root@vm-learning
~
$
open
ch-b1
PART B ▪ UNIT 1
06
Introduction: Artificial Intelligence for Everyone
Definition · Evolution · Types · Domains · Terminologies · Benefits & Limitations
Artificial Intelligence (AI) refers to the ability of a machine to learn patterns and make predictions. In its simplest form, AI combines computer science and robust datasets to enable problem-solving. AI does not replace human judgement — it adds value to it. Think of AI as a smart helper that can understand, learn from examples, and do tasks on its own without being told exactly what to do each time. According to Statista, the global AI market was worth £113.60 billion in 2023 and is on a continuous growth trajectory fuelled by massive investment.
Learning Outcome: Communicate AI concepts and applications, describe its history, differentiate types & domains, know ML/DL terminology, and evaluate AI's benefits & limitations
1.1 What is Artificial Intelligence?
At the most basic level, AI lets a computer do tasks that normally require human intelligence. Three everyday examples:
🗣️
Understand Language
Understand and respond to what you say — like Siri, Alexa, Google Assistant.🖼️
Recognise Images
Look at pictures and identify objects — e.g., identifying animals in photos.🔮
Make Predictions
Analyse data to predict outcomes — weather forecast, what to watch next on Netflix.AI Machines vs Non-AI Machines
| Non-AI Machine | AI Machine |
|---|---|
| Fixed-function hardware — microwave oven follows fixed rules. | Learns from data and makes its own decisions. |
| Non-interactive systems — basic fan doesn't change with new input. | Adapts to new information — smart washing machine adjusts settings based on clothes. |
| Basic sensors — collect data but don't analyse it. | Processes data, understands context and acts. |
Key difference: non-AI machines follow rules; AI machines learn, adapt and decide based on data.
1.2 Evolution of AI — Key Milestones
Although philosophers imagined artificial beings for thousands of years, modern AI began in the mid-20th century. The following timeline shows the major events.
| Period | Key Event / Development |
|---|---|
| 1950 | Alan Turing publishes "Computing Machinery and Intelligence". He proposes the Imitation Game — later called the Turing Test — to decide whether a machine can "think". |
| 1956 | Dartmouth Conference organised by John McCarthy — the birthplace of AI as a field. McCarthy coins the term "Artificial Intelligence" and is often called the "Father of AI". |
| 1980 – 1990 | Mixed optimism and scepticism. Breakthroughs in machine learning and neural networks, but also the so-called "AI Winter" when progress and funding slowed down. |
| 21st Century | Major resurgence. Advances in computing power, data availability, and algorithmic innovation — plus breakthroughs in machine learning, deep learning and reinforcement learning. Transformative applications in healthcare, finance, transportation, entertainment. |
"Data is the new oil" — 90 % of the world's data has been created in just the last two years (compared to the previous six million years of human existence).
1.3 Types of AI (by capability)
Computer scientists classify AI into three levels based on its ability to analyse data and make predictions.
🎯
1. Narrow AI
Focused on a single specific task. Very common in consumer apps — voice shopping, Siri, purchase prediction, scheduling. Effective but has no broader understanding.🔀
2. Broad AI
A midpoint between Narrow and General. Handles a wider range of related tasks. Used in businesses to integrate AI into specific processes with domain-specific data.🌐
3. General AI
Can do any intellectual task a human can. Requires abstract thinking, strategy, creativity. Does not yet exist. Beyond it lies Artificial Super-Intelligence (ASI).1.4 Types of Data
Data can be facts, statistics, opinions, voices, photos, names or even dance moves — anything recorded in some format. It is the fuel that AI runs on. Data falls into three categories:
📊
Structured Data
Neatly organised in rows and columns (like a table). Easy to analyse.Examples: names, dates, addresses, stock prices.
🎨
Unstructured Data
No specific organisation — harder to analyse. Needs specialised tools.Examples: images, text documents, customer comments, song lyrics.
🧩
Semi-structured Data
Between the other two — some organisation, not fully tabular. Easier than unstructured.Examples: JSON, XML, emails with headers.
1.5 Domains of AI
AI is applied in many areas, but the three core domains (recognized by CBSE curriculum and industry) map directly to the three data types.
📊
A. Statistical Data
Works on numerical / categorical / alphanumeric inputs. Uses statistical methods, ML algorithms, and data-visualisation techniques to extract insights.🗣️
B. Natural Language Processing (NLP)
Works on text and speech — enables machines to understand, interpret and generate human language.👁️
C. Computer Vision (CV)
Works on images and videos — lets machines see and understand the visual world.A. Statistical Data — Examples
- Search recommendations & Google Maps history — based on your previous interactions and location data.
