Class 9 AI Reflection, Project Cycle & EthicsNotes, Questions from CBSE Curriculum, MCQs & Q&A

Answer:

When a machine possesses the ability to mimic human traits—such as making decisions, predicting future events, learning, and improving on its own—it is said to have Artificial Intelligence.

Answer:

Natural Language Processing.

Answer:

1. Statistical Data, 2. Computer Vision (CV), and 3. Natural Language Processing (NLP).

Answer:

Who, What, Where, and Why.

Answer:

Overfitting is when an AI model memorizes the training data completely, causing it to predict training points perfectly while failing to generalize to new, unseen testing data.

Answer:

1. Efficiency: Allows teams to develop better AI solutions faster with less effort. 2. Modularity: Breaks complex challenges into simpler, self-contained sub-stages so that if a model fails, the team only needs to modify that specific stage rather than rebuilding everything.

Answer:

In a System Map, arrows show cause-and-effect relationships. A '+' sign indicates a direct relationship (if X increases, Y increases). A '-' sign indicates an inverse relationship (if X increases, Y decreases). Longer arrows represent time delays.

Answer:

In a Rule-Based model, developer-written rules strictly dictate behavior, and learning is static (e.g., Golf Play decision tree). In a Learning-Based model, the machine learns features and relationships directly from data and adapts dynamically to handle exceptions (e.g., Apple vs. Banana classifier).

Answer:

Morals are societal beliefs dictated by culture or tradition (e.g., 'always speak the truth'). Ethics are guiding principles that an individual chooses to evaluate good vs. bad in complex real-world situations (e.g., 'is it ethical to withhold truth if speaking it will endanger someone's life?').

Answer:

1. Problem Scoping: Identifying the problem, scoping goals using the 4Ws Canvas, and drafting a Problem Statement Template. 2. Data Acquisition: Gathering authentic, relevant training and testing data via surveys, sensors, APIs, or open-source portals (data.gov.in). 3. Data Exploration: Visualizing data trends, distributions, and patterns using graphs (bar, line, scatter) to guide model strategy. 4. Modeling: Selecting or engineering algorithms (Rule-Based vs. Learning-Based) and training them on acquired datasets. 5. Evaluation: Testing model performance on unseen test data using Confusion Matrix parameters (TP, FP, TN, FN) and ROC curves to guard against overfitting. 6. Deployment: Packaging the validated model into real-world applications (mobile apps, web services) with continuous monitoring and maintenance.

Answer:

1. Human Rights: Ensuring AI preserves freedom, prevents bodily or financial harm, and does not unjustly displace workers without reskilling. 2. Bias: Preventing partiality and unfair discrimination by curating balanced, representative training datasets (avoiding scenarios like search engines showing only women for 'personal secretary'). 3. Privacy: Strictly safeguarding personal user data with explicit consent, transparent data usage policies, and end-to-end security against breaches. 4. Inclusion: Designing accessible, equitable AI tools that empower diverse socioeconomic groups, regardless of wealth, language, or geography.

Answer:

• Problem Scoping: 6 million Indian farmers suffer crop loss from Pink Bollworms invisible to the naked eye. Goal: create an advisory app to optimize pesticide spraying. • Data Acquisition: Collecting photos of pests, farm sizes, farmer names, and pesticide usage records. • Data Exploration: Cleaning non-uniform images and graphing infestation patterns vs. seasonal weather. • Modeling: Developing a Computer Vision model trained on labeled pest photos. • Evaluation: Testing on blank sheets of paper with captured insects; initial 70% accuracy was refined with improved algorithms. • Deployment: Releasing the CottonAce mobile app, leading to a 26.5% jump in farm profit margins and a 38% drop in pesticide costs.

Answer:

• True Positive (TP): Reality = Fire has broken out; AI Prediction = Fire detected. (Correct alarm) • True Negative (TN): Reality = No fire; AI Prediction = No fire. (Correct quiet state) • False Positive (FP): Reality = No fire; AI Prediction = Fire detected. (False alarm, sending fire engines unnecessarily) • False Negative (FN): Reality = Fire has broken out; AI Prediction = No fire. (Dangerous miss, fire spreads unchecked without warning).

Answer:

This represents Algorithmic Bias. It occurred because the machine learning model was trained on historical company hiring data from prior years when male chefs dominated the applicant pool. The AI incorrectly learned that male candidates were statistically preferable, turning historical societal discrimination into an active algorithm rule.