🧠 Artificial Intelligence (AI) — Complete Study Guide

Your Ultimate Resource for Competitive Exam Preparation — UPSC, SSC, Banking, Railway & More

📚 100+ MCQs 🏛️ Govt Initiatives 📊 Comparison Tables 🎯 Previous Year Questions ⚡ Quick Revision

📋 Table of Contents

What is Artificial Intelligence?
History & Key Milestones of AI
Types of AI (Classification)
Branches of AI
Machine Learning — Types & Concepts
Key AI Terminology for Exams
AI in Banking & Finance
AI in Government & Public Sector (India)
Indian Govt Initiatives & Policies on AI
AI Ethics, Risks & Responsible AI
Latest AI Trends & Technologies (2024–2026)
AI & Cybersecurity
Previous Year Exam Questions
Practice MCQs — Chapter-wise (100 Questions)
Quick Revision One-Liners

🤖 1. What is Artificial Intelligence (AI)?

Artificial Intelligence (AI) is the ability of machines (especially computers) to perform tasks that normally require human intelligence — such as learning, reasoning, problem-solving, understanding language, and recognizing patterns.

In simple words: When a machine acts “smart” like a human brain — that’s AI.

🎯

Exam Tip: The term “Artificial Intelligence” was first coined by John McCarthy in 1956 at the Dartmouth Conference. He is known as the “Father of Artificial Intelligence.” — Frequently asked in SSC, Banking & UPSC Prelims.

Goals of AI

🧩

Learn

From data & experience

🔍

Reason

Logical thinking

💡

Decide

Smart decision-making

🗣️

Interact

Understand language

👁️

Perceive

See & hear like humans

Key Relationship: AI is the broadest concept → Machine Learning (ML) is a subset of AI → Deep Learning (DL) is a subset of ML.

Artificial Intelligence ⊃ Machine Learning ⊃ Deep Learning

📜 2. History & Key Milestones of AI

1950

Alan Turing’s “Turing Test”

Alan Turing proposed the Turing Test — if a machine can fool a human into thinking it’s human during a conversation, it can be considered “intelligent.” Published in the paper “Computing Machinery and Intelligence.”

1956

Birth of AI — Dartmouth Conference

John McCarthy organized the Dartmouth Conference and coined the term “Artificial Intelligence.” This is considered the official birth of AI as a field.

1966

ELIZA — First Chatbot

Joseph Weizenbaum created ELIZA, the first chatbot that could simulate a conversation with a therapist.

1974–1980

First AI Winter

Funding and interest in AI declined sharply due to unmet expectations. This period is called the “AI Winter.”

1997

IBM Deep Blue Defeats Chess Champion

IBM’s Deep Blue defeated world chess champion Garry Kasparov. A landmark moment for AI.

2011

IBM Watson Wins Jeopardy!

IBM Watson defeated two human champions on the TV quiz show Jeopardy!, showcasing NLP capabilities.

2016

Google AlphaGo Defeats Go Champion

Google DeepMind’s AlphaGo beat the world champion in Go — a game far more complex than chess.

2017

Transformer Architecture Introduced

Google published the revolutionary paper “Attention is All You Need” — the foundation for all modern LLMs like ChatGPT and Gemini.

2022

ChatGPT Launched

OpenAI released ChatGPT (based on GPT-3.5), bringing Generative AI to the mainstream. It became the fastest-growing consumer app in history.

2023–2026

AI Revolution Accelerates

GPT-4, Google Gemini, Claude, and India’s own BharatGPT / Hanooman launched. India announced IndiaAI Mission with ₹10,372 crore budget. EU passed AI Act.

⚡ Quick Recall — AI History

Father of AI → John McCarthy (1956)
Turing Test → Alan Turing (1950)
First Chatbot → ELIZA (1966)
Deep Blue vs Kasparov → 1997 (Chess)
AlphaGo → 2016 (Go Game, by Google DeepMind)
Transformer Paper → 2017 (“Attention is All You Need”)
ChatGPT Launch → November 2022 (OpenAI)
IndiaAI Mission → March 2024 (₹10,372 Crore)

📊 3. Types of Artificial Intelligence

Classification by Capability

Type	Also Called	Description	Example	Status
Narrow AI (ANI)	Weak AI	Designed for one specific task only. Cannot do anything beyond its training.	Siri, Alexa, Google Maps, Spam Filters, Face Unlock	Exists Today
General AI (AGI)	Strong AI	Can perform any intellectual task that a human can do. Equal to human intelligence.	No real example yet (Theoretical)	Theoretical
Super AI (ASI)	Superintelligence	Surpasses human intelligence in every field — science, creativity, emotions, social skills.	No real example (Hypothetical)	Hypothetical

🎯

Exam Tip: All AI we use today (ChatGPT, Alexa, self-driving cars) is Narrow AI (Weak AI). General AI and Super AI do not exist yet. This is frequently asked in banking and SSC exams.

