Artificial Intelligence Its Roots and Scope 1 1

Artificial Intelligence: Its Roots and Scope 1. 1 1. 2 From Eden to ENIAC: Attitudes 1. 3 Artificial Intelligence – A Summary toward Intelligence, 1. 4 Epilogue and References Knowledge, and Human Artifice 1. 5 Exercises Overview of AI Application Areas George F Luger ARTIFICIAL INTELLIGENCE 6 th edition Structures and Strategies for Complex Problem Solving Luger: Artificial Intelligence, 6 th edition. © Pearson Education Limited, 2009

A Definition of Artificial Intelligence An area of study concerned with understanding the activities of the human mind and replicating those activities on a computer. Note that we do not attempt to define “intelligence”.

Chapter 1: Early History and Applications 1. 1 From Eden to ENIAC: Attitudes toward Intelligence, Knowledge, and Human Artifice • Aeschylus [c. 525/524 – c. 456/455 BC], Myth of Prometheus • Gave humans the gift of nous: the rational mind • This angered Zeus, because he understood that the quest for intellectual knowledge can have adverse consequences

• Biblical Story of Eden [6 th century BC] • Adam and Eve ate fruit from the “tree of knowledge” • This angered God, for basically the same reason as in the myth of Prometheus • The belief that the desire for knowledge can lead to disaster has persisted throughout history. • Mary Shelly [1797– 1851], Frankestein: The Modern Prometheus, published in 1818 • Frankenstein employed scientific knowledge to create a living being that became a monster

• Modern Examples • Weapons of Mass Destruction • Nuclear • Biological • Surveillance State Technology: Autocracy • George Orwell’s 1984 (published 1949), “Big brother is watching you”, is now a reality • Invasion of Privacy • “Tallahassee Democrat” cartoon, 8/19/2020

• These same concerns apply to the emergence of AI • Many authors have warned about the dangers • The potential for human misuse of intellectual knowledge is no different now than it was 2500 years ago. • But the potential for disaster is far greater.

1. 1. 1 A Brief History of the Foundations for AI • Aristotle [384 -322 BC] • Science as the “study of things that change” • Distinction between “form” and “substance” • Substance, like the bronze in a statue, can be molded into a different form • Logic, syllogism, deductive forms • All men are mortal; Socrates is a man; therefore Socrates is mortal. • Modus Ponens: From P and P -> Q, infer Q • Modus Tollens: From ~Q and P -> Q, infer ~P • Surprisingly, no progress was made on the advancement of formal logic for the next 2000 years • Leibnitz, Boole, Frege; see the following

• Copernicus [1473 -1543] • Replaced earth-centered view of the universe with the solar-centered • The planets, including Earth, revolve around the sun • Contradicted established religious beliefs, which claimed that the world was created for the purpose of humans, with “man” as the center • Radical change in people’s awareness and perception of their relationship to their world • Separation between ideas and the world • Things are not necessarily as they appear • Bacon [1561 -1626] • The “form” of an object is the sum of its “features” • Has played a role in modern AI knowledge representation systems 8

• Schickard [1592 -1693] • Calculating Clock • A machine with gears that performed addition and rang a bell on each carry operation • Pascal [1623 -1662] • The Pascaline • Similar to Schikard’s calculating clock • Did both addition and subtraction • Used Bacon’s ideas as a basis for reasoning about objects • Leibnitz [1646 -1716] • Leibnitz Wheel • Addition, subtraction, multiplication, and division • Considered the possibility of automating logical reasoning

• Descartes [1596 -1650] • Distinction between “mind” and “body” • Mental: res cogitans, meaning mental (nonphysical) substance • Physical: res extensa, meaning extended thing • Each can be studied independently of the other • But how are they related? • By analogy with modern computing: mind is an information process running on the physical brain • Hobbes [1588 -1679] • “By ratiocination I mean computation” • “ratiocination” means “the process of exact thinking” • The view was emerging that mental processes, like physical processes, could be characterized through formal mathematics

1. 1. 2 AI and the Rationalist and Empiricist Traditions • Rationalists: Pascal, Descartes, Leibnitz • The world can be described by mathematical models • Empiricists: Hobbes [1558 -1679], Locke [1632 -1704], Hume [1711 -1776] • Knowledge begins with sense experience • It arises through associations established by repeated experiences (learning) • Kant [1724 -1804] • Synthesized the two views • Distinguished between two types of knowledge • a priori, prior to experience, innate • capacity for logic, mathematics, conceptual abstraction • a posteriori, coming from experience

