AI-01 Introduction to AI

What is AI?

• AI is like teaching a computer to think a bit like a human—it learns from examples, makes decisions, and helps solve problems
• Artificial Intelligence, or AI, is like teaching computers to think and learn like humans.
• Artificial intelligence is the study of how to make computers do things which, at moment people do better. It’s like giving computers a brain to figure things out!
• AI helps machines do things that usually need human brains, such as solving puzzles, learning new things, or understanding Human interaction
• Smart machines simulate human intelligence, as they are programmed
• AI helps computers see, learn, think, and make choices, so they can help us with tasks like talking, solving puzzles, and finding answers—just like a smart helper!
• John McCarthy, the father of AI, defined it as: "The science and engineering of making intelligent machines."

---

Evolution of AI

Year	Milestone	Explanation
1950s	Alan Turing & Turing Test	Questioned “Can machines think?” Introduced imitation game.
1956	Dartmouth Conference	Coined the term “Artificial Intelligence.”
1960-70s	Symbolic AI / Logic	Rule-based systems and logic programming (e.g., SHRDLU).
1980s	Expert Systems	E.g., MYCIN (diagnosis of blood infections). Used rules and facts.
1997	Deep Blue vs Kasparov	IBM’s Deep Blue beats world chess champion Garry Kasparov.
2010s-Present	Deep Learning Era	AI surpasses human-level performance in tasks like vision, NLP (e.g., GPT, AlphaGo).

---

Real Life Example

• Healthcare: AI reads X-rays to find diseases (like Google’s AI for eye problems in diabetes)
• Entertainment: Netflix suggests movies you’ll like using smart guessing
• Navigation: Google Maps shows traffic and best routes using past and live data
• Personal Assistants: Siri and Alexa understand voice using smart language tricks (NLP)
• Finance: AI helps catch fraud in bank transactions
• Robots: AI powers robots to clean homes, help in factories, or deliver packages
• Gaming: Games use AI to create smart enemies and exciting challenges
• Photos: Apps like Google Photos recognize faces and group your memories
• Education: AI apps give personal practice based on what students know
• Security: AI can help recognize faces or detect suspicious activity in CCTV

---

Difference between AI, ML, DL

• Artificial Intelligence:

  • Artificial Intelligence is the branch of computer science that focuses on creating machines that can think, learn, and act like humans.
  • It includes learning, reasoning, problem-solving, understanding language, and more.
  • Example: Chatbots, self-driving cars, voice assistants like Alexa.

• Machine Learning:

  • Machine Learning is a subset of AI that allows machines to learn from data and improve their performance without being explicitly programmed.
  • It finds patterns in data and makes decisions based on experience.
  • Example: Email spam detection, movie recommendations.

• Deep Learning:

  • Deep Learning is a subset of Machine Learning that uses neural networks with multiple layers (just like a human brain) to learn from large amounts of complex data like images, videos, or speech.
  • Example: Face recognition, voice assistants, self-driving car vision systems

---

Types of AI

Based on Capabilities

Narrow AI

• AI systems that are designed to perform one specific task.
• Narrow AI is everywhere today and helps us with specific tasks.
• They can’t perform tasks outside their defined function
• They simulate intelligence, but they don’t possess true understanding.
• These are AI systems designed to do one specific task really well. They can’t do things outside their programmed area.
• Siri or Alexa: They can answer questions, set alarms, or play music, but they can’t drive a car or cook food.
• Google Translate: It can translate languages but can’t write a story or solve math problems.
• ChatGPT: It can chat, write essays, or help with homework, but it can’t physically draw a picture or play sports.
• Facial Recognition: Used in phones to unlock them by recognizing your face, but it can’t recognize your voice.
• Recommendation Systems: Like Netflix suggesting movies or YouTube recommending videos based on what you’ve watched before.

General AI

• General AI is a type of AI that can think, learn, and do anything a human can do. It doesn’t exist yet, but scientists are working on it.
• A theoretical AI system that has general cognitive abilities like a human.
• Can learn, understand, and perform any intellectual task that a human can.
• An AI that can go to school, learn new subjects, and apply that knowledge in real life, just like a student.
• A machine that can paint a picture, write a poem, and solve a math problem, all while understanding emotions and feelings.
• Examples:
  • A robot that can learn to cook, clean, play games, and even have a conversation like a human.