Classification by Functionality

Type	Description	Example
Reactive Machines	Reacts to current input only. No memory, no past learning.	IBM Deep Blue
Limited Memory	Can use recent past data to make decisions. Most current AI falls here.	Self-driving cars, ChatGPT
Theory of Mind	Can understand emotions, beliefs, and thoughts of others. Not yet achieved.	Under research
Self-Aware AI	AI with its own consciousness and self-awareness. Purely hypothetical.	Science fiction concept

🌿 4. Branches of Artificial Intelligence

🤖 Machine Learning (ML)

Machines learn from data without being explicitly programmed. Uses algorithms to find patterns and make predictions.

🧠 Deep Learning (DL)

A subset of ML using multi-layer neural networks (inspired by the human brain) to process complex patterns like images and speech.

🗣️ Natural Language Processing (NLP)

Enables machines to understand, interpret, and generate human language. Powers chatbots, translators, and voice assistants.

👁️ Computer Vision

Allows machines to “see” and interpret images/videos. Used in face recognition, medical imaging, and self-driving cars.

🦾 Robotics

Combines AI with mechanical engineering to create robots that interact with the physical world. Used in manufacturing, surgery, etc.

💬 Expert Systems

AI programs that mimic the decision-making of a human expert in a specific field (e.g., medical diagnosis, financial advising).

🎮 Reinforcement Learning

AI learns by trial and error — receives rewards for correct actions and penalties for wrong ones. Used in game-playing AI.

🎨 Generative AI

AI that can create new content — text, images, music, code. Examples: ChatGPT, DALL-E, Midjourney, Google Gemini.

⚙️ 5. Machine Learning — Types & Key Concepts

Three Main Types of Machine Learning

Feature	Supervised Learning	Unsupervised Learning	Reinforcement Learning
Data Type	Labelled data (Input + Output given)	Unlabelled data (Only input given)	No data — learns by interaction
Goal	Predict/classify outcomes	Find hidden patterns/groups	Learn best actions via rewards
Techniques	Regression, Classification	Clustering, Association	Q-Learning, Policy Gradient
Example	Email spam detection, Loan default prediction	Customer segmentation, Market basket analysis	Game-playing AI (AlphaGo), Autonomous robots
Banking Use	Credit scoring, Fraud detection	Customer grouping for marketing	Dynamic pricing, Portfolio optimization

Important ML Concepts for Exams

Concept	Simple Meaning	Exam Relevance
Training Data	Data used to teach the model patterns	High
Testing Data	New data to check how well the model learned	High
Overfitting	Model memorizes training data but fails on new data	Very High
Underfitting	Model is too simple to capture patterns — poor performance everywhere	Medium
Feature	An individual measurable property of data (e.g., age, income, location)	Medium
Algorithm	Step-by-step method for solving a problem	Very High
Neural Network	Computing system inspired by the human brain’s neuron connections	Very High
Model Bias	When AI gives unfair results due to biased training data	Very High
Model Drift	Decline in model accuracy over time as real-world data changes	High
Black Box	AI model whose internal decision logic is not easily understandable	High

