PRIMER ON ROBOTICS AND AL Artificial Intelligence (Al) markable field that revolves around the development of intelligent machines and computer systems. In essence, AI aims to create systems that can simulate human intelligence and perform tasks that typically necessitate human cognitive functions like leaming, reasoning, problem- solving, perception, and —_language understanding. This chapter marks the beginning of our exploration into the world of Al, delving into its definition and the significant role it plays in our lives. The chapter includes meaning and brief history of AL, Applications of Al, Ethical considerations in Al - Bias & Fairness in AI Algorithms. Figure 20: Significance of Al 1.1 Meaning and brief history of AI AL, or Artificial Intelligence, refers to the development of computer systems that can perform tasks that typically require human intelligence. These tasks include leaming, perception, reasoning, problem-solving, speech recognition, and —_language understanding. LLL. Significance of A. Al holds immense significance by: + Enhancing Efficiency: Automating tasks to improve productivity * Enabling Innovation: Facilitating the development of advanced technologies. © Improving Decision-Making: Analyzing vast datasets for informed choices. © Personalizing Experiences: Customizing interactions based on user behavior. © Addressing Complex Challenges: Tackling intricate problems with machine learning. © Transforming Industries: Revolutionizing sectors like healthcare, finance, and manufacturing. Figure 21: Human Robot Interaction. 1.1.2. Brief History and Evolution of AT The journey of Artificial Intelligence (Al) stretches back through history, shaped by visionary thinkers, scientific breakthroughs, and periods of rapid advancement followed by periods of scepticism. Table 1: History and Evolution of AI. Year Milestone 1950 The Theseus, a remote- | controlled robotic mouse built by Claude Shannon, was able to find its way out of a labyrinth and remember its course. It is an early example of Al. 1960-1990 | Steady progress. in AT capabilities, including systems for playing chess, understanding natural language, and leaming from data. 2010 | Rapid advancements in Al, | particularly in areas like image recognition, language understanding, and speech recognition, surpassing |PRIMER ON ROBOTICS AND AI human performance in many standardized tes Rapid progress in AI- generated images, from primitive pixelate faces to photo realistic images based on text prompts. 2040 | Possibility of "transformative (projected) | Al" being developed, with capabilities potentially matching the human brain, according to forecasts by Al researcher Ajeya Cotra. 2014-2023 gy _lteligent mach x 2 Figure 22; Turing Test 1.2. Applications of AT Artificial Intelligence (AI) is all around us, making our lives easier and more exciting. 1.2.1. Voice Assistants: Your Al Friends The most popular voice assistants are Siri, Google Assistant and Alexa, These are like virtual AI friends that you can talk to! These virtual AI companions understand commands and assist with the below-mentioned features/tasks. Setting Reminders: The voice assistant can be set to provide reminders to complete the task. Asking Questions: The queries can be asked to find the answer, Playing Music: The voice assistant is helpful in terms of entertainment to play music as per individual choices. Meteorological Updates: The feature can provide the latest weather forecast. 1.2.2. Recommendation Systems: AT Understands Your Preferences After viewing a video on YouTube, one might notice subsequent suggestions for similar content. This exemplifies AI in action. Recommendation systems function as personal advisors for entertainment and shopping preferences: Movie Recommendations: On _ streaming platforms like Netflix, AI suggests movies and shows based on what you have watched before. t Music Playlists: Al can generate playlists featuring songs tailored to individual music preferences. Online Shopping: Notice recommendations like "Customers who bought this item also purchased..."2 AI assists in discovering products of potential interest. 1.2.3. Gaming and AI Opponents Al serves not only as a helper but also as an opponent in video games. In some games, AI characters play against players, demonstrating their intelligence by adapting to actions Example: Chess and Strategy Games: Al opponents offer challenges in games such as chess, adjusting strategies based on player moves. Virtual Pets: Al in virtual pet games simulates real animal behavior and interacts dynamically with players. 