- Amazon's personalised recommendations — driven by your shopping habits and browsing history.
- Social media activity, cloud storage, digital textbooks — all generate vast amounts of data analysable for insights.
B. Natural Language Processing (NLP)
NLP enables computers to understand, interpret and generate human language — covering tasks such as language translation, sentiment analysis, text summarisation, and speech recognition. Goal: a model that understands not just the words but also slang, sarcasm, inner meaning and contextual nuance.
NLP vs NLU vs NLG
| Abbreviation | Full form | What it does |
|---|---|---|
| NLP | Natural Language Processing | The umbrella — machines working with human language (text + speech). |
| NLU | Natural Language Understanding | Sub-field focused on reading and comprehending language — like reading a library of books. |
| NLG | Natural Language Generation | Sub-field focused on producing language — takes structured data as input, writes coherent text / speech. |
C. Computer Vision (CV)
CV gives computers the ability to see and understand the world through digital images and videos — the way humans use their eyes. Tasks include object detection, image classification, facial recognition, and scene understanding. Applications: autonomous vehicles, medical imaging, augmented reality.
Pixels and Resolution
A digital image is a grid of tiny coloured dots called pixels. Each pixel stores information about its colour and intensity. Resolution is the number of pixels along width × height — e.g., 1920 × 1080 means 1920 across × 1080 down. Higher resolution = sharper image.
🧭 Example — Categorise Applications into the 3 Domains
For practice, place each of the following into Statistical Data / NLP / Computer Vision.
| Statistical Data | Natural Language Processing | Computer Vision |
|---|---|---|
| Fraud detection in financial transactions · Sports analytics for performance · Customer segmentation for targeted marketing · Stock prediction · Recommendation systems for e-commerce | Text summarisation for news articles · Automated subtitles for videos · Google Translate · Grammarly · Chatbots like ChatGPT | Augmented-reality filters (Snapchat) · Object detection in autonomous vehicles · Medical image diagnosis · Facial recognition at airports · Google Lens |
1.6 AI Terminologies
Machine Learning (ML)
Machine Learning is a subset of AI focused on developing algorithms and models that let computers learn from data and make predictions or decisions without being explicitly programmed. The more data, the better the model.
Deep Learning (DL)
Deep Learning imitates how the human brain processes data and creates patterns for decision-making. It uses Artificial Neural Networks (ANNs) with multiple layers — if the number of layers (input + hidden + output) is more than three, it is called a Deep Neural Network.
🧠 Artificial Neural Networks (ANNs)
- A subset of Machine Learning and the core heart of modern ML/DL.
- Organised as node layers — an input layer, one or more hidden layers, and an output layer.
- If a node's output is above a set threshold, the node is activated and sends data to the next layer. Otherwise no data is passed along.
- More than 3 layers → Deep Neural Network.
📊 Machine Learning vs Deep Learning — Comparison
| Machine Learning | Deep Learning |
|---|---|
| Works on small datasets with good accuracy. | Works on large datasets. |
| Runs on low-end machines. | Heavily dependent on high-end machines (GPUs). |
| Divides task into sub-tasks, solves each, combines results. | Solves the problem end-to-end. |
| Takes less time to train. | Takes longer to train. |
| Testing time may be higher. | Less time to test the data. |
| Requires manual feature extraction. | Automatic feature extraction through layers. |
Sorting at an electronics store. Imagine you have dozens of products and very little time to sort them.
ML approach: label hundreds of products with size/shape/colour features and train a model — it matches new items to the right category. More data → better model.
DL approach: feed raw product photos into a deep neural network — layers automatically learn each product's visual features; no human feature-engineering needed.
1.7 Types of Machine Learning
📘 Supervised Learning
Learns from labelled data. Input is paired with correct output during training.Goal: learn a mapping from input → output so it can predict on new, unseen data.
Algorithms: linear regression, logistic regression, decision trees, SVM, neural networks.
🔍 Unsupervised Learning
Learns from unlabelled data — no correct output given.Goal: discover hidden structure — clusters, associations, anomalies.
Algorithms: k-means clustering, hierarchical clustering, PCA, autoencoders.
🎮 Reinforcement Learning
An agent learns by interacting with an environment — receives rewards or penalties.Goal: learn a policy that maximises cumulative reward over time.
