8. AI and its Applications: Questions and Answers

b) Artificial Intelligence

a) John McCarthy

He coined the term "Artificial Intelligence" at the Dartmouth Conference in 1956.

b) Labeled

Each input is already paired with a correct output.

b) Hidden layer

b) Process human language

d) All of the above

d) All of the above

a) Generate meaningful text

a) NLP and speech recognition

They also use cloud-based AI services.

b) Personalized learning

b) 1956

a) 1950

c) Reinforcement learning

b) Unsupervised learning

The data is unlabelled and the system finds clusters on its own.

b) Weight

b) Backpropagation

b) Tokenization

c) Controller

b) Image acquisition

b) Adaptive learning platform

John McCarthy.

Turing.

Expert.

Machine Learning.

Reinforcement.

Testing.

Nodes.

Forward.

Natural Language Processing (NLP).

Large Language.

Prompt.

Sensors; actuators.

Computer vision.

Facilitators.

Speech.

Discriminatory.

AI is a branch of computer science that enables machines to perform tasks which typically require human intelligence. It is the simulation of human intelligence in machines that are designed to think, make decisions, predict outcomes from patterns in data, and learn and improve through training, feedback or updated data.

Examples: voice assistants like Siri or Alexa, and Netflix recommending movies based on what you have watched.

• Thinks and learns like a human
• Mimics human intelligence
• Improves on its own
• Detects patterns and solves real-world problems
• Comprehends language and images
• Predicts and takes decisions on its own

Proposed by the British scientist Alan Turing in 1950, it is a test in which a human interacts through text with both a machine and a person. If the human cannot tell the machine from the other person, the machine passes the test, which suggests human-like conversational behaviour.

A famous meeting held in 1956 where the term "Artificial Intelligence" was coined by John McCarthy, one of the organizers. It marked the birth of AI as a field of study.

Systems of the 1980s that mimicked human decision-making using rules and knowledge from human experts. Early systems such as MYCIN (1970s) helped doctors diagnose diseases by asking questions. They were rule-based, whereas modern symptom checkers use Machine Learning.

• Smartphones: improving photo quality, filtering spam messages, powering personal assistants.
• Health care: medical diagnosis, personalized treatment plans, health trackers.
• Transportation: route-optimizing navigation apps, driver-assistance systems, self-driving vehicles.
• Finance: fraud detection, personalized financial advice, automated trading.
• Education: personalized learning, automated grading, better education tools.

AI does not have human senses such as taste or smell, so it cannot judge flavour or texture. For example, an AI-generated recipe may suggest ingredients that are not easily available or steps that are not practical. Its information may also be incomplete or unclear because of limited or biased data. AI can support humans but cannot completely replace human experience and judgment.

Ethical use ensures that AI is used for the good of society and does not cause harm to individuals or communities.

• Privacy: personal data such as names, photos and locations must be protected and not misused or shared without permission.
• Bias: if training data has unfair patterns, the AI may give discriminatory results.
• Accountability: it should be clear who is responsible for an AI decision.
• Employment: automation may replace repetitive jobs, so society must create new opportunities and help people learn new skills.

Machine Learning is a subset of AI that enables computers to learn from data without being explicitly programmed. Instead of step-by-step instructions, we provide examples and the machine recognizes patterns and makes predictions.

1. Supervised learning
2. Unsupervised learning
3. Reinforcement learning

The computer is trained using labeled data, i.e. each input is already associated with a correct output. The model learns the relationship between input and expected output and then predicts on new, unseen data.

• Identifying ripe and unripe mangoes from labeled images.
• Email spam detection from emails labeled "spam" or "not spam".

The computer works with unlabelled data, meaning there are no predefined answers. The model explores the data on its own to find hidden patterns, similarities or structures, often grouping similar items into clusters.

• Grouping students based on learning styles.
• Clustering e-commerce customers by shopping behaviour, preferences and purchase history.

The machine learns through trial and error. It performs actions, gets rewards for correct decisions and penalties for wrong ones, and improves its strategy to maximize rewards. It is commonly used in game development, robotics and control systems.

• A robot learning to navigate a maze, rewarded when it reaches the endpoint.
• AI playing chess or Go against itself and improving with every win or loss.

The machine picks up knowledge from its observations, so data should be accurate, complete and relevant. Predictions are unreliable if the training data is inaccurate, biased or lacking. For example, a facial recognition system trained mostly on images of people from one ethnicity may fail to recognize people from other ethnicities.

Training data is the initial dataset used to train an AI model so that it learns patterns or performs particular tasks. Testing data is used to evaluate the performance of the model.

Feature selection is choosing the most relevant features from a dataset to build a model. It improves performance, reduces complexity and makes the model easier to understand. Example: predicting student performance from attendance, study hours and previous grades.