📖 6. Key AI Terminology for Competitive Exams

Term	Full Form / Meaning	Simple Explanation
LLM	Large Language Model	AI models trained on massive text data to understand and generate language (e.g., GPT-4, Gemini)
GenAI	Generative AI	AI that creates new content — text, images, audio, video, code
NLP	Natural Language Processing	AI’s ability to understand and process human language
Chatbot	Conversational AI Agent	AI program that simulates human conversation
Transformer	Neural Network Architecture	AI model architecture using “attention mechanism” — powers all modern LLMs
GPT	Generative Pre-trained Transformer	Type of LLM by OpenAI that generates human-like text
AGI	Artificial General Intelligence	Hypothetical AI with human-level intelligence across all tasks
RLHF	Reinforcement Learning from Human Feedback	Training method using human ratings to improve AI responses
OCR	Optical Character Recognition	AI that converts printed/handwritten text in images to digital text
IoT	Internet of Things	Network of connected devices that communicate and share data
RPA	Robotic Process Automation	Software bots that automate repetitive, rule-based tasks
AML	Anti-Money Laundering	AI systems that detect suspicious financial transactions
KYC	Know Your Customer	AI automates identity verification for customer onboarding
Deepfake	Deep Learning + Fake	AI-generated fake video/audio that looks real — a cybersecurity concern
Prompt	User Input to AI	The question or instruction given to a GenAI model
Hallucination	AI Error	When AI generates confident but incorrect or made-up information
Federated Learning	Privacy-Preserving ML	Training AI on distributed data without centralizing it — protects privacy
Edge AI	On-Device AI	Running AI on local devices (phones, IoT) instead of cloud — faster, private

🏦 7. AI in Banking & Finance

🎯

Exam Alert: AI in Banking is one of the most frequently tested topics in IBPS, SBI PO/Clerk, and RBI Grade B exams. Focus on applications and RBI guidelines.

Key Applications of AI in Banking

Application Area	How AI Helps	Example
🔍 Fraud Detection	Detects unusual transaction patterns and anomalies in real-time	Flagging suspicious UPI transactions
📊 Credit Scoring	Predicts borrower risk using multiple data variables	Alternative credit scoring for new-to-credit customers
💬 Chatbots	Handles customer queries 24×7 using NLP	SBI’s SIA, HDFC’s EVA, ICICI’s iPal
📋 KYC Automation	Automates face matching and document verification	Video KYC, Aadhaar-based eKYC
🛡️ AML Monitoring	Monitors large volumes of transactions for suspicious activity	Compliance with PMLA regulations
💰 Robo-Advisors	Provides automated, personalized investment recommendations	Wealth management for retail investors
📑 Document Processing	Uses OCR + NLP to extract data from forms and documents	Loan application processing
📈 Risk Management	Provides early warning signals for stressed accounts	NPA prediction models
🔔 Personalization	Customizes product offers based on customer behaviour	Cross-selling, targeted marketing
🔍 Internal Audit	Analyses 100% transactions to highlight anomalies	Continuous auditing systems

AI Chatbots in Indian Banking

Bank	AI Chatbot Name	Launched
SBI	SIA (SBI Intelligent Assistant)	2017
HDFC Bank	EVA (Electronic Virtual Assistant)	2017
ICICI Bank	iPal	2018
Axis Bank	Axis Aha!	2018
Yes Bank	YES ROBOT	2019
Bank of Baroda	ADI	2018
Canara Bank	Mitra	2019
IndusInd Bank	Alexa-based banking	2019

🏛️ 8. AI in Government & Public Sector (India)

Sector	AI Application	Real Example in India
🌾 Agriculture	Crop disease detection, weather prediction, soil analysis	Kisan Suvidha App, AI-based crop advisory
🏥 Healthcare	Disease diagnosis, drug discovery, telemedicine	AI tools in Ayushman Bharat for screening
📚 Education	Personalized learning, automated grading, smart tutoring	DIKSHA platform, AI-based learning apps
🚦 Smart Cities	Traffic management, waste management, surveillance	AI-based traffic signals in Bengaluru, Delhi
🛡️ Defence	Surveillance, threat analysis, autonomous systems	DRDO’s AI-powered defence systems
⚖️ Judiciary	Case analysis, predicting case outcomes, legal research	SUPACE (Supreme Court Portal for Assistance in Court Efficiency)
🚆 Railways	Predictive maintenance, dynamic pricing, crowd management	Indian Railways’ AI for rake management
🏦 Taxation	Tax fraud detection, automated processing	Project Insight by Income Tax Department

🇮🇳 9. Indian Government Initiatives & Policies on AI

🎯

Exam Alert: Government AI initiatives are very important for UPSC, SSC CGL, Banking (IBPS/SBI), and Railway exams. Expect questions on IndiaAI Mission, NITI Aayog strategy, and budget allocations.

