AI Foundations — Unit 1 (Class IX) Summary & Study Notes

🤖 Understanding Artificial Intelligence

Artificial Intelligence (AI) is when a machine can mimic human traits: make decisions, predict, learn, and improve on its own. AI systems collect data, analyze it, learn patterns, and improve performance. The term “Artificial Intelligence” was coined by John McCarthy (1956). AI is both a field of study and a type of technology that aims to automate tasks that normally require human intelligence.

🧭 AI Domains

AI processes different kinds of data and can be grouped into three main domains:

• Data: Uses structured data to recognize patterns and make decisions (e.g., fraud detection, crop yield prediction).
• Computer Vision: Works with images and videos to identify objects, faces, or help in medical diagnosis.
• Natural Language Processing (NLP): Handles human language for understanding, interpreting, and generating text or speech (e.g., chatbots, translators).

🎮 AI in Games — Examples

AI is taught and demonstrated through simple games:

• Rock, Paper & Scissors: Uses historical move data to predict the next move (data-driven).
• Semantris: A word-association game that uses NLP to find related words.
• Quick, Draw: A Computer Vision task where a neural network guesses sketches.

🛠️ Applications of AI

Common real-world applications include speech recognition (Siri, Google Assistant), face recognition, autocorrect and grammar tools, search engines, recommendation systems (Netflix, Amazon), navigation and ride-sharing, robotics, and smart assistants (Alexa, Google Home).

🔄 AI Project Cycle

The AI Project Cycle is a step-by-step process to build AI solutions. Main stages:

• Problem Scoping: Define the problem, purpose, and goals. Use the 4Ws (Who, What, Where, Why) to identify stakeholders and scope.
• Data Acquisition: Collect authentic, relevant data from surveys, sensors, APIs, web scraping, cameras, public portals, or observations.
• Data Exploration: Clean data, fix errors, remove incomplete records, and visualize patterns with pie charts, bar graphs, histograms, and scatter plots. Use quick sketches (SketchyGraphs) for early insights.
• Modelling: Build models using either Rule-Based approaches (explicit rules) or Learning-Based approaches (models learn from data).
• Evaluation: Test models on separate testing data (not training data) to avoid overfitting. Key metrics include True Positive (TP), True Negative (TN), False Positive (FP), False Negative (FN), and tools like the ROC curve.
• Deployment: Integrate the model into real-world systems, and perform testing, monitoring, and maintenance.

🧾 Problem Scoping — 4Ws & Problem Statement

The 4Ws Problem Canvas helps frame a problem clearly:

• Who is affected?
• What is the problem?
• Where does it occur?
• Why must it be solved?

A clear Problem Statement Template captures these points so teams can revisit the basis of the project later.

🧰 Data Concepts — Training vs Testing

• Training data: Used to build and train the AI model. It contains inputs and their correct outputs so the model can learn patterns.
• Testing data: A separate dataset used to evaluate model performance on unseen examples.

🧩 System Maps and Cause–Effect

A System Map shows system components and boundaries, and how elements influence each other with directional arrows and signs (+ or −). A + sign means direct relationship (both increase together), and a − sign means inverse relationship.

🔬 Modelling: ML vs DL

• Artificial Intelligence (AI): Any technique enabling computers to mimic human intelligence.
• Machine Learning (ML): Enables machines to improve at tasks through experience.
• Deep Learning (DL): Uses large datasets and multi-layered neural networks to learn complex patterns; models can develop internal representations and features from raw data.

✅ Evaluation & Overfitting

Evaluation checks model reliability using testing data. Overfitting occurs when a model memorizes training data and performs poorly on unseen data. Use separate test data and evaluation metrics (TP, TN, FP, FN, accuracy, precision, recall, ROC) to detect and avoid overfitting.

🚀 Deployment Best Practices

Key steps in deployment:

• Test and validate the model thoroughly.
• Integrate with existing systems and user interfaces.
• Monitor performance and update the model as new data arrives.

⚖️ Ethics and Morality in AI

Important ethical principles:

• Human Rights: Ensure AI does not remove freedom, discriminate, or deprive people of basic rights.
• Bias: Bias often comes from skewed training data. Ask whether data represents all groups equally and whether the AI will discriminate.
• Privacy: Protect personal data, disclose what is collected, and ensure data is used safely.
• Inclusion: Make AI accessible and beneficial for diverse populations, not only the privileged. Consider usability and equitable benefit.

Consider these principles at every stage of the AI Project Cycle to build responsible systems.