Aristotle, Greek Philosopher (384–322 BCE) Syllogisms (Greek: syllogismós, “conclusion”, “inference”) use deductive reasoning to derive workable conclusions from two or more given propositions. Leonardo da Vinci, Italian Polymath (1452–1519) Leonardo da Vinci designed a hypothetical computing machine on paper even though it was never put into practice. The machine had 13 registers, René Descartes, French Philosopher (1596–1650) René Descartes believed that thinking and decision-making (rationality) can be explained using math and mechanics, and this idea is important in AI today because AI also uses math to make smart decisions. Thomas Hobbes, British Philosopher (1588–1679) in rational decision-making, humans employ operations similar to calcu lus, such that they can be formalized in a way that is analogous to mathematics (Flasiลski, 2016). David Hume studied how people learn by seeing things happen again and again (this is called repeated exposure). He asked questions like: “How do we know one thing causes another?” From this, he helped build the idea of the learning curve — the more we practice, the better (and faster) we get. The recent history of Artificial Intelligence (AI) began in 1956 at the Dartmouth conference, where the term "artificial intelligence" was first used. Since then, key scientists like Alan Turing, John McCarthy, Marvin Minsky, and Noam Chomsky helped shape AI by developing important ideas, tests, programming languages, and research labs. These contributions laid the foundation for AI as a scientific field that mimics human abilities like thinking, learning, and understanding language. ๐ง Timeline of Key AI Pioneers Name Contribution Alan Turing John McCa rthy Invented the Turing Test to check if a machine can "think" like a human Marvi n Minsk y Noam Chom sky Co-founded the MIT AI Lab, combined AI with cognitive science Coined the term "Artificial Intelligence", created LISP language, helped organize the Dartmouth Conference, and founded Stanford AI Lab Developed the Chomsky Hierarchy, helped AI understand language (NLP) Year (s) 1950 1956 – 1960 s 1959 1950 s– pres ent ๐ Key Event • Dartmouth Conference (1956): The official beginning of AI as a scientific field; term "Artificial Intelligence" was first used here. ๐ซ Key Institutions • • • • Dartmouth College: Where AI officially began in 1956. It still hosts important conferences. MIT (Massachusetts Institute of Technology): Many great AI scientists worked and taught here. IBM & Intel: Big tech companies that helped by building tools and supporting research. DARPA (a U.S. government research agency): Gave money and support to help AI grow, especially in military and science areas. Key Disciplines That Helped AI Develop 1. Decision Theory a. Helps AI make smart choices, especially when it’s not sure what will happen (using math and economics). 2. Game Theory a. Teaches AI how to win or cooperate, like in games or strategy situations. b. Based on the work of smart scientists like John von Neumann and Oskar Morgenstern. 3. Neuroscience a. Studies the brain and helps AI learn how to think like a human. b. Inspired the idea of artificial neural networks, which are like digital brains for computers! ๐ฃ๏ธ๐ง Natural Language Processing (NLP) • • NLP helps computers understand human language — both written (like this text) and spoken (like your voice). It combines linguistics (the study of language) and computer science. ๐ป Programming Languages in AI • • High-level programming languages are easier for humans to use than machine code. Some important languages used in AI are: Language Lisp Prolog Python What it's known for Created by John McCarthy, good for handling text and lists, used since the 1960s and still useful. Good at solving logic problems and proving theorems. Most popular today, easy to use, with many libraries that make AI projects faster and more powerful. ๐ Why AI Has Grown So Fast Recently Three big reasons: 1. Big Data: We now have tons of data to train AI (like photos, videos, emails, etc.). 