1. 1. 3 Development of Formal Logic • Leibnitz [1646 -1716] • Calculus Philosophicus • Introduced the first system of formal logic • Proposed a machine for automating logical reasoning • Euler [1707 -1783] • Graph theory • In AI, the foundation for state-space search

• Babbage [1791 -1871] • Difference Engine (partially built) • Computed the values of certain polynomials • Operated similarly to a weaving machine (the Jacquard loom), with instructions and data encoded on cards • Analytical Engine (designed but not built) • A general-purpose computing machine • Employed a “store” (memory) and a “mill” (processor) • Lovelace [1815 -1852] (Ada King, Countess of Lovelace) • Babbage’s friend, supporter, and collaborator • Considered to be the first computer programmer

• Boole [1815 -1864] • An Investigation of the Laws of Thought, on which are founded the Mathematical Theories of Logic and Probabilities (1854) • Boolean algebra and Boolean logic • Frege [1848 -1925] • Begriffschift (roughly, “concept script”), a formula language, modeled upon that of arithmetic, for pure thought (1879) • Foundations of Arithmetic • First complete formal logical system • Extended Boole by adding the implication operator, -> • Introduced quantifiers, “exists”, “for all” • Two dimensional diagrams for representing logic assertions and proofs

Frege’s definition of the implication operator (conditional)`

This describes to the following truth table: A B B -> A T T F T F F F T which is equivalent with the standard truth table: A B A -> B T T F F F T T F F T

• Whitehead [1861 -1947] and Russell [1872 -1947] • Principia Mathematica (3 vols. : 1910, 1912, 1913) • Revised (1925 -27) • Rewrote Frege’s system using a simple symbolic language (replacing the 2 -D diagrams) • Statements are represented as linear strings • Logical derivations are operations performed on strings (formal logic) • Nowadays called “First-Order Predicate Logic” (which we will study later in this course) • Showed that one can formulate all of the mathematical Theory of Sets in such a system • Set theory is the foundation for all of modern mathematics • This is why first-order logic has been adopted as a basic AI knowledge representation framework

• Tarski [1904 -1983] • Principia Mathematica is purely formal in that it works exclusively with uninterpreted symbolic strings • Tarski provided this with a semantics, comprising welldefined meanings for the symbols and expressions. • “Snow is white” it true if and only if snow is white. • This also will be studied later in this course • Von Neuman [1903 -1957] • Designed the basic computer architecture (following Babbage) employing a memory and processor • Was instrumental in creating the first digital computer • Separation of data and programs • Programs as data stored in memory, which take other data as input and produce data as output • Saw that programs can do both arithmetic and logical inference (early AI)

• Some views that emerged in the 1940’s • Intelligence is a form of information processing • Knowledge can be represented in formal language and manipulated by algorithms • Logical reasoning can be automated

1. 1. 4 The Turing Test • • Named after Alan Turing [1912 -1954] “Computing Machinery and Intelligence”, Mind, 1950 Described as an “Imitation Game” (1951) Attempts to give an objective notion of intelligence without becoming mired in the task of defining the meaning of “intelligence’’ • Provides the basis for many schemes used to evaluate modern AI programs (e. g. , expert systems)

Fig 1. 1 The Turing test. Luger: Artificial Intelligence, 6 th edition. © Pearson Education Limited, 2009

• Subject to various criticisms • Bias toward purely symbolic problem solving • Does not test perceptual skill or manual dexterity • Ford and Hayes, Turing Test Considered Harmful, 1999 • Trying to pass the Turing Test is not the proper aim of AI • Humans make mistakes, think slowly, are forgetful • The aim should be to augment human thinking, not to exactly mimic it

• Turing also addressed the issue of whether it is even possible to create a truly intelligent computer program • He raised two objections: • Lady Lovelace’s Objection, computers can only do what they are told to do, and therefore cannot be creative • Not everyone agrees with this • Argument from Informality of Behavior, it is impossible to create a set of rules that can tell an individual exactly what to do under every possible set of circumstances • A major challenge for self-driving cars

• Whether computer systems can exhibit the flexibility shown in living organisms is still subject to debate • Herbert Simon [1916 -2001] • The Sciences of the Artificial, 1981 • Regarding the behavior of ants: • As a physical system the ant is quite simple • The complexity of it’s behavior is function of its environment • Simon argued that this applies equally to humans • Intelligence is measured as an ability to interact with our environment • Culture plays a role in forming intelligence