• Healthcare: predicting disease outbreaks, diagnosing illnesses from medical images
• Finance: detecting fraud in credit card transactions
• E-commerce: product recommendation systems like Amazon or Flipkart
• Education: adaptive learning platforms that adjust lessons to student performance
• Agriculture: monitoring crop health and predicting yield

An Artificial Neural Network (ANN) is a computational model inspired by the human brain. The brain has billions of interconnected neurons that process information; similarly, an ANN uses artificial neurons (nodes) to analyze data, detect patterns and make predictions. Like humans, ANNs learn from experience.

• Input layer: entry point for data; each neuron represents a feature. In image recognition, each pixel's intensity may be an input neuron.
• Hidden layers: perform the crucial computations by transforming input through various functions; there can be several depending on task complexity.
• Output layer: generates the final prediction or output.
• Weights: the strength of each connection, which the network learns and tweaks during training.

1. Forward propagation: input data flows through the hidden layers to the output layer.
2. Error calculation: the difference between predicted and actual output is measured.
3. Backpropagation: weights of connections are adjusted to minimize the error.
4. Iteration: the process repeats until the network reaches maximum accuracy.

• Image recognition: Facebook tagging, Google Photos, autonomous vehicles
• Speech recognition: Siri, Google Assistant
• Medical diagnosis: detecting tumors or abnormalities in X-rays, MRI or CT scans
• Natural Language Processing: chatbots, translation tools, sentiment analysis
• Gaming: AI learning to play chess or Go

NLP is a branch of AI that helps computers understand, interpret and generate human language such as Malayalam, Hindi and English. It combines computer science, linguistics and machine learning. NLP enables computers to:

• Read and understand text such as emails, messages or documents
• Listen to and interpret speech for voice-based interaction
• Translate between languages
• Generate meaningful responses

• Tokenization: breaking text into smaller units like words or sentences
• Parsing: analyzing the grammatical structure of sentences
• Semantic analysis: understanding the meaning of words and sentences
• Response generation: producing meaningful, context-aware outputs

• Speech recognition: turns spoken words into text, e.g. voice-to-text typing
• Machine translation: translates text between languages, e.g. Google Translate
• Chatbots and virtual assistants: simulate human conversation, e.g. banking chatbots
• Sentiment analysis: detects positive, negative or neutral sentiment in posts, reviews or feedback
• Content generation: AI creates new text such as articles and poetry

Modern NLP systems use LLMs such as GPT, which are trained on massive datasets. This lets them understand context and generate meaningful responses, powering chatbots, writing assistance and more.

A prompt is a question or instruction given to an AI. The quality of the output largely depends on how well the prompt is formulated, and poorly worded prompts can give irrelevant or incorrect output.

• Be specific: define the scope and direction precisely. Example: "List three advantages and two disadvantages of AI in education."
• Provide context: background, target audience or prior information. Example: "As a teacher, explain how AI can help students improve reading."
• Define output requirements: word count, bullet points, format. Example: "Write a 100-word story suitable for class 11 students about robots in space exploration."
• Use examples: shows the desired tone, structure and detail; useful for complex or creative tasks.

• Do not ask AI to generate personal or sensitive information.
• Avoid spreading misinformation.
• Do not use AI for harmful or dishonest purposes.
• Promote learning, creativity and responsible use of AI.

• Education: explaining concepts, summarizing content, creating quizzes
• Content creation: articles, blogs, stories, poetry
• Programming: generating code snippets and debugging solutions
• Data analysis: summarizing trends and generating visualizations

A robot is a programmable machine that interacts with its surroundings and acts either fully or partially on its own.

• Sensors: perceive the environment by detecting light, heat, sound, motion or pressure. Example: obstacle detection sensors in an autonomous vehicle.
• Controllers: the "brain" of the robot; process information and decide the next action using AI or pre-programmed instructions.
• Actuators: perform movements or actions. Example: motors, wheels, arms or grippers.

Optional components: power supply (battery), communication modules, cameras and additional computing units.

1. Sensing: robots gather data from sensors.
2. Processing: the controller interprets data and decides on actions.
3. Acting: actuators execute movements or tasks.
4. Feedback: sensors continuously provide updated information for adjustments.

Applications

• Manufacturing: assembly line robots increase efficiency and precision
• Healthcare: surgical robots assist doctors in delicate operations
• Space exploration: rovers explore distant planets
• Disaster rescue: robots enter collapsed buildings or nuclear sites
• Education: robots help students learn coding, AI and automation

Ethical considerations

• Employment impact: automation may replace manual jobs
• Safety: robots must be programmed to avoid harming humans
• Responsibility: clear accountability for robot actions
• Privacy: robots with cameras or sensors must respect data privacy

Computer vision (CV) is a subfield of AI that gives computers the ability to recognize, evaluate and comprehend visual data such as pictures and videos, so machines can "see" their environment like humans do with their eyes.