2. Better Computers: Computers are now faster and stronger at handling all that data. 3. Smarter Ideas: New knowledge from math, philosophy, brain science, and machine learning helps build smarter AI systems. โ Main Idea AI has made huge progress because of: • • • Better tools to understand language (NLP), Powerful programming languages like Python, And 3 key growth boosters: lots of data, powerful computers, and new scientific ideas. AI WINTER ๐จ๏ธ What is an "AI Winter"? • • An AI Winter is a time when people lost interest in AI, and money and research stopped. The term sounds dramatic and comes from “nuclear winter,” where nothing can grow after a disaster. Similarly, in AI winter, progress stopped. โ๏ธ There were 2 main AI Winters: 1st AI Winter (1974–1980) • Goal: Translate languages automatically (like English to Russian). • Problem: The translations were terrible! Example: “Out of sight, out of mind” → “Invisible idiot” in Russian • Also, the AI models at that time (called perceptrons) couldn’t handle even basic logic tasks. • Result: The US government decided the research wasn’t worth it and stopped the funding. 2nd AI Winter (1987–1993) • By this time, people had hope in Lisp machines and expert systems. • Lisp machines = special computers built for AI. • Expert systems = programs that tried to act like human experts (like doctors). • Problems: o These machines failed in real-world use. o The systems couldn’t handle too much information. o They gave wrong answers when given new or unknown inputs. • Result: AI got a bad reputation again and funding stopped. ๐ง What Causes an AI Winter? AI needs 3 things to grow and succeed: 1. Good algorithms 2. Powerful computers 3. Lots of data In the past: • • We had algorithms, but not enough data or computing power. So AI couldn’t deliver what was promised, and support faded. ๐ฎ Will There Be Another AI Winter? • Today we have: o Big data o Supercomputers o Better algorithms • • So, AI is growing fast! BUT… if a popular AI idea gets too much hype and fails to work, it could cause another winter . โ Main Idea (In 1 Line): AI winters are times when people stopped believing in AI because it couldn’t meet big promises — but with today’s tech, AI has a stronger chance to succeed. โ Covered Information from Your Text: ๐น Definition of AI Winter: • • • Term coined in the 1980s. Drop in interest, funding, and research. Compared to "nuclear winter" – nothing grows. ๐น First AI Winter (1974–1980): • • • Triggered by failure in machine translation (e.g., “invisible idiot” mistake). Funded by the US during the Cold War. Perceptrons (early AI models) couldn't handle basic logic (like XOR). • Result: Projects failed → funding stopped. ๐น Second AI Winter (1987–1993): • • • • • Started when Lisp machines and expert systems failed. Expert systems couldn’t scale — too much data, bad with unknown inputs. Industry pessimism grew. Some claim AI winters are myths (Kurzweil's argument). AI still used in areas like credit card processing. ๐น Causes of AI Winters: • • • Three missing ingredients: o Algorithms o Computing power o Training data In the 1970s: not enough data or power. In the 1980s: more computing, but still not enough training data. ๐น Now and the Future: • • • All three conditions are now met (data + power + algorithms). Still possible to have future AI winters if hyped projects fail. BUT AI is now embedded in many fields, so total collapse is unlikely. โ What is an Expert System? An expert system is a type of AI program that tries to act like a human expert in a specific field (e.g., medicine, mechanics). It helps people solve problems by using expert knowledge and logical reasoning. Instead of memorizing solutions, it uses smart methods to apply past knowledge to new problems. ๐งฉ Main Parts of an Expert System: 1. Knowledge Base Stores expert knowledge (facts, rules, and information about a specific domain). 2. Inference Engine The “thinking brain” – it applies rules to the knowledge to make decisions or solve problems. 3. User Interface Allows regular users (non-experts) to interact with the system. ๐ Types of Expert Systems (based on how knowledge is stored): 1. Case-based Systems a. Store real-life examples and their solutions. b. Solve new problems by finding similar past cases. c. Needs a way to measure how similar two cases are. 2. Rule-based Systems a. Use IF–THEN rules (e.g., If patient has a fever, then consider infection). 