started & tested without having to completely produce the final specification. • In other words, most AI programs are initially poorly specified. • AI programming is inherently exploratory; the program is the vehicle through which we explore the problem area (domain) and discover solution strategies. 3 Support of Exploratory Programming Methodologies 34 CSNB234 Artificial Intelligence Essential Requirements for AI Language • Often, the problems addressed by AI program (such as Prolog program) require that the values of certain entities to remain unknown until sufficient information is gathered to determine the assignment. 4 Late Binding & Constraint Propagation • As constraints are accumulated, the set of possible values is reduced, ultimately converging on a solution. 35 CSNB234 Artificial Intelligence Essential Requirements for AI Language • Traditional computer languages are too complex in its programming constructs and semantic definitions. They are not subject to self-proof. • This could be achieved by developing new languages that do not (to certain extent) conform to the architecture underlying von Neumann computer and be on the foundation of mathematical formalisms such as logic (Prolog). 5 Clear and Welldefined Semantics 36 CSNB234 Artificial Intelligence AI Languages • In Europe and Japan, Prolog is the preferred choice while in America, LISP is usually the way to go. Prolog • Good for rapid prototyping. • Possible to write algorithms by augmenting logical sentences with information to control the inference process. Lisp • Flexible • Allows it to adapt as programming styles change. • It made complex programs easy and fast to write. 37 CSNB234 Artificial Intelligence 38 CSNB234 Artificial Intelligence Overview of AI Application Areas Game Playing Automated Reasoning and Theorem Proving Expert Systems Natural Language Understanding and Semantic Modeling Modeling Human Performance Planning and Robotics Languages and Environments for AI Machine Learning Alternative Representations: Neural Nets and Genetic Algorithms • AI and Philosophy • • • • • • • • • 39 CSNB234 Artificial Intelligence AI in Finance Trading Agent • Rates, news – Automated financial advisors and planners that assist users in making financial decisions – Smart wallets that monitor and learn users’ habits and needs and alert and coach users – Data-driven AI applications to make better informed lending decisions Stock market Bonds market Trades Commodities market Personalized Financial Services • New Management Decision Making – Analyze data to come up with recommended decisions • Reducing Fraud and Fighting Crime – learn and monitor users’ behavioral patterns to identify anomalies and warning signs of fraud attempts and occurrences, along with collection of evidence necessary 40 CSNB234 Artificial Intelligence AI in Robotics Camera, mic, touch Motors, voice, move wheel / arm • Most robots perform repeating tasks without ever moving an inch. • Autonomous robots are self supporting or in other words self contained. In a way they rely on their own brains. 41 CSNB234 Artificial Intelligence AI in Games Game Agent Your move You / Player Its own move • Smart objects are used to help implement the behaviors. • The object specifies how each character interacts with it, which has many scalability and workflow advantages over centralized logic. 42 CSNB234 Artificial Intelligence AI in Medicine Diagnostic Agent • Fast and accurate diagnostics Blood pressure, heart signal – ability to learn from past cases You / Doctor Diagnostics • Reduce errors related to human fatigue – assisting doctors by eliminating human error and relieving them of time consuming, monotonous tasks • Decrease in medical costs – Patients would be asked to submit data more frequently via online medical records, and the improved line of communication could result in less hospital visits 43 CSNB234 Artificial Intelligence AI in Website Crawler Web pages List of websites Query WWW • Crawl the web • AI understand what words you typed in and find the most relevant pages • Retrieve pages – stores in big DB inside crawler & analyze relevant pages for any possible queries 44 CSNB234 Artificial Intelligence Summary • Artificial Intelligence (AI) Background and History • AI Languages • Overview of AI Application Areas