1.2.4, Chatbots: Talking to Machines AT helps in having a conversation with a computer, Chatbots, AI programs designed for interaction, fulfill diverse roles Example: Customer Support: The chatbot assists a website in providing answers to the raised queries, Language Learning: It helps in learning a new language by chatting and practicing.PRIMER ON ROBOTICS AND AI 1.2.5, Alin Healthcare Al is transforming healthcare in amazing ways: Diagnosis Assistance: Al helps doctors analyse medical images like X-rays to identify diseases. Drug Discovery: Al can aid in predicting drugs’ bioactivity by leveraging machine learning algorithms to analyse vast amounts of data on known compounds and their biological activities. Personalized Treatment: Al __ suggests personalized treatment plans based on patients’ medical history and genetics. Figure 23: Alin Finance 1.2.6. AL in Finance sector * Algorithmic Trading: AI algorithms predict market trends and help traders make decisions. * Fraud Detection: AI identifies unusual patterns in transactions to catch fraudulent activities. © Customer Service: AI chatbots assist customers with account inquiries and financial advice, 1.3. Ethical Considerations in AI - Bias and Fairness in AI Algorithms Artificial Intelligence has the potential to do a lot of good, but it can also raise ethical concerns, One of the most significant concerns is bias and fairness in AI algorithms, 1.3.1. Bias in Al Algorithms What is Bias in AI? Bias in AI refers to the presence of unfair and prejudiced behaviour in the decisions made by Al systems. AI algorithms, like humans, can develop biases based on the data they are trained on. These biases can lead to unfair treatment of certain groups or individuals. How does bias happen in AI? Imagine you're teaching an AI to recognize whether people are doctors or not based on their photos. If you mostly show it pictures of male doctors, it might mistakenly think that doctors are mostly male. This is an example of bias. Bias can also happen when the data used for training contains unfair stereotypes or is. not representative of the real world. Consequences of Bias in Al: Unfair Treatment: Al systems can make unfair decisions, such as denying loans or job opportunities to certain groups. Reinforcing Stereotypes: Bias can perpetuate harmful stereotypes, which is not only unfair but also harmful to society. Legal and Ethical Issues: Biased AI systems can raise legal and ethical concems, potentially leading to lawsuits, regulatory scrutiny, and reputational damage for organizations deploying these systems. Inaccurate Decisions: Bias can result in inaccurate predictions or decisions, such as denying opportunities or services to deserving individuals or groups. 1.3.2, Fairness in AI Algorithms What is Fairness in AI? Faimess means that AI systems treat all individuals or groups fairly and without discrimination. Fair AI algorithms aim to provide equal opportunities and outcomes to everyone, How Can AI Be Fair? AI faimess involves making sure that the AI: -Does not discriminate against any group. -Is transparent in the decision-making process -Is testing to validate fairness in its choices.Figure 24: AI Ethics 1 Interactive Activity: AT Ethics Discussion Now, let's dive into an interactive discussion about Al ethics. Some ethical dilemmas related to Al will be presented for the group to discuss and decide on the best course of action Ethical Dilemma 1: Al Hiring Imagine a company using AI to screen job applications. However, the AI has learned from historical hiring data, which might have contained biases. As a result, it tends to reject applications from a certain gender. What should the company do to address this issue? Ethical Dilemma 2: Predictive Policing ‘A city uses AI to predict where crimes are likely to happen. However, this AI system ends up patrolling certain neighbourhood more often than others, leading to more arrests in those areas. Is this fair? How can the city ensure fairness in its policing AI?PRIMER ON ROBOTICS AND AL In this chapter, we will explore the vital role of data and information in Artificial Intelligence (AD and understand how data is crucial for Al projects. We'll also briefly touch on the evolution of computing, 2.1 Data and Information: 2.1.1. Types of Data: Data comes in various forms, each valuable for different AI tasks: Audio Data: This includes sounds, music, and spoken words, which can be used for speech recognition and audio analysis. Visual Data: Images and videos are visual data. Al can analyse them for tasks like object detection, facial recognition, and image classification. Numeric Data: Numeric data consists of numbers and measurements, often used in statistical analysis and predictive modelling. Text Data: Textual data includes written words and documents. Natural Language Processing (NLP) techniques are applied to analyse and understand text data. 2.1.2. Data to Information: Data becomes valuable when it is processed, analysed, and transformed into meaningful information. AI algorithms are designed to extract insights, patterns, and knowledge from data, converting it into actionable information. Aspect Data Information Raw, Pier nized, Natwreand || etre, Ror figures, meaningful symbols __| context Output of