3. Decision Tree Systems a. Represent knowledge as yes/no questions that branch out like a tree. b. Help in step-by-step decisions. ๐ง How Did Expert Systems Start? • • 1950s: Researchers like Herbert Simon and Allen Newell tried to build a general problem solver. It didn’t fully work but inspired later systems. Edward Feigenbaum at Stanford created the first real expert system and introduced the term. ๐งช Early Applications: • • DENDRAL: Identified chemical molecules. MYCIN: Diagnosed diseases using 600+ rules. These proved expert systems could work well. ๐ Why Were Expert Systems So Important? 1. They separated expert knowledge from the computer code. This means: a. Non-programmers could understand or update the knowledge. b. Anyone could rapidly create new systems by just changing the rules. 2. Helped pioneer the idea of "explainable AI" — systems that can show how they made a decision. โ ๏ธ Why Did Expert Systems Lose Popularity? 1. Too much information = too slow a. As knowledge bases grew bigger, the system took too long to find answers. 2. Hard to manage big knowledge bases a. It became difficult to check if all the rules made sense or didn’t contradict each other. 3. They couldn’t learn a. Once rules were set, the system couldn’t update itself or learn from mistakes. b. Only a human expert could change the rules. Expert Systems Summary Chart Aspect Definition Goal Details AI systems that simulate expert human decision-making in specific domains. Help non-experts solve problems using expert knowledge. ๐งฉ Main Components Component Knowledge Base Inference Engine User Interface Function Stores domain-specific facts and rules. Applies logic to draw conclusions from knowledge base. Allows user to input queries and receive solutions. ๐ง Types of Expert Systems Type Case-based Rule-based Decision Tree Explanation Uses past cases to solve new problems. Uses “IF-THEN” rules to infer solutions. Uses structured decision trees to make choices based on examples. ๐ Development History Key Milestones 1950s–60s 1970s 1980s Details Idea born from General Problem Solver research (Simon & Newell). First systems (e.g., DENDRAL, MYCIN). Peak of popularity; used commercially. โ๏ธ Advantages • • • Explicit and modular knowledge representation Accessible to non-programmers Supports rapid prototyping โ ๏ธ Limitations Problem Scalability Consistency Checking No Learning Capability Description Slow response as rules/facts increase. Hard to ensure no rule contradictions. Can’t learn or update rules without human input. ๐ Notable Advances in AI This section talks about how AI has progressed over time, from early ideas to modern applications. ๐ง 1. Nascent AI (1956–1974) – The Beginning • • Early AI focused on symbolic logic – using rules to imitate thinking. Programs solved problems step by step, using search strategies. • • First attempts at language processing, robot control, and computer vision were made, but only in very simple settings (called "microworlds"). Scientists also started building models of how neurons might work using computers. ๐ 2. Knowledge Representation (1980–1987) – Teaching AI Facts • • • • Scientists realized logic alone isn’t enough; AI needs real-world knowledge to make smart decisions. They built expert systems – software filled with expert knowledge. Countries invested in AI research (e.g., UK’s Alvey project, Japan’s Fifth Generation Project). Neural networks improved using a method called backpropagation, making learning possible. 3. Learning from Data (Since 1993) – Real Progress Begins • AI beat a human chess champion in 1997 (Deep Blue vs. Kasparov). • • AI began being used in real-world apps like search engines, spam filters, etc. Concepts like intelligent agents became popular – programs that make decisions on their own. Thanks to the Internet, we now have a massive amount of data. In 2012, deep learning (AI learning from huge data sets using layered networks) caused a big breakthrough. AI became better at speech recognition, image classification, and more. • • • ๐ง๐ฌ Related Fields that Support AI 1. Linguistics • • Studying how people use language helps AI understand and use it too. Language is deeply connected to thought, creativity, and intelligence. 2. Cognition • AI models how humans think, learn, and decide. • Computers use inputs (stimuli), form internal ideas, then act. • A key goal today: make AI explain its decisions clearly (called explainability), especially in deep learning, which is often a "black box." 3. Games • • Games like chess or Go are ideal for AI testing, because they involve logic, decisionmaking, and strategy. AlphaGo (2015) and AlphaZero (2017) beat the world’s best Go players by teaching themselves how to play using reinforcement learning (learning by trial and error). 