ilovepdf_merged.pdf

Introduction to Artificial Intelligence Artificial Intelligence Definition, History, and Applications Artificial Intelligence and Human Intelligence Artificial Intelligence Definition Artificial Intelligence Foundations Artificial Intelligence Characteristics and Advantages Artificial Intelligenc...

Introduction to Artificial Intelligence Artificial Intelligence Definition, History, and Applications Artificial Intelligence and Human Intelligence Artificial Intelligence Definition Artificial Intelligence Foundations Artificial Intelligence Characteristics and Advantages Artificial Intelligence Applications Artificial Intelligence History and The Turing Test What is Artificial Intelligence? What is the definition of artificial intelligence, and what are the most important branches? Artificial Intelligence and Human Intelligence To build systems capable of simulating human intelligence, it needs to understand how humans perform intelligent tasks. For example, to build a robot that can move intelligently as a human from one place to another, it must first study how a human can move from an intellectual, scientific, psychological, and technical perspective. Definition of Intelligence At the beginning, we will define Intelligence and Human intelligence. Intelligence: the ability to learn from experience, solve problems, think, make sense, remember inspiring information, and deal with the demands of everyday life. Human intelligence: the ability to analyze and make conclusions, the ability to categorize and visually perceive, the ability to understand and generate spoken and read languages, the ability to analyze emotions, the ability to solve mathematical problems, and the ability to diagnose and discover diseases. These skills and abilities are Applications of Artificial Intelligence in the current era. Artificial Intelligence Definition Artificial Intelligence: It’s a science that aims to build a system that simulates human behavior in its ability to learn, understand, make decisions, and solve problems. Artificial Intelligence Theories Artificial Intelligence Theories The Researchers' efforts in artificial intelligence theories were divided into four sections: 1. Thinking Rationally: Define AI as Computational studies that make it possible to perceive, think, and act. 2. Thinking Humanly: Define AI as the automation of all the activities related to human thinking. 3. Acting Rationally: Define AI as a field study the intelligent behavior in the industry. 4. Acting Humanly: Define AI as the art of creating machines that can do all the activities that need intelligence. Artificial Intelligence Foundations Artificial Intelligence Foundations AI is a modern science related to different science fields: 1. Philosophy: ▪ Contributed to the establishment of many concepts in artificial intelligence as the rationality concept. ▪ It formulated its elementary rules by Aristotle. 2. Mathematics: ▪ Many of the concepts used in artificial intelligence have been built in the sciences of mathematics, statistics, and logic. ▪ use the theory of probability and completeness. Artificial Intelligence Foundations 3. Psychology: ▪ Contributed to the studies of human vision and perception in building the models. ▪ Contributed with AI in Cognitive psychology, which looks at the brain as a device for processing information, has also contributed to the knowledge- based agent system. 4. Neuroscience: ▪ Contributed to AI in the study of the nervous system and the brain. ▪ Neuroscience research has contributed to the development of several Neural Network models used in the fields of machine learning and deep learning.  AI Branches (Strong/Weak) The researchers consider that there are two branches of artificial intelligence, and they are what is known as artificial intelligence, the strong and the weak: Weak AI Strong AI Define Artificial Narrow Intelligence (ANI): Define Artificial General Intelligence (AGI): when the machine simulates narrow and limited tasks or when a machine can simulate multiple human bits of capabilities of human tasks and their cognitive ability. intelligence without being tied to a single task. A weak or limited artificial intelligence approach is used in: General artificial intelligence has potential applications in building artificial intelligence systems at present. robotics, where machines can think, make decisions on their own, and make them more efficient and productive. Examples: personal assistant programs in smartphones Examples: Google and Deep Mind (Apple’s Siri) or autonomous vehicles. Narrow rangeability holding perspective that executes Artificial Super Intelligence (ASI): specific focused tasks, without the ability to the self- operates beyond human-level intelligence, capable of expand mechanism(functionality). outsmarting human beings in potentially every field of knowledge and activity. However, it’s currently a hypothetical concept because no system has yet achieved ASI. The characteristics of AI. Why AI is important? Automation 1. Automation: uses technology to perform tasks automatically with minimal or no human intervention. Automation helps in : 1. Reduce the cost of hiring and training many human resources. 