1. Image acquisition: capture images using cameras or sensors.
2. Pre-processing: remove noise or adjust brightness.
3. Feature extraction: identify edges, corners or textures.
4. Recognition and classification: use features to identify objects or patterns.
5. Decision making: use the results to act, e.g. guide a robot or flag something.

Example: a smartphone camera captures a face, identifies its features, compares them with stored data and unlocks the phone if they match.

Applications

• Medical imaging: detecting tumors, fractures or anomalies in X-rays, MRI and CT scans
• Autonomous vehicles: detecting obstacles, lane markings and traffic signs
• Security systems: face recognition for surveillance and access control
• Industrial quality control: inspecting products for defects
• Agriculture: monitoring crops, identifying diseased plants, estimating fruit ripeness

Challenges

• Privacy: surveillance and facial recognition may intrude on personal privacy
• Bias: misidentification if trained on non-diverse datasets
• Safety: autonomous vehicles rely on accurate computer vision to avoid accidents

Personalized learning means educational tools analyze a student's strengths, weaknesses, learning pace and style, and then suggest exercises, lessons or resources tailored to that student.

• Adaptive learning platforms: adjust pace and difficulty based on performance. Example: Khan Academy suggests lessons from prior quiz performance.
• Digital tutors and chatbots: instant feedback and personalized help.
• Learning analytics: tracks progress and highlights areas needing improvement.
• Automated grading: grades assignments and quizzes, reducing teacher workload.
• Inclusive education: voice recognition, text-to-speech and visual aids support students with disabilities.

Teachers become facilitators who guide students through the personalized paths AI suggests. They focus on critical thinking, creativity and social skills, guide, mentor and motivate students, and monitor the ethical use of AI tools. Teachers remain central to the learning process.

Ethical considerations:

• Data privacy: student information must be protected.
• Transparency: students should understand how AI tools make recommendations.
• Equity: AI systems should not favour certain groups over others.

A voice assistant is an AI program that comprehends and reacts to spoken instructions, such as playing music, setting reminders, answering questions and controlling smart devices. It combines speech recognition, NLP and cloud-based AI services.

1. Speech recognition: converts voice into text.
2. Natural Language Processing: interprets the meaning and intent, considering context.
3. Decision making: determines the most appropriate response or action.
4. Response generation: creates a clear, personalized spoken or written response.

Example: when you say "Hey Siri, what's the weather today?", Siri converts the speech to text, analyses the query, fetches weather data from the internet and replies aloud with the temperature.

• Smart phones and smart homes: controlling lights, fans, thermostats or security systems by voice
• Healthcare: helping elderly or differently-abled people set reminders or call for help
• Navigation: voice commands in GPS systems for routes and traffic updates
• Productivity: setting alarms and reminders, creating calendar events, sending messages

Alexa, Amazon's voice assistant, lets users shop online, play music and control smart home devices through voice commands. Over time it learns user preferences, making interactions faster and more personalized.

Opportunities

• Healthcare: faster diagnosis from medical images and personalized treatment.
• Education: adaptive learning and instant feedback from digital tutors.
• Finance: fraud detection in credit card transactions.
• Agriculture: monitoring crop health and predicting yield.
• Accessibility: voice assistants and text-to-speech help people with disabilities.

Challenges

• Privacy: large amounts of personal data are collected and may be misused.
• Bias: biased training data gives unfair results, e.g. face recognition failing for some groups.
• Accountability: unclear responsibility for harmful AI decisions.
• Employment: repetitive jobs may be replaced.
• Dependency: over-reliance may reduce critical thinking.

AI should therefore be used responsibly, with new job opportunities and skill training created alongside it.

Idea: an AI-based study-support system for a school (sample answer).

• Data needed: attendance records, study hours, marks in earlier tests and quiz results.
• How it works: a supervised machine learning model is trained on past data to predict which students may need extra help. A chatbot then suggests practice exercises matched to each student's weak topics.
• Benefits: teachers identify struggling students early, students learn at their own pace, and time is saved on repetitive tracking.
• Ethical care: keep student data private, explain to students how the system works and make sure it treats all groups fairly.

No. AI is likely to support teachers rather than replace them (sample answer).

• Teachers provide mentoring, motivation and emotional support that AI cannot.
• Critical thinking, creativity and social skills develop best through human interaction.
• Teachers must monitor ethical use of AI tools and check for bias.
• AI lacks human judgment and may give incomplete or biased information.
• AI helps by grading, giving insights through learning analytics and offering personalized content, so teachers can focus on guiding students.

• Train speech recognition on large, diverse voice samples covering many languages (e.g. Malayalam, Hindi, English) and regional accents and dialects.
• Use NLP models that understand context and intent in each language, including mixed-language speech.
• Add machine translation so the assistant can switch between languages.
• Use a feedback loop so the assistant keeps learning from corrections.
• Test with users from different regions to detect bias and misinterpretation of accents.
• Protect voice data and ask permission before storing it.