4. Internet of Things (IoT) • • Many devices (phones, smartwatches, etc.) are now connected and constantly generate data. AI can use this data to make these devices smarter, but this also raises privacy and ethics issues. โ๏ธ Quantum Computing & AI • • • Quantum computing is based on quantum mechanics, where particles can be in multiple states at once. This allows new types of super-fast computing, which could speed up AI. But quantum computers are still in the early stages and need more research. ๐ฎ The Future of AI & the Gartner Hype Cycle It’s hard to predict the future of AI, but we can look at trends. One tool used is the Gartner Hype Cycle, which shows how new tech evolves over time: 1. 2. 3. 4. 5. Discovery: A breakthrough excites everyone. Peak of Expectations: Everyone expects it to change everything. Disillusionment: It doesn’t meet all the hype, people lose interest. Enlightenment: People learn what it can really do (and what it can’t). Productivity: The tech becomes useful and is widely used – either by experts (niche) or by everyone (mass market). ๐ What Is the Hype Cycle? Think of the hype cycle as the emotional rollercoaster of a new technology. Every new tech (like AI) goes through phases—like a new product people get super excited about, then disappointed, and finally, they understand how to really use it. Imagine someone invents a magic smart robot that everyone hears about. ๐ The 5 Phases of the Hype Cycle (Gartner Model) Phase 1. Innovation Trigger 2. Peak of Inflated Expectations 3. Trough of Disillusionm ent 4. Slope of Enlightenme nt What Happens A new technology is born! People start talking. People expect too much. Excitement is everywhere. Reality hits. Tech doesn't work as well as people thought. Developers improve it. People learn what it's really good at. Story/Example Scientists build a new smart robot. It can talk, move, and learn. Everyone is curious! Headlines say, “This robot will replace teachers, doctors, and chefs!” People think it's perfect. The robot breaks a dish, mishears commands, and can’t cook pasta. People get disappointed. Now, it helps in hospitals or helps kids with learning disabilities. It’s not perfect, but helpful. 5. Plateau of Productivity The tech is useful and accepted. It becomes normal. The robot is now used in schools and homes. Everyone knows how to use it the right way. Note: Not every technology makes it to the last phase. ๐ค What Happened in 2021 with AI? Gartner analyzed different AI technologies and showed where they are on this rollercoaster: ๐ก Phase 1 – Innovation Trigger (new tech, rising interest) • • • • Composite AI – combining different AI methods for better results. General AI – a machine that can think and learn like a human. Human-Centered AI – AI that considers human values, feelings, fairness. Responsible AI – focused on making AI ethical and explainable. โถ These are new ideas. People are excited but real results are not there yet. ๐บ Phase 2 – Peak of Inflated Expectations (overhyped) • • • Deep Neural Networks – powerful AI models used in voice assistants, self-driving, etc. Knowledge Graphs – tech that helps AI understand relationships between information (used by Google). Smart Robots – advanced robots expected to behave like humans. โถ People expect too much from them, but they’re not perfect yet. ๐ป Phase 3 – Disillusionment (reality check) • Autonomous Vehicles – self-driving cars faced problems, crashes, and lost funding. โถ Many believed cars would drive themselves by now… but it’s harder than expected. โฐ๏ธ Plateau of Productivity? • None yet: No AI tech had fully matured and become part of everyday life in 2021—but many are getting close. ๐ง A Simple Trick to Remember: Just remember the journey of a new superhero AI: 1. 2. 3. 4. 5. Birth (Innovation Trigger): People say, “Wow!” Fame (Peak): Everyone cheers, but hopes are too high. Criticism (Disillusionment): “This isn’t that great...” Comeback (Enlightenment): “Hey, it’s useful after all.” Success (Plateau): It becomes part of life.
0
You can add this document to your study collection(s)
Sign in Available only to authorized usersYou can add this document to your saved list
Sign in Available only to authorized users(For complaints, use another form )