2. It helps human resources to complete tasks continuously which increases productivity. Examples: Use machines to transport and deliver parts and packages across factory floors or programs. Reliability and Accuracy 2. Reliability and Accuracy AI increases data analysis capabilities, allowing for large amounts of information to be quickly and accurately analyzed to identify patterns, trends, and insights. This can be especially useful in fields such as science and healthcare, where the accuracy and reliability of data is critical. Examples: Systems that require high accuracy like the system for identifying cancer cells of the brain and cervix in medical imaging. Availability 3. Availability A customer service employee can provide support to only one person at a time for a specific period of no more than hours. Machines or software that simulate intelligent human behaviour can operate continuously and efficiently. An example of this: a chatbot service that can answer customer questions 24 hours a day. Efficiency 4. Efficiency Computers can provide information and respond faster than humans, especially when dealing with a huge amount of data that may be unorganized or obtained from different resources at the same time. Example: accident monitoring systems that can analyse large amounts of data from many different sources, such as obtaining information from sensors, databases, and blogs. Risk Mitigation 5. Risk Mitigation Is the ability to replace humans in dangerous and critical situations. For example, using robots to defuse bombs, as well as using robots to explore the deepest parts of the ocean or dangerous or difficult-to-reach places. AI Applications Speech Recognition 1. Speech Recognition Applications can simulate, understand, and speech recognition. Smart software hears the spoken sentences and understand them, then implements or responds with relevant information. This is converting the sound waves into texts then analysing them and generating the appropriate responses. Example: Personal assistant applications in smartphones and some electronic devices. Computer Vision 2. Computer Vision Enable a machine to simulate the human ability to understand and distinguish the content of digital images or videos. Computer vision tasks include methods for : Obtaining processing, analyzing, understanding digital images, and extracting high-dimensional data from real- world images to obtain digital or symbolic information. Example: 1. People identification applications through images. 2. Automatic examination applications in factories to detect manufacturing problems. 3. Medical image analysis applications to detect diseases. Natural Language Processing 3.NLP Is related to different corrections such as linguistics, computer science, and information engineering. Its applications aim to read, analyse, process, and generate human languages to achieve many goals. Examples : 1. Classifying documents and analysing opinions and feelings. 2. Natural language processing techniques are used in several areas, including answering questions, machine translation, automatic corrective, spam filtering, and many others. 3. Most natural language processing techniques rely on machine learning. Robotics 4. Robotic: The goal of robotics is to design intelligent machines that can help humans, facilitate their daily lives, and keep everyone safe. These systems can perform many tasks that differ according to the purpose for which they are manufactured. Example: 1. Robots can sense and see things, measure temperature, and contain processors that perform complex operations. 2. Robots are used in dangerous environments, manufacturing processes, or places where humans cannot exist. History and Evolution of the AI The start of the AI 1956 to 1943 1950 1956 1960 1970 Warren Alan Turing John McCarthy at Lutfi Zadeh McCulloch (Mathematician) Dartmouth (Scientist) John McCarthy (Philosopher and College Developed the Lisp medical scientist ) language and produced Walter Bates a paper entitled (Mathematician) Wrote a scientific Published his "Mainstream Logic article called Did a workshop research paper Programs "Computing studying machine "Fuzzy Groups", Proposed a program Proposed the first Machines and intelligence, which is called The Advisor model of artificial Intelligence” artificial neural considered one of Taker to search for neural networks. networks, and the bases of the solutions to common Each neuron is automation theories on which problems in the world. assumed to be a theory. intelligent systems machine with two are built. states that can be Frank