Introduction to Artificial Intelligence PDF

COE 292 Introduction to Artificial Intelligence Introduction 2/4/2025 COE292 - INTRODUCTION TO AI 1 What is AI? What is Artificial Intelligence? What are the Application Areas of AI? A (Short) History of AI AI Solution Strategies 2/4/2025...

COE 292 Introduction to Artificial Intelligence Introduction 2/4/2025 COE292 - INTRODUCTION TO AI 1 What is AI? What is Artificial Intelligence? What are the Application Areas of AI? A (Short) History of AI AI Solution Strategies 2/4/2025 COE292 - INTRODUCTION TO AI 2 What is AI? The definition of AI has been evolving with the advancements in technology Artificial Intelligence is the theory and development of computer systems able to perform tasks normally requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages 2/4/2025 COE292 - INTRODUCTION TO AI 3 What are the application Areas of AI? # Area/Field/Things # Area/Field/Things 1 Virtual Assistant or Chatbots 7 Manufacturing and Production 2 Agriculture and Farming 8 Livestock and Inventory Management 3 Autonomous Flying 9 Self-driving Cars or Autonomous Vehicles 4 Retail, Shopping and Fashion 10 Healthcare and Medical Imaging Analysis 5 Security and Surveillance 11 Warehousing and Logistic Supply Chain 6 Sports Analytics and Activities Any area where reasoning is to be performed by a machine 2/4/2025 COE292 - INTRODUCTION TO AI 4 Natural Language  Speech technologies (e.g. Siri)  Automatic speech recognition (ASR)  Text-to-speech synthesis (TTS)  Language processing technologies  Question answering  Machine translation  Web search  Text classification, spam filtering, etc…  General Purposed Language Model?  GPT-3 by OpenAI: better funny stories here https://www.gwern.net/GPT-3  Jukebox by OpenAI: interesting music here https://openai.com/blog/jukebox/ 2/4/2025 COE292 - INTRODUCTION TO AI 5 Computer Vision Image Captioning Image Segmentation Facial Recognition (security) 2/4/2025 COE292 - INTRODUCTION TO AI 6 Robotics  Robotics  Part Mech. Eng. and Part AI  Reality much harder than simulations!  Technologies  Self-driving cars  Rescue  Help in the home  Lots of automation… 2/4/2025 COE292 - INTRODUCTION TO AI 7  Decision Making Applied AI involves many kinds of automation  Scheduling, e.g. airline routing, military  Route planning, e.g. Google maps  Medical diagnosis  Web search engines  Automated help desks  Fraud detection  Product recommendations  … Lots more! This slide was created by Dan Klein and Pieter Abbeel for CS188 Intro to AI at UC Berkeley 2/4/2025 COE292 - INTRODUCTION TO AI 8 A (Short) History of AI Event in History​ Time Period Alan Turing Wrote the Turing Test​ 1950s​ Marvin Minsky: Steps towards Artificial Intelligence​ 1960s​​ Symbolic Integration by James Slagle​ Expert and Ruled based systems​ 1970s​ Deep Blue Systems (AI) beats world champion of chess​ 1980s-1990s​ Robotics, Computer Vision, Natural Language Processing, Deep 2000-now​ Learning... etc.​ 2/4/2025 COE292 - INTRODUCTION TO AI 9 Acting like a Human – Turing Test Scientific start of artificial intelligence (AI) which laid the foundation of a method to determine if a computer can behave like a human being What is the Turing Test: If you don't know who is on the other side, Can you tell the machine from a human by interacting only? 2/4/2025 COE292 - INTRODUCTION TO AI 10 www.educba.com Turing Test  Assesses the ability of an agent to exhibit intelligence.  Computer programs are considered to think if their responses can deceive a human into perceiving them as human.  Not everyone accepts the validity of the Turing Test, but passing it remains a challenge to developers of AI.  Turing test limitations include:  Requires a controlled environment,  Lacks the implicit definition of intelligence, and  Needs adaptation to new advancements COE292 - INTRODUCTION TO AI 11 ELIZA Chatbot In 1966, Prof. Weizenbaum at MIT developed the program ELIZA, which aimed at tricking its users by making them believe that they were having a conversation with a real human being. ELIZA was designed to imitate a therapist who would ask open- ended questions and even respond with follow-ups. Try it out: http://psych.fullerton.edu/mbirnbaum/psych101/eliza.htm 2/4/2025 COE292 - INTRODUCTION TO AI 12 How AI was Viewed over Time 2/4/2025 COE292 - INTRODUCTION TO AI 13  What is AI? The science of making machines that: (2) Think like (1) Think rationally people (3) Act like people (4) Act rationally This slide was created by Dan Klein and Pieter Abbeel for CS188 Intro to AI at UC Berkeley 2/4/2025 COE292 - INTRODUCTION TO AI 14 Definitions Agent: A thing that takes inputs using sensors and acts based on these inputs. Utility in AI: Preferred Outcome or Result. In other words, the mathematical treatment of “preferred outcomes” is called utility. Example: An agent's preferences over possible outcomes can be captured by numbers; the higher the number the more that agent likes that outcome. 