Super AI

• A hypothetical AI that surpasses human intelligence in all fields.
• Super AI is like a super-smart robot from the future that we see in movies but don’t have in real life yet!
• This would be an AI that is smarter than humans in every way.
  • Creativity
  • Emotions
  • Problem-solving
  • Decision-making
• It’s still just an idea and doesn’t exist in real life, but it’s often shown in movies and stories.
• Examples:
  • Ultron (Avengers: Age of Ultron): A super-intelligent robot that wants to take over the world.
  • Skynet (Terminator): An AI that becomes self-aware and tries to control humanity.
  • J.A.R.V.I.S. (Iron Man): A highly advanced AI that helps Tony Stark with almost everything, but in a good way

Based on Functionality

Reactive Machines

• Oldest and most basic type of AI.
• Cannot use past experiences to make future decisions.
• They operate purely based on the present input, with no memory or ability to learn from past experiences.
• No memory, no learning, purely reactive.
• Like a calculator—it just does what you tell it, right away!
• These systems analyze the current situation and respond according to pre-programmed rules or algorithms
• Example: Picture a robot playing a game of tic-tac-toe.
• React Machines are
  • Rule-based
  • Highly predictable
  • No adaptability

Limited Memory AI

• Temporary memory, learns from past data, improves over time.
• This is like a robot with a short memory.
• It can look at what’s happening now and remember a little bit of what happened before, so it can make smarter choices. But it doesn’t keep those memories forever—just for a little while.
• Self-driving cars. They use sensors to react to the environment
• These AIs use past data temporarily to make better decisions.
• Most modern AI systems fall into this category.
• Example:
  • Imagine a robot car driving down the road.
  • It sees a kid on a bike and remembers that the bike was moving fast a second ago, so it slows down to be safe.
  • It’s like how you might remember where you hid your toys yesterday, but not last month!

Theory of Mind AI

• Emotional intelligence, understanding of intentions, social awareness.
• This robot is like a friend who can guess how you’re feeling. It doesn’t just see what you do—it tries to understand why you’re doing it, like if you’re happy, sad, or mad.
• Theory of Mind AI is a hypothetical next step where AI can understand and model the emotions, intentions, and thoughts of humans or other agents.
• Requires a deep understanding of psychology and context; still largely theoretical.

Self Aware AI

• Consciousness, self
• An understanding of its own existence, goals, and limitations.
• This is the sci-fi dream of machines that think and feel like humans.
• Understand its own state
• Think independently
• Form goals and desires

---

Turing Test

• A method to determine whether a machine can demonstrate human intelligence.

• Proposed by: Alan Turing in his 1950 paper “Computing Machinery and Intelligence”.

• Main Question: “Can machines think?”

• Imagine a game with 3 players:

  • A = A real person (human)
  • B = A computer (machine)
  • C = A judge (also a human)

• How they play:

  • The judge chats with both the person (A) and the computer (B).
  • The judge can’t see or hear them — only text messages (like using WhatsApp or chat apps).

• What’s the goal?

  • The judge has to guess: "Which one is the human, and which one is the machine?"

• How does Turing Test works?

  • The judge talks to both — a human and a machine — using text only (like chatting on a phone). No voice, no pictures — just words
  • The judge’s job is to guess who is human.
  • If the machine fools the judge and the judge can’t guess correctly more than half the time.
  • The machine passes the Turing Test — it acts smart enough to feel human!

• Purpose of the Turing Test:

  • The Turing Test was created to check if a machine can think and talk like a human
  • It tests whether a computer can answer questions like a human would.
  • If a human judge can't tell if they’re chatting with a person or a machine, the machine passes the test
  • It helps us see how smart or human-like AI has become.
  • The Turing Test is a benchmark for checking how smart an AI really is
  • It’s not about how the machine works, but what it can do — can it talk and respond like a human?

• It checks if the AI can:

  • Understand natural language (like chatting in real sentences)
  • Represent knowledge (remember facts or information)
  • Reason (think and give smart answers)
  • Learn (improve from past experiences)

• Examples:

  • Chatbots like ChatGPT, Google Bard, or Cleverbot are smart programs that can chat just like humans
  • These AIs are trained to understand and respond using natural language — the way we talk every day.
  • Many people can’t tell if they’re chatting with a real person or a machine at first
  • That’s exactly what the Turing Test is all about

• Limitation:

  • Focuses only on behavior
    • It checks what the machine says, not how it thinks or understands.
    • A machine can fool a judge with smart replies, but that doesn’t mean it truly understands!
  • Can be tricked with clever scripts
    • Some chatbots are good at pretending — using funny or vague answers to confuse the judge.
  • Short conversations
    • In short chats, it’s easier for machines to act smart. Longer talks often show the machine’s real limits.
  • Biased by the judge
    • Judges are human too! They might guess wrong, or be fooled easily
  • Doesn’t test full intelligence
    • It only checks language skills — not things like emotions, creativity, or solving real-world problems.

---

AI Challenges

• Data Bias:

  • AI systems learn from data. If the data reflects human biases, the AI will replicate or even amplify those biases.
  • Example: A hiring algorithm trained on past employee data may prefer male applicants if the historical dataset is biased.
  • Impact: Discrimination in hiring, lending, healthcare.

