CSE1100 Lecture 3C - Artificial Intelligence (AI) PDF

Artificial Intelligence (AI) Adapted from Udacity.com1  Outline What is AI? Background to AI Intelligent Agents Applications of AI AI Terminology AI as Uncertainty Management Examples of AI 2 ...

Artificial Intelligence (AI) Adapted from Udacity.com1  Outline What is AI? Background to AI Intelligent Agents Applications of AI AI Terminology AI as Uncertainty Management Examples of AI 2  Quiz 1. An AI program is called: a. Intelliware b. Formulaware c. Intelligent Agent d. Formula Agent 3  What is AI? An intelligence displayed by machines It is the study of intelligent agents Accomplished through machines that mimic human functions e.g. cognitive ability, learn, problem solve AI has become a critical area of the technology industry by helping to solve a number of difficult problems 4  Background to AI In earlier years, algorithms followed step by step reasoning process By the 1980’s and into the 1990’s, AI had developed ways of dealing with uncertainty or incomplete information by utilising probability and economics However, algorithms require lots of memory and computational power, thus there was a need for more efficient techniques 5  Background to AI Human beings commonly use intuitive judgments, rather than step by step approaches AI was able to model this strategy by using – intelligent agents that rely on sensory-motor skills for higher order reasoning – neural networks for simulating structures inside the brain – statistical approaches e.g. probability 6  Intelligent Agents Intelligent agents interact with an environment The agent can perceive the environment through its sensors The agent can affect its state through its actuators The perception-action cycle is the circular flow of information that takes place between the agent and its environment. Environment AGENT SENSORS Control Policy ACTUATORS 7  Quiz 2. AI has successfully been used in: a. Finance b. Robotics c. Games d. Medicine e. The Web f. None of the above 8  Applications of AI AI in Finance TRADING AGENT RATES, NEWS Predictions Stock Market Decision Bonds Controlmaking Policy Commodities TRADES (buy or sell) 9  Applications of AI AI in Robotics CAMERAS ROBOT MICROPHONES TOUCH Pick up object, OBJECTS Avoid Obstacles, OBSTACLES Follow a path MOTORS PATH coloured in RED VOICE 10  Applications of AI AI in Games GAME AGENT YOUR MOVES Decisions YOU geared Control Policy Beating you AGENT’S OWN MOVES 11  Applications of AI AI in Medicine YOU DIAGNOSTIC AGENT VITAL SIGNALS Nature of your health Control by Policy examining DIAGNOSTICS vital signs, DOCTOR symptoms 12  Applications of AI AI and the Web CRAWLER WEB PAGES WORLD DB WIDE WEB QUERY LIST OF SITES YOU 13  AI Terminology Fully versus Partially Observable Fully – what your agent can sense at any point in time is completely sufficient to make the optimal decision e.g. all cards are faced up and cards are matched Partially Observable – In this case, you need memory on the side of the agent to make the best possible decision e.g. in the game of Trump, all the cards are not open on the table and memorizing past moves can help you make a better decision 14  AI Terminology Deterministic versus Stochastic Deterministic – actions uniquely determine the outcome e.g. in the game of chess there is no randomness in moving a playing piece (rules for moving a piece are predefined) Stochastic – outcome involves some randomness e.g. rolling dice 15  AI Terminology Discrete versus Continuous Discrete – finite choices, finite number of things to sense e.g. in chess there are a finite number of board positions Continuous – space of possible actions or things to sense may be infinite e.g. with throwing darts there are an infinite distance along with a combination of angles you can throw them along 16  AI Terminology Benign versus Adversarial Benign - the environment might be random but it has no objective of its own that would contradict your own objective e.g. weather might affect the outcome of your actions but it is not intentionally targeting you Adversarial – in this environment, e.g. in chess your opponent's objective is to compete against you. In other words, your opponent observes you and tries to counteract what you are striving to achieve. 17  Quiz 3. Which applies to the game of Checkers? a. Partially observable b. Stochastic c. Continuous d. Adversarial 18  Quiz 4. Which applies to the game of Poker? a. Partially observable b. Stochastic c. Continuous d. Adversarial 19  Quiz 5. Which applies to a Robot car? a. Partially observable b. Stochastic c. Continuous d. Adversarial 20  AI as Uncertainty Management AI is the technique of uncertainty management in computer software AI is the discipline that you apply when you want to know ‘what to do when you don’t know what to do’ Reasons for uncertainty – Sensor limit – Adversaries – Stochastic environments – Laziness – Ignorance 21  Examples of AI At Google – Machine Translation system 22  Examples of AI Google’s Machine translation system is built for 50 different languages Apply machine learning techniques rather than trying to build them by hand Works by going out and collecting examples of text e.g. find newspapers that publishes in more than one language and use this as a standard to take pattern from (Translation Model) When given some text to translate, this model is used to find the most probable translation 23 Examples of AI 24 Examples of AI 25 Examples of AI 26 Examples of AI 27 Examples of AI 28 Examples of AI 29 Examples of AI 30  Summary AI has become a critical area of the technology industry by helping to solve difficult problems AI mimics human functions Intelligent agents interacts with an environment through sensors and can affect its state through actuators The perception-action cycle is the circular flow of information that takes place between the agent and its environment. 31  Summary Fully versus Partially Observable Deterministic versus Stochastic Discrete versus Continuous Benign versus Adversarial AI is the technique of uncertainty management in computer software 32