2/4/2025 COE292 - INTRODUCTION TO AI 15 Definitions Rational Agent: is one that acts to achieve the best outcome or, when there is uncertainty, the best expected outcome. A rational agent often has clear preference. It models uncertainty and acts in a way to maximize its Utility (performance) with all possible actions. 2/4/2025 COE292 - INTRODUCTION TO AI 16 Designing Rational Agents An agent is an entity that perceives and acts. A rational agent selects actions that maximize its (expected) utility. Characteristics of the percepts, environment, and action space dictate techniques for selecting rational actions 2/4/2025 COE292 - INTRODUCTION TO AI 17 Human Brains Very good in making rational decisions but not perfect It is not completely understood how human brain works. Difficulty in copying its behavior to a machine How brain works? Has a huge amount of data (lessons learned across the life of a human) Can simulate, compute or predict the outcome based on available data or by inference. i.e., unroll the consequences without trying them. 2/4/2025 COE292 - INTRODUCTION TO AI 18 What is this Course About 1. Get introduced to a suite of representations that will help you make programs that are intelligent 2. Use representations to model a problem and produce useful solutions or predictions 2/4/2025 COE292 - INTRODUCTION TO AI 19 You will Learn Constructing Algorithms (or methods, or procedures) by Exposing the constraints of the problem and Using a representation that supports making of models which facilitate understanding of: 1. Perception 2. Thinking 3. Action 2/4/2025 COE292 - INTRODUCTION TO AI 20 AI Solution Strategies Next, we will show the application of two AI strategies in solving problems: Use of Perception to solve problems Use of “Generate and Test” strategy to solve problems 2/4/2025 COE292 - INTRODUCTION TO AI 21 Example1: Use of Perception to Solve Problems Problem: A person has one pet lion, one lamb and a bundle of grass. He wants to cross a river but there is only one boat, and it can't sustain the weight of more than two articles at time. Also, he must make sure that the lion doesn't eat the lamb and the lamb doesn't eat the grass. The questions we want to answer are: 1. How will he get to the other side of the river with all three living beings intact?. 2. How will you tell (or program) a computer to solve this problem? (the bigger question) 3. How many possible ways are there to solve such a problem? 2/4/2025 COE292 - INTRODUCTION TO AI 22 Example1: Use of Perception to Solve Problems First, we must understand the problem Using the right perception will aid our understanding What is the best representation for the problem? Visual perception: drawing the situation so we use our visual capabilities 2/4/2025 COE292 - INTRODUCTION TO AI 23 Example1: Use of Perception to Solve Problems How many more States (or situation) can there be? List of all states is shown below: 2/4/2025 COE292 - INTRODUCTION TO AI 24 Example 1: Use of Perception to Solve Problems Expose constraints Lion will eat sheep  state not allowed () Sheep will eat grass  state not allowed () We now detect those states in which we find someone eating someone else and delete them 2/4/2025 COE292 - INTRODUCTION TO AI 25 Example 1: Use of Perception to Solve Problems Identify the states that violate constraints 2/4/2025 COE292 - INTRODUCTION TO AI 26 Example 1: Use of Perception to Solve Problems Removing the states with anomaly will yield 2/4/2025 COE292 - INTRODUCTION TO AI 27 Example 1: Use of Perception to Solve Problems  Add the final constraints Make your state equal to the initial State Repeat until you reach to the final state: From one state to another, If you are on the left side of the only one OR no item can river, then accompany the person Next state = State with single Reorder states to go from item moved with you to the other side initial to final states with Else the above constraint Next state = State with you  Apply Reordering moving to the other side with or without item Algorithm End If End Repeat 2/4/2025 COE292 - INTRODUCTION TO AI 28 Example 1: Use of Perception to