Initiative / Policy	Year	Key Details
NITI Aayog — National Strategy for AI	2018	Released the paper “National Strategy for Artificial Intelligence #AIForAll”. Identified 5 focus sectors: Healthcare, Agriculture, Education, Smart Cities, Smart Mobility.
National AI Portal (INDIAai)	2020	Launched by MeitY & NASSCOM as a knowledge hub for AI-related news, articles, and resources. Website: indiaai.gov.in
Responsible AI Principles	2021	NITI Aayog released two-part approach paper on Responsible AI — covering principles like safety, equality, inclusivity, privacy, transparency, accountability.
IndiaAI Mission	2024	Approved by Cabinet with budget of ₹10,372 crore. 7 pillars: IndiaAI Compute, IndiaAI Innovation Centre, IndiaAI Datasets Platform, IndiaAI Application Development, IndiaAI FutureSkills, IndiaAI Startup Financing, Safe & Trusted AI.
IndiaAI Compute Capacity	2024	Building 10,000+ GPU compute infrastructure for AI research and startups. Public-Private Partnership model.
AI Centres of Excellence (CoE)	2024–25	Establishing AI CoEs in key sectors — Healthcare, Agriculture, Sustainable Cities. Collaborative research with IITs and IIITs.
RBI on AI in Banking	2023–25	RBI issued guidelines on Responsible AI for financial institutions. Focus on explainability, fairness, model governance, and customer protection.
Digital Personal Data Protection Act (DPDPA)	2023	India’s data protection law — directly impacts how AI systems collect and use personal data. Establishes Data Protection Board of India.
BharatGPT / Hanooman	2024	India’s own multilingual LLM supporting 22+ Indian languages. Developed by IIT Bombay consortium under BharatGPT initiative.
SUPACE	2021	Supreme Court Portal for Assistance in Court Efficiency — AI tool to assist Supreme Court judges with legal research.

IndiaAI Mission — 7 Pillars (Remember as “C-I-D-A-F-S-S”):

Compute Capacity — 10,000+ GPUs
Innovation Centre — Research & development
Datasets Platform — Quality AI training data
Application Development — Real-world AI solutions
FutureSkills — AI workforce training
Startup Financing — Funding for AI startups
Safe & Trusted AI — Responsible AI framework

⚖️ 10. AI Ethics, Risks & Responsible AI

Key Ethical Principles of AI

🔍 Transparency

AI decisions should be open and understandable. Users should know when they’re interacting with AI.

⚖️ Fairness

AI should not discriminate based on gender, race, religion, or socioeconomic status.

🔐 Privacy

Personal data must be protected. AI should comply with data protection laws (DPDPA in India).

📋 Accountability

Someone (person or organization) must be responsible for AI’s decisions and outcomes.

🛡️ Safety

AI systems must be safe, reliable, and should not cause harm to users or society.

🤝 Inclusivity

AI should benefit all sections of society, not just the privileged. Access should be equal.

Major Risks of AI

Risk	Description	Example
Bias & Discrimination	Unfair outcomes due to biased training data	AI rejecting loans for specific demographics
Job Displacement	Automation replacing human workers	AI replacing data entry and call centre jobs
Privacy Violation	Unauthorized collection/use of personal data	AI surveillance without consent
Deepfakes	AI-generated fake videos/audio used for fraud	Fake videos of political leaders, CEO fraud
AI Hallucination	AI generates false information confidently	ChatGPT inventing fake court cases, references
Lack of Explainability	“Black box” models give decisions without reasons	Credit denial without clear explanation
Security Threats	AI systems can be hacked or manipulated	Adversarial attacks fooling image recognition

“Human-in-the-Loop” (HITL): This means keeping human oversight in AI decision-making, especially for high-risk decisions like loans, legal judgments, and medical diagnoses. It reduces the risk of fully automated wrong decisions. Very Important for Exams

🚀 11. Latest AI Trends & Technologies (2024–2026)