• Ethical: impact on workers' jobs (plan retraining and new roles); clear accountability for robot actions; respect privacy if cameras are used.
• Safety: robots must be programmed to avoid harming humans; safe zones, emergency stop and regular maintenance.
• Efficiency: computer vision can inspect products for defects faster and more precisely and work continuously; check the cost, accuracy and reliability of sensors and controllers.

Possible misuse: spreading misinformation, generating harmful or dishonest content, copying work and cheating in exams, asking for or exposing personal or sensitive information, and over-reliance that reduces critical thinking.

Preventive measures:

• Follow ethical guidelines for prompting.
• Verify AI output against reliable sources.
• Teach responsible use in schools.
• Protect personal data.
• Set clear policies and accountability for developers and users.
• Train models on fair, good-quality data to reduce bias.

Computer vision can monitor crops, identify diseased plants and estimate fruit ripeness.

Scenario (sample): a farmer flies a drone or uses a phone camera over a paddy field.

1. Images are captured (image acquisition).
2. Noise and brightness are corrected (pre-processing).
3. Features such as leaf colour, spots and texture are extracted.
4. The system classifies leaves as healthy or diseased.
5. The farmer is alerted about the affected area, so only that part is treated.

This saves cost, protects yield and reduces unnecessary use of chemicals.

Scenario (sample): a student types "Write about pollution." The AI gives a generic answer mixing air, water and soil pollution that does not suit the assignment.

Correction: make the prompt specific, give context and state output requirements, e.g. "As a class 11 student, write five sentences explaining how air pollution affects human health, with examples."

• Education: voice recognition, text-to-speech and visual aids support participation; digital tutors give personalized, patient help.
• Daily life: voice assistants allow hands-free control of lights, fans and devices, and help set reminders or call for help.
• Healthcare: medication reminders and monitoring for elderly or differently-abled people.
• Limitations: voice assistants may misunderstand some accents and voice data raises privacy concerns, so design and testing must include diverse users.

Overall, AI increases independence and inclusion when it is designed fairly.

Analysis: the specific prompt defines scope, audience, length and format, so the result is relevant and ready to use. The vague prompt needs follow-up prompts to get something useful.

 +-----------+      +--------------+      +-------------+
 |  SENSORS  | ---> |  CONTROLLER  | ---> |  ACTUATORS  |
 | (camera,  |      | (processing  |      | (motors,    |
 |  mic etc) |      |  & decision) |      |  wheels,    |
 +-----------+      +--------------+      |  grippers)  |
       ^                                  +-------------+
       |                                         |
       +------- Feedback loop (continuous) ------+

Sensors collect data, the controller processes it and decides, actuators carry out the action, and the feedback loop continuously updates the sensors' readings.

Healthcare: diagnosing illness from X-rays (sample answer). A neural network is trained on thousands of labeled X-ray images (supervised learning). Pixels enter the input layer, hidden layers detect features such as edges and shapes, and the output layer predicts whether an abnormality is present. A doctor reviews the result, so AI supports rather than replaces human judgment.

• Computer vision on satellite or drone images can detect deforestation, water pollution or waste dumping.
• Machine learning on sensor data can predict air quality and weather-related risks.
• Ethical use: use only data collected with permission, protect privacy (avoid unnecessary surveillance of individuals), make sure data covers all regions fairly, be transparent about how results are produced and keep humans responsible for decisions.

Sample answer:

• Smartphone voice assistant: sets alarms and answers spoken questions.
• Video or music recommendations: suggests content based on what you watched or listened to earlier.
• Email spam filter: separates spam from genuine mail.

Discussion points:

• Smart traffic management and better navigation to cut travel time.
• Early prediction of disease outbreaks and faster diagnosis in local hospitals.
• Personalized learning for students in schools.
• Fraud detection in digital payments.
• Crop monitoring to support local farmers.

Ask a simple fact ("What is the capital of India?"), a context-based request ("Remind me to submit my record book at 4 pm") and a vague one ("Tell me about it"). Evaluate each for accuracy, relevance, clarity and whether the assistant understood the intent. Usually the first two are answered well, while the vague one gets a generic or clarifying reply, which shows the importance of good prompts.

Discussion points:

• Review and delete stored voice recordings regularly.
• Turn off the microphone when not in use.
• Do not speak passwords or sensitive details to the assistant.
• Give app permissions only when necessary.
• Use trusted devices and keep software updated.

Discussion points:

• Get consent before capturing or storing people's images.
• Train systems on diverse datasets to avoid bias and misidentification.
• Protect stored image data from misuse and security risks.
• Test systems thoroughly where safety is critical, e.g. autonomous vehicles.
• Keep human oversight and clear accountability.