Rosenblat represented by 0 Developed the or 1 or on/off. simplest model of artificial neuron called Albertson. The Expert System Technology 1970-1980 During 1970 to Mid- the 1970s 1980 1980s The most important development in A major decline in support for the field the 1970s was the realization that the of artificial intelligence, because the field of smart devices should be Great progress has been made in all systems at that time were unable to sufficiently specific. Trying to areas of artificial intelligence. solve large-scale problems from the find solutions to large-scale, highly real world. complex, or real-world problems was an unachievable goal. Many important results and 1. The expert system that was developments have emerged in developed in this era is the Dondrell several areas, including: program. Machine Learning The Dondrell program is an expert Intelligent Teaching systems program designed to assist organic chemists in identifying Multi-agent Planning unknown organic molecules by Uncertain Thinking, Data Mining analyzing their mass spectra using Understanding And Translating knowledge of chemistry. Natural Language Computer Vision Virtual Reality. 2. The expert system that was developed is the MYCIN system. It is an expert system that aims to diagnose infectious blood diseases and provides the doctor with a therapeutic brochure in an easy way. Machine learning and Deep learning 1982 1986 1989 1999 1. Great Mater Rummelhart, Yan Lecon development took John Hopfield Hinton, and place in the field of McClelland deep learning techniques, and this 1. Developed the acceleration in Introduced neural Major development backpropagation developments networks with occurred in the field algorithm with a occurred when data feedback or what is of artificiality when filtering neural processing became known as field they do network. faster Use of graphics networks, which are backpropagation processing units GPU. the basis of the algorithm, which is recurrent neural one of the methods 2. This approach is network. 2. Artificial neural of machine learning widely used in network models in artificial neural computer vision began to compete networks. applications. with the support vector machine model. Turing Test Alan Turing proposed a test called the "Turing imitation Test" to measure the ability of machines to imitate human behaviour. The test steps: 1. Interaction or textual dialogue takes place with a terminal that is either connected to a human or a computer. 2. If the dialogue with the terminal continues for a sufficient period being unable to determine whether the terminal is connected to a human, a computer or a smart program. Turing Test In the following are the skills that should be included in Turing's test to success: 1. Natural Language Processing: For the program to be able to speak to the resident, it needs to analyze the sentence, extract the context, and then generate the appropriate answer. 2. Knowledge representation: For the program to be able to answer the questions, a large number of information must be provided and stored before the conversation. It also requires saving the information that was discussed during the conversation. 3. Inference: It requires analyzing the information that is stored and drawing conclusions in a short time. 4. Machine Learning: It requires adapting to new conditions and instantaneous learning during the conversation. The machine needs to analyze and detect patterns so that it can conclude. Thank You Introduction to Artificial Intelligence Practice Questions Lecture 1 1. Which of the following options is the science that aims to build a system that simulates human behavior in its ability? A. Human Intelligence B. Intelligence C. Human D. Artificial Intelligence 2. Which of the following theories can define AI as a field study the intelligent behavior in the industry? A. Thinking Rationally B. Thinking Humanly C. Acting Rationally D. Acting Humanly 3. Which of the following Artificial Intelligence Foundations has contributed to the studies of human vision and perception in building AI models? A. Philosophy B. Mathematics C. Neuroscience D. Psychology 4. Which of the following AI applications hears the spoken sentences and understands them then responds with relevant information? A. Speech Recognition B. Computer Vision C. NLP Processing D. spam filtering 5. Which of the following AI characteristics expresses the ability to replace humans in dangerous and critical situations? A. Automation B. Risk Mitigation C. Availability D. Efficiency 1 Introduction to Artificial Intelligence​ Problems and methods of artificial intelligence and knowledge representation Appropriate Tasks to Artificial Intelligence.​ Artificial intelligence methods.​ Knowledge Representation:​ - Graphs.​ - Semantic Trees and Search Trees.