Trend / Technology	What It Is	Why It Matters
Generative AI (GenAI)	AI creating new text, images, code, audio, video	ChatGPT, Gemini, Claude — transforming every industry
Multimodal AI	AI that can process text + images + audio + video together	Google Gemini, GPT-4o can “see, hear, and speak”
AI Agents	AI that can autonomously plan and execute multi-step tasks	Next evolution beyond chatbots — can browse, code, shop
Small Language Models (SLMs)	Efficient, smaller AI models that run on devices	Microsoft Phi, Google Gemma — AI on phones without internet
Edge AI	Running AI on local devices instead of cloud	Faster response, better privacy — used in IoT, autonomous cars
Quantum AI	Combining quantum computing with AI for exponential speed	Google’s quantum chip “Willow” — solving problems in minutes that would take billions of years
EU AI Act	World’s first comprehensive AI regulation law	Classifies AI systems by risk level; enforced from 2025
Agentic AI	AI that sets its own goals and works independently	Can manage workflows, make decisions with minimal human input
AI in Cybersecurity	AI detecting threats, preventing attacks, responding autonomously	Real-time threat detection, phishing email detection
Responsible AI / AI Governance	Frameworks for ethical, fair, and transparent AI	NITI Aayog guidelines, RBI model governance

Major AI Models & Companies

AI Model	Company	Key Feature
ChatGPT / GPT-4	OpenAI	Most popular GenAI chatbot; text, image, code generation
Gemini	Google DeepMind	Multimodal AI; integrated into Google Search, Android
Claude	Anthropic	Focus on safety and helpfulness; long-context processing
LLaMA	Meta	Open-source LLM for research and developers
Copilot	Microsoft	AI assistant integrated into Windows, Office, Edge
DALL-E / Sora	OpenAI	Image generation (DALL-E) and video generation (Sora)
Hanooman / BharatGPT	IIT Bombay Consortium	India’s multilingual LLM in 22+ Indian languages
DeepSeek	DeepSeek (China)	Open-source AI model competing with GPT-4

🛡️ 12. AI & Cybersecurity

AI and Cybersecurity are deeply connected. AI is used both to defend against cyber attacks and unfortunately, also to launch more sophisticated attacks. This dual nature makes it a critical topic for competitive exams.

🛡️ AI for Cyber Defence

Real-time threat detection
Automated malware analysis
Phishing email detection using NLP
User behaviour analytics (UBA)
Fraud detection in banking transactions
Network anomaly detection
Automated incident response

⚠️ AI for Cyber Attacks

Deepfake creation for fraud
AI-powered phishing emails (more convincing)
Automated vulnerability scanning
Password cracking using ML
Social engineering with AI chatbots
Evading security systems
Generating polymorphic malware

🎯

Exam Tip: Questions on Deepfakes, AI-based phishing, and AI in fraud detection are increasingly common in competitive exams. Know both the defensive and offensive uses of AI in cybersecurity.

📝 13. Previous Year Exam Questions on AI

Note: AI questions have appeared in IBPS IT Officer, SBI PO, UPSC Prelims (Science & Tech section), SSC CGL, and Railway exams. Below are question patterns based on previous exam trends.

IBPS IT Officer / SBI Specialist Officer

PYQ 1 IBPS IT Officer

The term “Artificial Intelligence” was first coined by:

a) Alan Turing
b) Charles Babbage
c) John McCarthy ✓
d) Tim Berners-Lee

Answer: c) John McCarthy He coined the term at the Dartmouth Conference in 1956 and is called the “Father of AI.”

PYQ 2 Banking Exam

Which of the following is NOT a type of Machine Learning?

a) Supervised Learning
b) Unsupervised Learning
c) Reinforcement Learning
d) Structured Learning ✓

Answer: d) Structured Learning The three main types of ML are Supervised, Unsupervised, and Reinforcement Learning.

PYQ 3 IBPS IT Officer

What is the primary goal of Robotics in AI?

a) To reduce internet speed
b) To enable machines to interact with the physical world ✓
c) To create websites
d) To store databases

Answer: b) To enable machines to interact with the physical world Robotics combines AI with mechanical systems for physical task execution.

UPSC / SSC Pattern Questions

PYQ 4 UPSC Pattern

Consider the following statements about Artificial Intelligence:
1. Narrow AI can perform any intellectual task like a human.
2. The Turing Test was proposed by Alan Turing in 1950.
3. Deep Learning is a subset of Machine Learning.

Which of the above statements is/are correct?

a) 1 and 2 only
b) 2 and 3 only ✓
c) 1 and 3 only
d) 1, 2 and 3

Answer: b) 2 and 3 only Statement 1 is wrong — Narrow AI is designed for specific tasks only. General AI (AGI) aims for human-level intelligence.

