AIE111 Lecture 2 - Intelligent Agents PDF

Lecture 2: Intelligent Agents Dr. Laila Shoukry  Course Outline Tentative W L Lecture Topic CS50 (extra) Lab Courser...

Lecture 2: Intelligent Agents Dr. Laila Shoukry  Course Outline Tentative W L Lecture Topic CS50 (extra) Lab Coursera Assessments Python W1, W2, W3 Coursera submission (4pts) Grading 1 1 Introduction Python for AI 1 Intro to AI W1 Lab Work (1pt) 2 Agents Python W4, W5, W6 Coursera submissions (4 pts) Assessment % Grade 2 Python for AI 2 Intro to AI W2 Lab Work (1pt) Tutorials and Coursera certificate (2pts) Assignments 10% 3 3 Uninformed Search W0 Search 8-puzzle, Maze Escape Lab Work (1pt) Assign1 (2.5pts) on L3 Best 3 out of 4 Quiz1 (5pts) on L1,2,3 Quizzes 4 4 Informed Search W0 Search Maze Escape 2, Tic Tac Toe 15% Lab Quiz 1 (3 pts) on lab 1,2,3 Best 3 out of 4 Assign2 (2.5pts) on L4 Lab Work + Lab 5 5 Knowledge-based Agent Clue Knowledge Base Lab Work (1pt) Quizzes 10% Propositional Logic and Quiz2 (5pts) on L4,5 6 6 W1 Knowledge PL Inference in Python Best 2 out of 3 Inference Lab Quiz 2 (3 pts) on lab 4,5 7 Midterm Exam Midterm Midterm 15% First-Order Logic and Coursera 8 7 W1 Knowledge FOL Inference in Python Tut Work (1pt) Inference Submissions 10% Reasoning with Forward and Backward Lab Quiz 3 (3pts) on lab 6,8 and Certificate 9 8 W1 Knowledge Chaining using Python Assign3 (2.5pts) on L5,6,7 Certainty Best Quiz3 (5 pts) on L5,6,7,8 attendance and 10 9 Uncertainty W2 Uncertainty Probability in Python Lab Work (1pt) participation, Reasoning with 11 10 W2 Uncertainty Bayes Rule & Net using Python Assign4 (2.5pts)on L9,10 additional extrac Uncertainty 1 certificates, 11 Reasoning with Sampling & Likelihood 12 Weighting using Python Quiz4 (5pts) on L9,10 extracurricular Uncertainty 2 project 13 12 Planning Planning in Python (POP) Extrac Project Final 40% Questions on Last lecture What is Intelligence? Artificial Intelligence? Turing Test? Total Turing Test? Chinese Room Experiment? Machine Learning? An agent?  What is AI? https://www.wired.com/story/its-called-artificial-intelligence-but-what-is-intelligence/  Total Turing Test (TTT) https://www.gopichandrakesan.com/acting-humanly-the-turing-test-approach-artificial-intelligence/  Chinese Room Experiment https://www.infineon.com/cms/en/discoveries/definition-artificial-intelligence/ What is an agent? According to Russell & Norvig: Agents are entities that can be viewed as perceiving and acting upon their environment. Agents use sensors to perceive, and effectors/actuators to act. Examples: Animals. Robots. Calculators. Thermostats. Clocks.  Software Agents Software agents can have keystrokes or file contents as sensory input, act on those inputs and display output on the screen. Percepts https://dorozhko-anton.github.io/nsu-rl101.github.io/ Robotic Vacuum Cleaner Ad https://www.youtube.com/watch?v=fFy69i-QRYc  Was this an agent? https://medium.com/knoldus/ai-rational-agents-and-operating-environments-2bef1e3c3738  Structure of an AI Agent Agent = Architecture + Agent program Architecture: machinery that an AI agent executes on. Agent Function: to map a percept to an action. f: P* → A Agent program: an implementation of agent function. An agent program executes on the physical architecture to produce function f.  Agent Function maps a percept to an action. f: P* → A https://skilllx.com/intelligent-agent-in-artificial-intelligence/  Vacuum Cleaner Example Percepts: location and contents e.g. [A, Dirty] Actions: Left, Right, Suck, NoOp https://www.edx.org/course/artificial-intelligence-ai Simple Vacuum Cleaner Program https://www.edx.org/course/artificial-intelligence-ai Rational Agent An agent must strive to “do the right thing,” based on what he can perceive and the actions he can take. How did we measure the success? Robotic Vacuum Cleaner Example  How to measure rationality? Rationality is relative to a performance measure. Judge rationality based on: The performance measure that defines the criterion of success. The agent prior knowledge of the environment. The possible actions that the agent can perform. The agent’s percept sequence to date.  