Input for —_| processed Purpose generating | data, ads information | decision- making ‘Numbers, Reports, Representation text, aummaties ] regis [ee | Usage and Pen ite Interpretation fasceey interpretable, » | charts, symbols | Sraphs Lacks na | Provides Contest and | ting conte Meaning eal relevance, ose processed _| PU DIKW MODEL Figure 25: DIKW model The DIKW model, which stands for Data, Information, Knowledge, and Wisdom, is a framework often used to describe the hierarchy of information processing and understanding, While it's not inherently an Al-specific model, it's relevant to Al because AI systems often deal with processing data into higher forms of understanding, Here’s a breakdown of each level: Data: Data refers to raw, unorganized facts or observations. Its the most basic level of information and lacks context or meaning on its own, For example, a list of numbers, words, or symbols without any interpretation or analysis represents data, Information: Information is data that has been processed, organized, or structured in some way to make it meaningful. It provides context or relevance to the data, This could involvePRIMER ON ROBOTICS ANE categorizing, sorting, summarizing, or filter the data to extract useful insights, For fistineS arranging a list of numbers in ascending order OF grouping similar items together tums raw data into information, Knowledge: Knowledge involves understanding and interpreting information within a specific context. It goes beyond the mere accumulation of facts and involves comprehension, analysis, and synthesis, pccenrienes is acquired through experience, learning, and reasoning. In the context of Al knowledge might involve recognizing Pattems, making inferences, or drawing conclusions from the information provided. Wisdom: Wisdom is the highest level of understanding and involves the ability to make sound judgments, decisions, and choices based on knowledge and experience. It requires a deep understanding of underlying principles, values, and long-term consequences. Wisdom often involves the application of ethical considerations and a broader perspective beyond immediate circumstances. In AI, the goal is often to progress from data to wisdom through various computational techniques such as data mining, machine learning, natural language processing, and knowledge representation. AI systems aim to not only process and analyze data but also to extract meaningful information, derive knowledge, and ultimately make intelligent decisions or recommendations akin to human wisdom. 2.1.3. Identifying Relevant Data for AI Acquiring Data: Data can be collected from various sources such as sensors, websites, databases, or user interactions. Exploring Data: Data exploration involves understanding the data’s structure, relationships, and potential insights. Techniques like data visualization and statistical analysis help in data exploration. . Evolution of Computing From Mainframes to Al: Computing has evolved from room-sized mainframes to powerful and compact devices. Al’s development was greatly facilitated by advances in computing power and data storage AI Project Cycle Precursor: Understanding the evolution of computing and data’s role is essential as a precursor to the AI project cycle. Computing advancements enable Al algorithms to process large datasets and make complex decisions. 2.2.2. Evolution of Computing: Pre AI/ML Binary Logic System: The foundation of modern computing lies in binary logic, using 0s and Is to represent data. Binary logic allows computers to process and manipulate data efficiently through electronic switches. Conditional Gates: In early computing, logical operations were performed using conditional gates such as AND, OR, and NOT. These gates allowed for decision-making and data processing based on logical conditions. > a 2 ws NOT _NAND___XOR Figure 26: Logic gates, Deterministic Computing for Deterministic Problems: Early computers were primarily designed for deterministic problems, where the outcome is entirely determined by the input and operations. These computers were used fortasks like numerical calculations, data processing, and solving equations, 2.2.3. Relevance to Al and ML Foundation for AI and ML: The evolution of computing, especially the development of binary logic and conditional gates, laid the foundation for Al and ML. Transition to Probabilistic Problems: While early computing was suited for deterministic problems, Al and ML often deal with probabilistic problems where outcomes are uncertain.AI and ML algorithms use statistical methods to make predictions and decisions based on data. 2.2.4, Understanding the Transition Computing Paradigm Shift: Understanding the transition ae deterministic computing to proba’ a AVML is crucial for appreciating the complexity of modern AI systems. Aland ML algorithms can handle uncertainties and make informed decisions in dynamic, real-world scenarios.