PYQ 5 UPSC/SSC Pattern

Which Indian initiative is associated with building AI compute infrastructure and promoting AI startups?

a) Digital India Programme
b) IndiaAI Mission ✓
c) Make in India
d) Startup India

Answer: b) IndiaAI Mission Approved in March 2024 with ₹10,372 crore budget. Has 7 pillars including compute capacity, startup financing, and safe AI.

📋 14. Practice MCQs — Chapter-wise (100 Questions)

Basics of Artificial Intelligence

Questions 1–20 • Foundation concepts, definitions & history

Artificial Intelligence (AI) mainly refers to:

a) Machines doing only arithmetic calculations
b) Machines that can think and act intelligently like humans to some extent ✓
c) Only robots moving physically
d) Only storing large amounts of data

Answer: b) AI enables machines to mimic human-like intelligence — learning, reasoning, and decision-making.

Who is known as the “Father of Artificial Intelligence”?

a) Alan Turing
b) Charles Babbage
c) John McCarthy ✓
d) Bill Gates

Answer: c) John McCarthy He coined the term “Artificial Intelligence” in 1956 at the Dartmouth Conference.

Which of the following is the BEST example of AI in daily life?

a) Manual ledger posting
b) Basic calculator performing addition
c) Google Assistant answering your questions ✓
d) Photocopy machine

Answer: c) Voice assistants like Google Assistant use NLP and ML to understand and respond to queries.

AI designed for a specific task only is called:

a) General AI (AGI)
b) Super AI (ASI)
c) Narrow AI (ANI) ✓
d) Strong AI

Answer: c) Narrow AI (ANI) All current AI — Siri, Alexa, ChatGPT — is Narrow AI. It excels at one task but cannot do others.

The Turing Test is used to determine:

a) A computer’s processing speed
b) Whether a machine can exhibit intelligent behaviour indistinguishable from a human ✓
c) The storage capacity of a hard disk
d) The battery life of a device

Answer: b) Proposed by Alan Turing in 1950. If a machine fools a human into thinking it’s human, it passes the test.

Which of the following is NOT a goal of Artificial Intelligence?

a) Learning from data
b) Decision making
c) Pattern recognition
d) Increasing manual paperwork ✓

Answer: d) AI aims to automate and reduce manual work, not increase it.

Machine Learning (ML) is a branch of AI that:

a) Requires all rules to be manually coded
b) Allows systems to learn from data and improve automatically ✓
c) Works only on mechanical devices
d) Can function without any data

Answer: b) ML models automatically learn patterns from past data without explicit programming.

The Dartmouth Conference of 1956 is significant because:

a) The internet was invented
b) The first computer was built
c) The term “Artificial Intelligence” was coined ✓
d) The first mobile phone was launched

Answer: c) John McCarthy organized this conference and introduced the term AI, marking the birth of AI as a field.

Which AI system defeated world chess champion Garry Kasparov in 1997?

a) AlphaGo
b) Watson
c) Deep Blue ✓
d) ChatGPT

Answer: c) Deep Blue IBM’s Deep Blue was a chess-playing AI that made history by defeating Kasparov.

Q10

The relationship between AI, ML, and Deep Learning is:

a) ML contains AI, DL is separate
b) AI is the broadest, ML is a subset of AI, DL is a subset of ML ✓
c) All three are the same thing
d) DL is the broadest concept

Answer: b) AI ⊃ ML ⊃ DL. This hierarchy is one of the most frequently tested facts in exams.

Q11

Natural Language Processing (NLP) is an AI technique used to:

a) Process chemical formulas
b) Understand, interpret, and generate human language ✓
c) Manage power backup
d) Design office furniture

Answer: b) NLP powers chatbots, translators, sentiment analysis, and voice assistants.

Q12

An AI system that can perform any intellectual task a human can do is called:

a) Narrow AI
b) Artificial General Intelligence (AGI) ✓
c) Reactive AI
d) Edge AI

Answer: b) AGI AGI is theoretical — no AI system has achieved human-level general intelligence yet.

Q13

Training an AI model means:

a) Paying salary to the model
b) Feeding data and adjusting model parameters to learn patterns ✓
c) Only changing hardware
d) Printing training reports

Answer: b) During training, models process data and adjust internal parameters to recognize patterns.

