Chapter 13 Artificial Intelligence Chapter Goals • Discuss types of problems that – humans do best – computers do best • Turing test 13-2 Chapter Goals • Knowledge representation and semantic networks • Search Trees • Expert systems • Biological and artificial neural networks 13-3 Chapter Goals • Natural language processing • Natural language comprehension ambiguities • Understand the main kinds of AI used in autonomous robots 13-4 What is Artificial Intelligence? 13-5 All I have to offer is the truth… • The truth is, Hollywood movies are great, but they are not reality! 13-6 AI Today • No single computer today is close to being considered “intelligent” like a human • However, computers can solve particular human-like tasks are typically considered to require intelligence, such as: – Playing chess – Diagnosing diseases – Identifying objects in images 13-7 What is Artificial Intelligence? • This is an important question… • Computer Science has a particular viewpoint • There are 2 Big Slices: – A Practical Definition (Task oriented) – A Philosophical Definition (Perception Oriented) 13-8 What is Artificial Intelligence? • Practical Definition: Making a computer do things tasks that are easy for humans to do, like: –Find kitty (aka “Machine Vision”) –Use natural human languages –Apply expert human knowledge –Play chess 13-9 What is Artificial Intelligence? • Philosophical Definition: –Making a computer fool humans into thinking it is a human (Turing test). –(I perceive that you are intelligent). 13-10 The Turing Test 13-11 The Turing Test • Alan Turing wrote a landmark paper: “Can machines think?” • How will we know when we’ve succeeded? • The Turing test used empirically determine if a computer is “intelligent” 13-12 The Turing Test Figure 13.2 In a Turing test, the interrogator must determine which respondent is the computer and which is the human 13-13 Tasks 13-14 Us vs. The Machines • • • • Some tasks are easy for computers Some tasks are hard for computers Some tasks are easy for humans Some tasks are hard for humans 13-15 Easy For a Computer… • Adding a thousand four-digit numbers • Counting the letters in a book • Searching a list of 1,000,000 numbers for duplicates • Matching finger prints 13-16 Easy for a Human… Where’s Kitty? • A computer would have difficulty pointing out the cat in this picture, which is easy for a human 13-17 Specific AI Tasks 13-18 SPECIFIC AI TECHNIQUES • !MAGIC FREE ZONE! • We will look at techniques to perform specific tasks that we consider to be “intelligent” 13-19 Knowledge Representation • Humans use knowledge for certain tasks • AI systems must have a way to represent knowledge 13-20 Semantic Networks • Semantic network A knowledge representation technique that focuses on the relationships between objects • A directed graph is used to represent a semantic network or net 13-21 Semantic Networks Figure 13.3 A semantic network 13-22 Search Trees • Search tree A structure that represents all possible moves in a game, for both you and your opponent • The paths down a search tree represent a series of decisions made by the players 13-23 Search Tree Example: Nim Figure 13.4 A search tree for a simplified version of Nim 13-24 Search Trees • Search tree also work for more complicated games such as chess • Because these trees are so large, only a fraction of the tree can be analyzed in a reasonable time limit, even with modern computing power • Now, the biggest supercomputers compete against each other 13-25 Expert Systems • Simulates a Human Expert – Car Mechanic – Medical Doctor – Gardener • Using: – A Set of Rules – The “Data” – An Inference Engine – The SW that asks questions and applies the rules 13-26 Expert Systems • Example: What type of treatment should I put on my lawn? – NONE—apply no treatment at this time – TURF—apply a turf-building treatment – WEED—apply a weed-killing treatment – BUG—apply a bug-killing treatment – FEED—apply a basic fertilizer treatment – WEEDFEED—apply a weed-killing and fertilizer combination treatment 13-27 Expert Systems • Questions: – BARE—the lawn has large, bare areas – SPARSE—the lawn is generally thin – WEEDS—the lawn contains many weeds – BUGS—the lawn shows evidence of bugs 13-28 Expert Systems • Rules: – if (BARE) then TURF – if (SPARSE and not WEEDS) then FEED – if (BUGS and not SPARSE) then BUG – if (WEEDS and not SPARSE) then WEED – if (WEEDS and SPARSE) then WEEDFEED 13-29 Expert Systems • An execution of our inference engine – – – – – – – – – System: Does the lawn have large, bare areas? User: No System: Does the lawn show evidence of bugs? User: No System: Is the lawn generally thin? User: Yes System: Does the lawn contain significant weeds? User: Yes System: You should apply a weed-killing and fertilizer combination treatment. 