​ - Production Systems. - Semantic Networks. Appropriate Tasks to Artificial Intelligence Many problems can be solved using traditional computer techniques that involve simple decision-making or accurate calculations ,it can produce by Artificial intelligence to get it done better. There are three common characteristics of most AI problems: 1. The problems of artificial intelligence tend to be significant. 2. AI problems are computationally complex and cannot be solved by simple algorithms. 3. AI problems and fields tend to embody a great deal of human experience. Appropriate Tasks to Artificial Intelligence To determine what problems are appropriate for AI there are some examples: 1. Medical diagnosis. 2. Barcode shopping. 3. ATM machines. 4. Strategy games such as chess. Appropriate Tasks to Artificial Intelligence First, medical diagnosis is one of the scientific fields using expert systems. Medical diagnosis is suitable to be represented through expert systems techniques because it contains specialized information in a specific field, which can be represented hierarchically. Expert systems are the most successful artificial intelligence techniques in achieving comprehensive and effective results. Example MYCIN, an expert system designed to aid in the diagnosis of bacterial infections in the blood. MYCIN presents the probabilities of the disease as well as determines the degree of certainty that the correct diagnosis. Appropriate Tasks to Artificial Intelligence Second, shopping using a barcode scanner: The shopping process using a barcode to scan products and transfer them to the cash register. The shopping process can be developed to be more effective with smart machines by the following example : The machine may suggest purchasing products based on previous purchases. “Do you need a box of detergents Laundry?" Or suggest special foods suitable for the buyer's diet. Third, automatic teller machines : (ATMs), current ATMs developed to provide you with a general financial advisor, such as tracking a person’s spending, analyzing consumer purchasing behavior, studying categories and repetition of elements, and the machine can suggest options that provide more financial security. Appropriate Tasks to Artificial Intelligence Fourth, strategic games such as chess: using AI to build an intelligent system capable of playing bilateral strategic games such as chess, specifically machine learning techniques. For the smart chess player programmed to be able to play at the professional level, the AI must be programmed on the rules of play, and it must also be able to interpret the motives and reasons for the opponent's moves. An example is the Deep Blue program, by IBM. It is the first system that can beat the world champion in the Chess game. AlphaZero developed by DeepMind Corporation, using self-learning techniques, has been able to beat the world champion in many matches. AI Methods To build an intelligent system that simulates human behavior and can work in the real world, first a specific approach must be followed to build a system. We will mention the most used approaches to building intelligent systems, which are: 1. Knowledge representation. 2. Search algorithms. 3. Rule-based reasoning. 4. Uncertainty reasoning. 5. Machine Learning. Knowledge Representation Machines or computers process and store huge amounts of information. Data There is a hierarchical relationship between data, facts, information, and Facts knowledge. From data, we can build facts, and from facts we get information and Information Knowledge can be considered as processing information to help make knowledge an intelligent decision. Knowledge Representation Representing knowledge methods differ according to the type of problem, the method of solving it, and the programming language used to build the system. For example: Data is usually represented in smart games in the form of search trees. While tables are used in machine learning to quickly and accurately retrieve data. There is a difference between data, facts, information, and knowledge. Data is usually numbers without a specific meaning or units associated with them. Facts are numbers with units. Information is the conversion of facts into meaning. Knowledge is the processing of information to help make complex decisions. Knowledge Representation For example, we are trying to determine the appropriate conditions for swimming. The data will be the numerical value, and when we add a unit to the data, we get facts, and to convert these facts into information, we get knowledge. Example Data Fact Information Knowledge If the 20-degree The temperature outside temperature above Swimming status 20 Celsius is 20 degree Celsius 20 degree Celsius, I can swim Knowledge Representation Knowledge representation systems consist of two parts: 1. Data structures that contain trees, lists, and stacks. 2. Inferential procedures for the use of knowledge, such as searching, sorting, and collecting. The elements that knowledge bases include: 1. Objects: Physical objects and concepts. 