PEAS of A Rational Agent PEAS stands for: Performance Environment Actuators Sensors  PEAS of A Vacuum Cleaner Performance: cleanness, efficiency, obstacle detection, noise,.. Environment: wood floor, carpet, corners, different obstacles,… Actuators: wheels, brushes, vacuum,.. Sensors: camera, dirt detection, infrared, bump sensors PEAS of A Self-Driving Car https://skilllx.com/intelligent-agent-in-artificial-intelligence/ PEAS of Software Agents https://skilllx.com/intelligent-agent-in-artificial-intelligence/  Types of Environments Fully Obersvable An agent’s sensors give it access to the complete state of the environment at each point in time. https://tunpixelblog.wordpress.com/2016/02/25/fully-vs-partially-observable-environment/  Types of Environments Fully Obersvable An agent’s sensors give it access to the complete state of the environment at each point in time. https://www.perchingtree.com/intuition-difference-fully-partially-observable-environments-ai/  Types of Environments Partially Obersvable The agent needs a memory in order to make the best possible decision. https://tunpixelblog.wordpress.com/2016/02/25/fully-vs-partially-observable-environment/  Types of Environments Partially Obersvable The agent needs a memory in order to make the best possible decision. https://tunpixelblog.wordpress.com/2016/02/25/fully-vs-partially-observable-environment/  Types of Environments Deterministic The next state of the environment is completely determined by the current state and the action executed by the agent. https://skilllx.com/intelligent-agent-in-artificial-intelligence/  Types of Environments Stochastic (Non-Deterministic) A stochastic environment is random in nature and cannot be determined completely by an agent. https://skilllx.com/intelligent-agent-in-artificial-intelligence/ Nice debate about this topic: https://www.reddit.com/r/aiclass/comments/l66ss/question_about_deterministic_vs_stochastic/ Types of Environments Episodic vs. Sequential Episodic The agent’s experience is divided into atomic "episodes" (each episode consists of the agent perceiving and then performing a single action) https://slideplayer.com/slide/15212488/ Types of Environments Episodic vs. Sequential Sequential Current decisions affect future decisions or rely on previous ones (Coherent sequence of observations and actions) Types of Environments Static The Environment does not change while the agent is thinking Types of Environments Dynamic The Environment can change with time while the agent is thinking e.g. Taxi Driving Semi-Dynamic The environment itself does not change with the passage of time but the agent’s performance score does. e.g. Chess with Score Types of Environments Discrete One action or one measurement per step, Defined actions Types of Environments Continuous Actions continually changing Types of Environments Types of Environments Types of Agents Simple-Reflex Agents Simple-Reflex Agents No memory No planning Action based on current state only Simple-Reflex Agents Model-based Agents Model-based Agents Memory Model Goal-based Agents Goal-based Agents Planning based on goals Utility-based Agents Utility-based Agents Planning based on utility Learning Agents Learning Agents  Summary Comparing Intelligent Agents Performance Measures PEAS of Agents Types of Environments Types of Agents Thank you! Questions? http://mba-blog.audencia.com/2018/08/20/can-ai-save-humanity/  Our Focus in this course https://ec.europa.eu/futurium/en/system/files/ged/ai_hleg_definition_of_ai_18_december_1.pdf