Artificial Intelligence - The Very Idea
|
Study Guide
Please note that the precise schedule is subject to change. The lecture slides (and lecture notes, if any) are updated after the lecture.
Lecture 1. Artificial Intelligence - The very idea
Course overview. What is AI about? Why should we care? A little history of artificial intelligence and computer science. Hobbes, Leibniz, Hilbert, Turing, and the Dartmouth Workshop - the birth of AI. Goals of AI. Working hypothesis of AI.
Required readings
Recommended materials
Lecture 2. Natural and Artificial Intelligence
What is intelligence? Theories and measures of intelligence. Turing Test. Searle's Objection (Chinese Room Argument).
Required materials
Recommended materials
Lecture 3. Natural and Artificial Intelligence (continued)
Variants of the Turing Test - Winograd's Schema Challenge, Visual Turing Test, Video Turing Test, Lovelace Test, Physically Embodied Turing Test, Total Turing Test (Harnad Test). Animal Intelligence.
Required materials
Recommended materials
- Read: Artificial Intelligence - The Very Idea (Draft) Chapter 2 Vasant Honavar
- Watch: Rethinking Thinking: How Intelligent Are Other Animals?, World Science Festival
- Read: Turing Test: 50 Years later, Saygin A., Cicekli, I. and Akman, V. (2000). Minds and Machines, Kluwer.
- Read: Chinese Room Argument Stanford encyclopedia of philosophy.
- Read: AI Magazine Special Issue: Beyond Turing Test, Gary Marcus, Francesca Rossi, Manuela Veloso (Ed), 2016
Lecture 4. Computation, Machines, Languages
Origins of the Theory of Computation. Algorithms and Computation. Boolean Logic, Finite Automata and Turing machines, Algorithms, Languages, and Programs. Turing Machines. Universality of Turing Machines. Church-Turing Thesis. Limits of Computation.
Required materials
Recommended materials
Lecture 5. On Artificial Intelligence
Origins of AI. Functionalist view of intelligence. Computational theory of mind. Brief recent history of AI from the Dartmouth workshop to present. Some early lessons of AI.
Required materials
Recommended materials
Lecture 6. On Artificial Intelligence - continued
Different Goals of AI. Relation of AI to other disciplines. AI for building intelligent agents. What is an agent? What is a rational agent? Role of performance measure.
Required materials
Recommended materials
Lecture 7. Intelligent Agents
Agents and their environments. What is an agent? What is a rational agent? Types of agents: Simple reflex agents, Reflex agents with state; Goal-based agents; Optimal plan seeking agents; Utilitarian agents; Learning Agents.
Required materials
Recommended materials
- Read: Wooldridge, Michael Intelligent Agents In: Multiagent systems: A modern approach to distributed artificial intelligence 1 (1999): 27-73.
- Read: Holland, J. Genetic AlgorithmScientific American, vol. 267, pp 262-273. 1992
- Read: Daniel C. Dennett Kinds of Minds, 1997, Basic Books
Lecture 8. Knowledge Representation
What is knowledge representation? Form versus content of representation. Role of knowledge representation as a proxy or surrogate for the real world. Semantics of representation. Why all representations are bad but some are useful. Examples of representation and the types of inference they support. Feature space, relational network, state transition network, production rules, frames, logic, procedures, isomorphic representations.
Required materials
Recommended materials
- Read: Davis, Randall, Howard Shrobe, and Peter Szolovits. What is a knowledge representation?. AI magazine 14, no. 1 (1993): 17-.
- Read: Dretske, Fred. Machines and the Mental. In Proceedings and Addresses of the American Philosophical Association, vol. 59, no. 1, pp. 23-33. American Philosophical Association, 1985.
- Read: Harnad, S., 1990. The symbol grounding problem. Physica D: Nonlinear Phenomena, 42(1-3), pp.335-346.