2. Events: The element of time, cause-and-effect relationships. 3. Performance: Information about how something is, the intended steps, and the logic or algorithm that governs performance. 4. Meta-Knowledge: About knowledge, reliability, and relative importance of facts. Knowledge Representation Knowledge representation is considered one of the most important basics for building computer systems. There is a limited number of data structures used, such as tables, arrays, stacks, and lists which a problem can be represented and solved. In the field of artificial intelligence, there are many ways to represent complex problems and solutions to them, and the traditional types of representation used in computer science and artificial intelligence science, including: 1. Graphs. 2. Search trees. 3. Semantic networks. 4. Production systems. Graphs The graph contains a set of vertical vertices or nodes and a group of edges or links. The nodes represent the objects, while the links represent the relationships between the objects. Each edge consists of a pair of nodes. If the edge "e", that is consists of the vertical vertices value {u, v}, then we often write: 1. e=(u, v). 2. u is also linked to v. 3. v is linked to u. 4. u and v are contiguous. Graphs can also be: 1. Directed: The links are arrows, which 2. Undirected: without arrows. means they have their direction. u e v u e v Directed graph Undirected graph Graphs For example, in the graph here, Ahmed teaches Omar, not the other way around. Teaches Ahmed Omar Directed Labeled Graph The links can contain : 1. Labels - The links are named to indicate the type of relationship 2. Weights- The weights to indicate the strength of the link between objects or nodes. 25 Home Shop Directed Weight Graph Graphs The graphs are a tool in representing knowledge because they represent states, alternatives, and paths used for measurement in problem search. The correct choice of representation of a problem Omar has a significant impact on how the problem is solved. Ahmed Chain of nodes: it’s one link from the first node to the second and then from the second to the third, and so on. The knowledge method will be ineffective because it means visiting each node Khalid until the required node is found. One of the disadvantages of the graphs is unable to represent negation, for example, We cannot represent Ahmed does not teach Khalid in the graph here. Semantic Networks Semantic networks are a type of graph and one of the methods for representing knowledge =multipurpose relation. It aims to represent knowledge in a way like the functioning of human associative memory. Semantic networks help provide a general definition of an object, concept, event, or situation. In semantic networks the objects of knowledge are represented in the form of nodes and the relationship between nodes is represented in the form of arrows. The relation (IS-A) is often used in semantic networks, although it does not always correctly represent real-world facts. Sometimes it may represent a specific membership, and other times it may mean equality. Example: the penguin is (IS- A) a bird, and birds can fly, but a penguin cannot. IS-A can PENGUIN BIRD FLY Semantic Trees A semantic tree: is a type of semantic network. Semantic tree has the following characteristics : 1. Each node has exactly one predecessor (parent) and one or more successors (children). Root node 2. In the figure, node A is the predecessor of node B: node A connects by one edge to node B and comes before it in the tree. The successors of node C, nodes E and F, connect A directly (by one edge each) to node C and come after it in the tree. We can write these relationships as: succ (B) = D and pred (B) = A. B C 3. The nonsymmetric nature of this relationship means that a semantic tree is a directed graph. 4. One node has no predecessors. This node is called the root node. D E F The root node: the starting point when searching in a semantic tree. Leaf nodes: Do not have children. Leaf nodes Goal node: One or more from the leaf nodes. Goal node The disadvantage of this approach in semantic networks is that it does not lead to any quick solution. Semantic Trees The path is taken to cross the tree, A which may consist of ( only one node = length 0) or more. The path of (path consists of a node B C and a branch = length 1 ) that leads to the next node. D E F Complete path: The path that leads from the root nodes to the goal node. Partial path: the path that starts from Semantic Tree the root node leads to the leaf node (that is not the goal node ) A C Branch: The connecting link that connects two nodes. The difference between semantic B networks and semantic trees is : E That semantic network can contain F cycles that visit the same node more than once. semantic network Search Trees A The searching for the goal node in C semantic