Algorithms: Q-learning, Deep Q-Networks (DQN), policy gradients, actor-critic.
Common Uses of Each Type
| Type | Typical scenarios |
|---|---|
| Supervised | Spam filtering, credit scoring, price prediction, medical diagnosis. |
| Unsupervised | Customer segmentation, market basket analysis, anomaly detection. |
| Reinforcement | Playing games (AlphaGo, chess), robot control, portfolio management, self-driving cars. |
1.8 Benefits of AI
⚡ Efficiency & ProductivityAutomates tasks, analyses data faster, optimises processes across sectors.
🧠 Better Decision-MakingAnalyses vast data, spots patterns humans miss — leads to data-driven decisions.
💡 Innovation & CreativityGenerates new ideas, automates repetitive work, frees humans for creative pursuits.
🩺 Science & HealthcareAids drug discovery, medical diagnosis, personalised medicine — powers research breakthroughs.
1.9 Limitations of AI
🏭 Job DisplacementAutomation raises concerns about lost jobs and the need for re-training / up-skilling the workforce.
⚖️ Ethical ConcernsBias in algorithms · misuse for surveillance or manipulation · need for ethical guidelines and regulation.
❓ Lack of ExplainabilityComplex models are opaque "black boxes" — it's hard to understand how they reach a decision.
🔒 Data Privacy & SecurityLarge-scale data collection raises concerns about privacy leaks and security vulnerabilities.
1.10 Practical & Extension Activities (from Handbook)
Practical (from Syllabus):
- Categorise given applications into the three domains (Statistical / NLP / CV).
- Find examples of Machine Learning & Reinforcement Learning in daily life.
- Earn a credential on IBM SkillsBuild — Introduction to AI.
🎮 AI Games and Experiments
| Experiment | AI Domain | What it does |
|---|---|---|
| Semantris (Google) | NLP | Word-association game powered by semantic search. |
| Quick, Draw! (Google) | CV / Neural Net | You draw; a neural network tries to guess what it is. |
| AutoDraw (Google) | CV | Identifies what you draw and suggests polished related images. |
📚 Extension Activities
- AI in the News: research recent articles on AI advancements, ethical dilemmas or industry applications; present to class.
- AI Applications Showcase: in small groups, take one technology (virtual assistants, self-driving cars, healthcare diagnostics) — create a poster / presentation showing how it works, its benefits, drawbacks and real-world use.
- AI Coding Projects: introduce Python + libraries like TensorFlow / scikit-learn; build simple AI models — image classifiers, chatbots.
- AI Film Analysis: watch and discuss AI-themed films such as Ex Machina · Her · I, Robot · The Social Dilemma; analyse how AI is portrayed and the ethical issues raised.
Quick Revision — Key Points to Remember
- AI = machine's ability to learn patterns and make predictions. It adds value to human judgement, not replaces it.
- 3 things AI can do: understand language · recognise images · make predictions.
- AI machines learn, adapt, decide — non-AI machines only follow fixed rules.
- AI history: 1950 Turing "Computing Machinery & Intelligence" (Turing Test) · 1956 Dartmouth Conference — McCarthy coins "Artificial Intelligence" (Father of AI) · 1980-90 AI Winter · 21st century resurgence.
- "Data is the new oil" — 90 % of data created in the last 2 years.
- 3 types of AI (capability): Narrow (single task) · Broad (wider related tasks) · General (any human task) — beyond = ASI.
- 3 types of data: Structured (tables) · Unstructured (images/text/video) · Semi-structured (JSON/XML/email).
- 3 domains of AI: Statistical Data (numeric) · NLP (text/speech) · Computer Vision (images/videos).
- NLP sub-areas: NLU = understanding · NLG = generation.
- CV basics: image = grid of pixels; resolution = width × height in pixels.
- Machine Learning: learns from data, no explicit programming.
- Deep Learning: imitates brain via Neural Networks; >3 layers = Deep Neural Network.
- ML vs DL: ML = small data, low-end hardware, manual features · DL = large data, GPUs, automatic features.
- 3 types of ML: Supervised (labelled) · Unsupervised (unlabelled, finds patterns) · Reinforcement (agent + reward).
- Benefits: efficiency · better decisions · innovation · healthcare advances.
- Limitations: job displacement · ethical concerns · explainability · data privacy & security.
- Practical games: Semantris (NLP) · Quick Draw (CV/NN) · AutoDraw (CV).
- Certification: IBM SkillsBuild — Introduction to AI.