1. 1. 5 Biological and Social Models of Intelligence: Agent Theories • Challenges to the rationalist tradition represented by Plato/Aristotle, Galileo/Copernicus, Descartes, Leibnitz, etc. • Break from reliance on mathematics and logic as a basis for knowledge representation and intelligent reasoning • Two philosophical traditions that emerged • Wittgenstein • Linguistic relativism: words do not have objective meanings; all meanings are subjective, relative to individual experience • Husserl and Heidegger • Phenomenology: knowledge is not knowing what is true, but knowing how to cope with a changing and evolving environment (adaptability)

• The rationalist/logical agenda is sometimes referred to as GOFAI, Good Old Fashioned AI • Alternatives that reflect the relativist and phenomenological views are sometimes called social, or agent-based, or situational theories • Emergent new models of computing • Connectionist learning: neural models of intelligence, neural nets, deep learning • Genetic algorithms (aka artificial life): knowledge evolves according to established principles • Social Systems: populations of individual acting independently exhibit collective intelligence • The latter leads to agent-based models used in: • Economics • Sociology

Important Research and Application Areas 1. 2. 1 Game Playing 1. 2. 2 Automated Reasoning and Theorem Proving 1. 2. 3 Expert Systems 1. 2. 4 Natural Language Understanding and Semantic Modeling 1. 2. 5 Modeling Human Performance 1. 2. 6 Planning and Robotics 1. 2. 7 Languages and Environments for AI 1. 2. 8 Machine Learning 1. 2. 9 Alternative Representations: Neural Nets and Genetic Algorithms 1. 2. 10 AI and Philosophy Luger: Artificial Intelligence, 6 th edition. © Pearson Education Limited, 2009

1. 2. 10 AI and Philosophy • Newell and Simon, 1976 Turing Award lecture • Physical Symbol System Hypothesis: the necessary and sufficient condition for a physical system to exhibit intelligence is that it be a physical symbol system. • Searle, 1980 Chinese Room argument • “Understanding” is essential to intelligence • Purely symbolic systems don’t have understanding • Therfore, Newell and Simon’s hypothesis is incorrect. • Symbolic representation and reasoning may be necessary, but are not sufficient for true intelligence

• Searle defined: • Strong AI , the hypothesis that true machine understanding/ intelligence is possible • Weak AI, the position that this can only be simulated • Argued that only weak AI is possible

The AI Hype • Since the 1950’s top researchers at top universities have been predicting that we will have computer systems that fully replicate human mental capabilities with “the next 20 years”. • Some instances, past and present: • 1961 MIT documentary “The Thinking Machine” • https: //techtv. mit. edu/videos/10268 -thethinking-machine-1961 ---mit-centennial-film • Herbert Simon (Nobel, Turing awards), 1963 • 1970’s and 1980’s: Expert Systems • Generated a lot of false expectations • Mid 1980’s and 1990’s: Neural Nets (learning) • Mid 2010’s: Deep Learning

• Some recent relevant publications: • See the link “The New AI Hype” on the course website • This article was published in IEEE Spectrum, December 2019 • G. Marcus and E. Davis, Rebooting AI: Building Artificial Intelligence We Can Trust, Vintage Book, 2019 ($16. 95 on Amazon)

Important Features of Artificial Intelligence 1. The use of computers to do reasoning, pattern recognition, learning, or some other form of inference. 2. A focus on problems that do not respond to algorithmic solutions. This underlies the reliance on heuristic search as an AI problem-solving technique. 3. A concern with problem-solving using inexact, missing, or poorly defined information and the use of representational formalisms that enable the programmer to compensate for these problems. 4. Reasoning about the significant qualitative features of a situation. 5. An attempt to deal with issues of semantic meaning as well as syntactic form. 6. Answers that are neither exact nor optimal, but are in some sense “sufficient”. This is a result of the essential reliance on heuristic problem-solving methods in situations where optimal or exact results are either too expensive or not possible. 7. The use of large amounts of domain-specific knowledge in solving problems. This is the basis of expert systems. 8. The use of meta-level knowledge to effect more sophisticated control of problemsolving strategies. Although this is a very difficult problem, addressed in relatively few current systems, it is emerging as an essential are of research. Luger: Artificial Intelligence, 6 th edition. © Pearson Education Limited, 2009