networks requires going across the network and examining the nodes. B E Therefore, we can represent the possible F paths in a semantic network in the form of a search tree. semantic network Search tree: is a type of semantic tree. A The search tree: represents the possible B C paths through the semantic network. B E Each node in the tree represents a path, F C F where the successive layers in the tree F A represent longer paths and include C F E periodic paths, which means that some F A branches in The search tree end with leaf A nodes that are not the goal nodes. A A C Search Tree A Decision Tree Decision Tree is a type of search tree system that can be used to find a solution to a problem by choosing from alternatives starting from the root nodes. The decision tree divides the problem Yes space logically into separate paths that can No Has be followed independently in searching for family a solution or obtaining a result. An example of this is an attempt to limit the number of employees who have No Has medical Yes families and do not have medical insurance insurance. The area of ​all employees is divided and those who have a family are sorted, and then this area is divided into those who have medical insurance. +0 +0 +1 Production System Production system: For systems or people to make intelligent decisions, we need a way to evaluate the data or the situation. One of the Cognitive representations using production systems, which consists of a set of rules that take the form: IF [condition] THEN [action] Production systems: are used with control systems to translate rules for tracking inference and saving data. The database acts as a temporary store, which allows recording the conditions under the rules. For example: IF [You drive a car and you see a school bus] THEN [Stop the car until the bus passes] Thank You Introduction to Artificial Intelligence Practice Questions Lecture 2 1. Which of the following options represents the data in the knowledge representation? A. 1000 B. 1000 $ C. The cost of the phone is 1000$ D. If I work in the summer, I can buy a phone for 1000$. 2. Which of the following options represents the diagram below? A. Directed label graph. B. Undirected label graph. C. Directed Weighted graph. Sara Trains Lama D. Undirected Weighted graph. 3. Which of the following options is defined as a branch in the semantic tree below? A. Short Trip/Long Trip B. SUV/RV C. Island/Airport2 D. Airport 1/Airport2 1 Introduction to Artificial Intelligence Practice Questions 4. What does the picture below represent? A. Decision Tree B. Semantic Tree C. Search Tree D. Semantic network. 5. Which of the following options is considered a root node? A. Train B. SUV C. Airport2 D. Airport 1 2 Introduction to Artificial Intelligence Inference based on expert rules and systems Rule-based Reasoning. Knowledge Representation in Rule-based Systems. Rules in Rule-based Systems. Components of Rule-based Systems. Examples of Appropriate Implementations of Rule-based Systems. Characteristics of Rule-based Systems. Forward Chaining or Backward Chaining. Rule-based Systems Knowledge is a theoretical and practical understanding of a specific topic or field. And the sum of what is currently known in one of the specific fields. Those who possess knowledge are called experts. An expert: a person who has a deep knowledge of the facts and rules and strong practical experience in a specific field and is a skilled person who can do things that others cannot do, such as a doctor who specializes in a specific field such as blood diseases. The expert express their knowledge in AI in the form of rules, to build expert systems that simulate the knowledge of human cognitive ability. The most common and easy approach to use: is rule-based systems, which use symbolic rules to represent data and make inferences. Rules A rule in artificial intelligence can be defined in the form IF – THEN that information or facts given in the condition (IF) part are related to action in the result (THEN). The rule is presented after some description of how to solve a problem that is relatively easy to create and understand. IF THEN The rule consists of two parts: 1. The part IF called the hypothesis or condition. 2. The part THEN called the result, where everything represents the basis of the conditions of the rule in which the action represents the result or conclusion. A rule can contain multiple conditions associated with AND or OR keywords or a combination of both. Rules The conditional part (IF) consists of two parts object and value. And they are linked to each other by the operator. IF the object here is “traffic light” its value is “red”. The operator identifies the object and assigns its value. The factor can be (is, are, is not, are not) like: IF THEN Expert systems can use arithmetic operators to define an object as numeric and assign its numerical value. for example: IF