Lecture 9. Problem Solving Machines
Goal-Based Agents. Problem-solving as state space search. Basic search algorithms - Breadth-first search, depth-first search, interative deepening search. Finding optimal solutions - Branch and bound search
Required materials
Recommended materials
Lecture 10. Problem Solving Machines (continued)
Informed search methods. Heuristics. How to design heuristics. How to use heuristics. Best-first search. A* search. Properties of A*. Variants of A*.
Required materials
Recommended materials
- Read: Korf, Richard. Artificial Intelligence Search Algorithms, In: Artificial intelligence search algorithms. CRC Handbook of Algorithms and Theory of Computation. CRC Press, 2010
- Read: Ginsberg, Matt. Essentials of Artificial Intelligence, Chapter 4, Morgan Kaufmann, 1994
- Read: Chang, C.L. and Slagle, J.R., An admissible and optimal algorithm for searching AND/OR graphs. Artificial Intelligence, 2(2), pp.117-128.
Lecture 11. Problem Solving Machines (continued)
Problem reduction representation. AND-OR trees. Extensions of basic (uninformed) search methods to AND-OR graphs. Extensions of branch-and-bound and A* search to AND-OR graphs (AO* algorithm for finding optimal problem reduction solutions.
Constraint satisfaction. Examples of constraint satisfaction problems. Special properties of constraint satisfaction problems. Variants of depth-first search with backtracking for efficient solution of constraint satisfaction problems.
Required materials
Recommended materials
Lecture 12-13. AI and society (Guest Lectures by Dana Calacci and Sarah Rajtmajer)
AI and the future of work.
AI and privacy.
Required materials
Recommended materials
Lecture 14. On reasoning with knowledge. Rule-based expert systems. IF-then rules. Reasoning using forward chaining and backward chaining. Rule-based deductive versus reactive systems. Structure of expert systems - knowledge base, working memory, inference engine, agenda etc. Rule matching, Rule triggering, rule firing. Expert systems in action. Examples. Generating explanations. Increasing efficiency using Rete network. Expert systems in practice. Representative commercially successful expert systems e.g., R1. Knowledge engineering bottleneck.
Required materials
Recommended materials
Lecture 15-16. On reasoning with knowledge. Knowledge representation using propositional logic; Syntax and (Model-based or Possible Worlds) Semantics; Basic laws of propositional logic. Logical notions of Truth and Falsehood; Logical Entailment; Inference rules; Modus ponens; Soundness and Completeness of inference rules. Resolution principle. Theorem-proving as search. Forward and backward chaining.
Required materials
Recommended materials
Lecture 17. Beyond Propositional logic. Introduction to first-order predicate logic (FOPL). Terms, Predicates, Relations, Functions. Expressing assertions in FOPL. Resolution proofs in FOPL.
Required materials
Recommended materials
Lecture 18-19. Representing and reasoning under uncertainty. Generalizing propositional logic to probability theory. Syntax and (possible world) semantics. Probability, joint probability and conditional probability. Marginalization, Bayes rule, independence and conditional independence. Representing and reasoning with probability distributions using Bayes networks.
Required materials
Recommended materials
Lecture 20-21. Learning. Why should machines learn? What is machine learning? Major types of machine learning. Supervised learning from labeled examples. Nearest neighbor classifier. Naive Bayes classifier. Decision tree learner - playing the game of 20 questions against nature. Information and entropy. Linear classifier. Limitations of linear classifiers and how to overcome them.
Required materials
Recommended materials
Lecture 22-23. Language Models. What are language models? The power of next word prediction. Applications of language models. Simple language models. Low-order Markov models - unigrams, bigrams, trigrams.. N-grams. Large language models (LLM). Transformer architecture? Power of scale. Prompting LLMs. Customizing LLM through fine-tuning. In context and few-shot training of LLMs. Chain-of-thought prompting. What LLMs are and what they are not. Good and bad uses of LLM.
Required materials
Recommended materials
|