Solve Problems Available States Make your state equal to the initial State Repeat until you reach to the final state: If you are on the left side of the river, then Next state = State with single item moved with you to the other side Else Solution Next state = State with you moving to the other side with or without item End If End Repeat 2/4/2025 COE292 - INTRODUCTION TO AI 29 Example 1: Use of Perception to Solve Problems Available State Make your state equal to the initial State Repeat until you reach to the final state: If you are on the left side of the river, then Next state = State with single item moved with you to the other side Else Solution Next state = State with you moving to the other side with or without item End If End Repeat 2/4/2025 COE292 - INTRODUCTION TO AI 30 Example 1: Use of Perception to Solve Problems Available State Make your state equal to the initial State Repeat until you reach to the final state: If you are on the left side of the river, then Next state = State with single item moved with you to the other side Else Solution Next state = State with you moving to the other side with or without item End If End Repeat 2/4/2025 COE292 - INTRODUCTION TO AI 31 Example 1: Use of Perception to Solve Problems Available State Make your state equal to the initial State Repeat until you reach to the final state: If you are on the left side of the river, then Next state = State with single item moved with you to the other side Else Solution Next state = State with you moving to the other side with or without item End If End Repeat 2/4/2025 COE292 - INTRODUCTION TO AI 32 Example 1: Use of Perception to Solve Problems Result 2/4/2025 COE292 - INTRODUCTION TO AI 33 Example 2: Generate and Test in AI Problem: What is the name of the fish shown? Before thinking of AI and if you don’t already know it, what would you do? 2/4/2025 COE292 - INTRODUCTION TO AI 34 Example 2: Generate and Test in AI Go online and search for the fish by flipping web pages until it is found Try to find a match by extracting some features (shape, size, color … etc.) and matching it with pictures of the fish on the web This step is a form of feature extraction Good features are often discriminating features and are unique about the object under consideration. 2/4/2025 COE292 - INTRODUCTION TO AI 35 Example 2: Generate and Test in AI Examples of Good features (for the fish under investigation): Has a square head (since not many fish have square shaped head) Comes in different colors and has patterns on its body Mouth is pointing downwards (many fish don't have pointy mouth) Eyes on the side of the square head (eyes are close to the head edge) … etc. 2/4/2025 COE292 - INTRODUCTION TO AI 36 Example 2: Generate and Test in AI Examples of Bad features (for the fish under investigation): Lives in water or sea (all fish live in water or sea, so this feature does not add any value) Swims close to rocks (features should give unique information about the object that help in identifying fish, again this is something common to all fish) Lives only in lighted area (to take good pictures we need to have a light but that does not mean that this fish cannot live in the dark) 2/4/2025 COE292 - INTRODUCTION TO AI 37 Example 2: Generate and Test in AI Generally, Good features Should take you closer to the target, while bad features do not add any value Should help us or the computer to identify unique things about the fish. Since one feature may not be enough to identify the fish, the goal is to have the minimum number of features to uniquely identify the fish. 2/4/2025 COE292 - INTRODUCTION TO AI 38 Example 2: Generate and Test in AI Procedure 2/4/2025 COE292 - INTRODUCTION TO AI 39 Generate and Test Properties A generator must have good properties: 1. Should not be redundant: i.e., should not generate the same case more than once.  In the example of the fish, we don't want to test the same fish more than once. 2. Should be informable: supply knowledge related to the matter or subject in hand.  In the fish example, we don't want the generator to produce pictures of dogs, cats, planes or cars Extract meaningful test features to determine success vs failure Test should be fast and efficient 2/4/2025 COE292 - INTRODUCTION TO AI 40 Summary We introduced the subject of AI Went over some definitions Agents in AI Rationality in AI Rational Agents We presented two ways to solve problems Using representation and constraints to find a solution Generate and test methodology 2/4/2025 COE292 - INTRODUCTION TO AI 41