Q14

Google’s AlphaGo defeated the world champion of which board game?

a) Chess
b) Go ✓
c) Checkers
d) Scrabble

Answer: b) Go Google DeepMind’s AlphaGo beat Lee Sedol in 2016. Go is far more complex than chess.

Q15

The first chatbot in the history of AI was:

a) Siri
b) Alexa
c) ELIZA ✓
d) ChatGPT

Answer: c) ELIZA Created by Joseph Weizenbaum at MIT in 1966. It simulated a psychotherapist conversation.

Q16

The main requirement for AI/ML systems to function effectively is:

a) Large amount of quality data ✓
b) Manual registers
c) Physical office space
d) Printed documents

Answer: a) Data is the fuel for AI. The quality and quantity of data directly impacts model performance.

Q17

Computer Vision enables machines to:

a) Listen to music
b) Interpret and make decisions based on visual data like images and videos ✓
c) Store text files
d) Connect to the internet

Answer: b) Computer Vision is used in face recognition, self-driving cars, medical imaging, and quality inspection.

Q18

The period of reduced AI funding and interest is called:

a) AI Spring
b) AI Winter ✓
c) AI Revolution
d) AI Bubble

Answer: b) AI Winter The first AI Winter occurred during 1974–1980 due to unmet expectations and reduced funding.

Q19

Which company developed the AI system “Watson”?

a) Google
b) Microsoft
c) IBM ✓
d) Apple

Answer: c) IBM IBM Watson won the TV quiz show Jeopardy! in 2011, demonstrating advanced NLP capabilities.

Q20

Expert Systems in AI are designed to:

a) Play video games
b) Mimic the decision-making ability of a human expert in a specific field ✓
c) Store customer photographs
d) Manage electricity supply

Answer: b) Expert systems encode domain knowledge as rules and are used in medical diagnosis, financial advising, etc.

Advanced Concepts & Terminology

Questions 21–40 • ML types, neural networks, GenAI & modern concepts

Q21

Supervised Learning in AI requires:

a) No data labels at all
b) Only numeric data
c) Input data with correct output labels ✓
d) Only images

Answer: c) In supervised learning, the model learns from labelled examples — both input and expected output are provided.

Q22

Unsupervised Learning is mainly used for:

a) Predicting exact outcomes
b) Linear regression
c) Clustering and pattern discovery ✓
d) Payroll processing

Answer: c) It groups similar data without predefined labels — useful for customer segmentation.

Q23

A Neural Network is inspired by:

a) Human brain and nervous system ✓
b) Car engine
c) Solar panels
d) Filing cabinets

Answer: a) Neural networks are loosely modelled on how biological neurons are connected and process information.

Q24

Overfitting in AI models means:

a) Model performs well on training data but poorly on new data ✓
b) Model is too simple
c) Model uses insufficient data
d) Model is not trained at all

Answer: a) The model “memorizes” training data instead of learning general patterns — a key risk in ML.

Q25

Generative AI refers to AI systems that:

a) Only classify existing data
b) Create new content such as text, images, audio & video ✓
c) Store historical data
d) Manage hardware resources

Answer: b) GenAI models like ChatGPT, DALL-E, and Gemini generate entirely new content based on prompts.

Q26

LLM in AI stands for:

a) Large Language Model ✓
b) Loaded Learning Mechanism
c) Long Latency Module
d) Large Loop Memory

Answer: a) Large Language Model LLMs are trained on massive text data to understand and generate human-like language. Examples: GPT-4, Gemini.

Q27

The Transformer architecture, which powers modern LLMs, was introduced in the paper:

a) “Computing Machinery and Intelligence”
b) “Attention Is All You Need” ✓
c) “The Art of Computer Programming”
d) “A Mathematical Theory of Communication”

Answer: b) Published by Google researchers in 2017. This architecture uses the “attention mechanism” and is the foundation of ChatGPT, Gemini, etc.

Q28

Bias in AI models arises mainly due to:

a) High hardware speed
b) Unrepresentative or unfair training data ✓
c) Low internet speed
d) Manual authorization

Answer: b) If training data is biased (e.g., underrepresents certain groups), the AI model will produce biased outcomes.