13-30 Artificial Neural Network • Attempts to mimic the actions of the neural networks of the human brain • Good at things like: – Will it rain? – Is that a Kitty? – Is that a male or female face? 13-31 Biological Neurons Figure 13.6 A biological neuron 13-32 Neural Networks – Each neuron has multiple input tentacles called dendrites and one primary output tentacle called an axon – A series of connected neurons forms a pathway – The gap between axons and dendrites is called a synapse – Strong connections creates a strong pathway 13-33 Biological Neural Nets • Your brain 13-34 Neural Networks • Each connection between elements has a particular strength • Particular combinations of inputs will make it through the network and produce an output 13-35 Artificial Neural Nets 13-36 Artificial Neural Networks • The process of adjusting the connection strength is called training • A neural net can be trained to produce whatever results are required 13-37 Natural Language Processing • Three separate problems – Voice synthesis • Recreating human speech • Making computers talk • Easy to do – Voice recognition • recognizing human words • Making computers listen • Harder to do – Voice comprehension • Making computers understanding • Very hard to do 13-38 Voice Recognition is HARD • The sounds that each person makes when speaking are unique – unique shape to our mouth, tongue, throat, and nasal cavities that affect the pitch and resonance of our spoken voice – mumbling, volume, regional accents, complicate the problem 13-39 Voice Recognition is HARD • Humans speak in a continuous, flowing manner – Words are strung together into sentences – Sometimes it’s difficult to distinguish between phrases like “ice cream” and “I scream” – Also, homonyms such as “I” and “eye” or “see” and “sea” • Humans can often clarify these situations by the context of the sentence, but that processing requires another level of comprehension 13-40 Voice Comprehension is HARD • Human speech is inherently ambiguous • 3 kinds of ambiguity – Lexical – Syntactic – Referential 13-41 Lexical Ambiguity – The meaning of individual words Time flies like an arrow. – What do you mean, “flies” ? – The computer gets confused. 13-42 Syntactic Ambiguity • Phrases can be put together in various ways I saw the Grand Canyon flying to New York. • What is flying, the Grand Canyon or me? • The computer gets confused. 13-43 Referential Ambiguity • When using pronouns, for example: The brick fell on the computer but it is not broken. • To what does “it” refer to? • The computer gets confused. 13-44 Robots and AI 13-45 13-46 Autonomy • Some robots use AI • Some robots do NOT use AI • The difference is AUTONOMY • Autonomy – the ability to adapt to new situations without outside help Autonomy requires AI 13-47 Robotics • Not Autonomous – Uses “Simpler” algorithm consisting of a list of steps • Autonomous – Uses more “Complex” Artificial Intelligence 13-48 2 Robotic AI Architectures • Sense Plan Act – The “old way” – A Top-Down approach • Subsumption – The “newer way” – A Bottom-Up approach – Similar to how nature works 13-49 Robotics - Sense Plan Act Architecture • In the sense-plan-act (SPA) paradigm the world is represented in a complex semantic net in which the sensors on the robot are used to capture the data to build up the net • The “old way” • Very hard to do… 13-50 Robotics - Subsumption Architecture • Rather than trying to model the entire world all the time, the robot is given a simple set of behaviors each associated with the part of the world necessary for that behavior 13-51