• Lack of Explainability

  • Modern AI models (like Deep Neural Networks) are complex black boxes.
  • It is often not clear how a model arrived at a decision.
  • Example: Why did a model reject a loan application? It’s hard to explain clearly to the user.
  • Need: Explainable AI (XAI) for trust and transparency.

• Ethical Concerns:

  • Privacy: AI surveillance systems raise concerns about individual rights.
  • Autonomous Weapons: Can AI make life-or-death decisions in war?
  • Courtroom AI: Risk of using biased data for legal decision-making.
  • Solution: Ethical frameworks and AI governance are critical.

• Large Data & Compute Requirements:

  • Deep learning models require massive labeled datasets.
  • They also need powerful GPUs or TPUs for training.
  • Challenge: Tech giants have resources; small entities struggle to keep up.

• Security Threats:

  • Deepfakes: AI-generated fake videos/images can spread misinformation.
  • Adversarial Attacks: Tiny input changes (invisible to humans) can fool AI systems.
  • Example: Altering pixels in a stop sign image can cause a self-driving car to ignore it.

---

Simple Real World AI Problems

• Spam Detection

  • ML models like Naïve Bayes or SVM are trained on labeled email data.
  • They classify emails as spam or not spam based on features like keywords, sender, time, and links.

• Product Recommendation

  • Used by platforms like Amazon, Netflix, Flipkart, YouTube.
  • Collaborative Filtering: Suggests items liked by similar users.
  • Content-Based Filtering: Suggests items similar to what the user liked earlier.

• Route Finding

  • Used in GPS systems like Google Maps.
  • Combines graph algorithms (like Dijkstra, A) with AI to find optimal routes.
  • Factors include: real-time traffic, user history, roadblocks, and construction reports.

• Game Playing AI – Tic Tac Toe

  • Uses the Minimax Algorithm to play intelligently.
  • AI explores all possible moves to maximize wins and minimize losses.
  • Also applied in games like Chess and Checkers.

---

AI Techniques

Rule-Based Systems

• A Rule-Based System (RBS) is an AI approach that applies predefined rules to process data and make decisions.

• It follows simple IF-THEN logic, where each rule connects a condition to an action or output.

• Key features:

  • Based on IF-THEN logic:
    • IF a certain condition is met, THEN a specific action is taken.
    • Example: IF temperature > 100°F THEN alert = “High Fever”
  • Easy to build and implement
    • You don’t need machine learning or data training.
    • Just define your rules and actions clearly.
  • Used in Expert Systems
    • Popular in early AI, especially for decision-making.
    • Example: MYCIN, which helped doctors with medical diagnoses
  • Mimics human experts
    • Works like how an expert would think:
    • "IF this symptom exists, THEN suggest this medicine."
  • Transparent and explainable
    • Every decision can be traced to a specific rule.
    • Makes it easy to debug, audit, or explain to others.
  • Deterministic
    • Always gives the same output for the same input.
    • No randomness or learning—it's rule-driven.

• Advantages:

  • Simple to understand and manage
  • Works well for small, structured domains
  • No need for large datasets
  • Predictable and consistent behavior

• Limitations:

  • Rigid – Can’t adapt or learn on its own
  • Not suitable for complex or dynamic environments
  • Rule explosion – Too many rules can make the system hard to manage
  • Cannot handle uncertainty or fuzzy situations

• Real-World Examples

  • MYCIN – Diagnosed bacterial infections
  • Troubleshooting systems – For printers, cars, or electronics
  • Home automation rules – IF it’s dark, THEN turn on the lights
  • Shopping alerts – IF item in cart > ₹1000 THEN apply 10% discount

Simple Decision-Making Using Conditions

• Decision-making using conditions is the foundation of programming and AI logic.

• It means making choices based on whether a condition is true or false — just like how we make decisions in real life!

• Key features:

  • Based on "if-else" or conditional logic

• Why It’s Important in AI & Coding?

  • Helps systems take actions: Based on inputs or situations, the program decides what to do next.
  • Guides the flow of a program: Determines which block of code runs and which doesn’t.
  • Mimics real-life decisions: Just like how humans think before acting

• Advantages:

  • Simple to understand
  • Easy to implement
  • Works in every programming language
  • Foundation for AI, automation & robotics

• Limitations:

  • Only handles simple situations
  • Becomes complex when too many conditions are added
  • Doesn’t learn or adapt (needs logic written by human)

• Real-Life Examples:

  • IF you finish homework → THEN you get ice cream
  • IF bus is late → THEN wait or take auto
  • IF battery < 10% → THEN stop playing game

---