Q29

Deep Learning is a subset of:

a) Cloud Computing
b) Machine Learning ✓
c) Blockchain
d) Cryptography

Answer: b) Machine Learning Deep Learning uses multi-layered neural networks to learn complex patterns from large datasets.

Q30

RLHF in AI training stands for:

a) Real Learning for Human Finance
b) Reinforcement Learning from Human Feedback ✓
c) Real-Life Hardware Function
d) Rapid Language High Framework

Answer: b) RLHF is used to fine-tune LLMs (like ChatGPT) using human evaluators’ feedback to improve response quality.

Q31

“Black box” in AI refers to:

a) A locked server room
b) A model whose internal decision logic is not easily interpretable ✓
c) ATM safe box
d) CCTV recording device

Answer: b) Many complex deep learning models make decisions that are difficult to explain — this is called the “black box” problem.

Q32

Model “drift” in AI means:

a) Staff transfer
b) Gradual decline in model performance over time ✓
c) Power fluctuation
d) Network shifting

Answer: b) As real-world data patterns change over time, a model trained on old data becomes less accurate — this is model drift.

Q33

Which AI technique is MOST suitable for sentiment analysis of customer feedback?

a) Robotics
b) NLP combined with Machine Learning ✓
c) Computer Vision
d) Optical storage

Answer: b) NLP processes the text to understand meaning, while ML classifies it as positive, negative, or neutral sentiment.

Q34

A model used to predict whether a transaction is fraud or not is an example of:

a) Clustering
b) Classification ✓
c) Sorting
d) Sampling

Answer: b) Classification Classification assigns items to categories (fraud / not fraud). It is a supervised learning technique.

Q35

Federated Learning allows:

a) Sharing all raw customer data openly
b) Training AI models on distributed data without centralizing it ✓
c) Only manual data collection
d) Offline-only training

Answer: b) Federated Learning enables privacy-preserving AI by keeping data on local devices and only sharing model updates.

Q36

“AI Hallucination” means:

a) AI having consciousness
b) AI generating confident but incorrect or made-up information ✓
c) AI running slowly
d) AI crashing due to error

Answer: b) LLMs sometimes fabricate facts, cite non-existent sources, or give wrong answers with high confidence — this is called hallucination.

Q37

Deepfake technology uses AI to:

a) Improve image quality only
b) Create realistic but fake videos, audio, or images of people ✓
c) Compress video files
d) Stream live video

Answer: b) Deepfakes use deep learning to swap faces or clone voices — a major cybersecurity and misinformation concern.

Q38

Edge AI refers to:

a) AI running only on cloud servers
b) Running AI on local devices like phones and IoT sensors ✓
c) AI on the edge of failure
d) AI used in border security only

Answer: b) Edge AI processes data locally on the device, providing faster response and better privacy without relying on cloud.

Q39

Multimodal AI can process:

a) Only text
b) Only images
c) Multiple types of data — text, images, audio, and video together ✓
d) Only numerical data

Answer: c) Models like GPT-4o and Google Gemini are multimodal — they can “see, hear, read, and speak.”

Q40

Quantum AI combines:

a) Classical physics and biology
b) Quantum computing and artificial intelligence ✓
c) Nuclear energy and computing
d) Traditional banking and IT

Answer: b) Quantum AI uses quantum computing’s massive parallel processing to solve AI problems exponentially faster.

AI in Banking, Finance & Government

Questions 41–65 • Applications, Indian banking AI, RBI guidelines, government initiatives

Q41

AI-based credit scoring models help banks to:

a) Decide branch location only
b) Assess borrower risk using multiple data variables ✓
c) Only print passbooks
d) Manage lockers

Answer: b) AI analyses income, spending patterns, repayment history, and many other variables for accurate credit assessment.

Q42

SBI’s AI-powered chatbot is called:

a) EVA
b) SIA (SBI Intelligent Assistant) ✓
c) iPal
d) Mitra

Answer: b) SIA SIA handles millions of customer queries. HDFC Bank’s chatbot is EVA, ICICI Bank’s is iPal.

Q43

AI-based fraud detection systems primarily look for:

a) Random transactions
b) Patterns and anomalies in transaction data ✓
c) Cheque size only
d) Branch code only

Answer: b) AI monitors transactions in real-time and flags unusual behaviour (sudden large transfers, unusual locations, etc.).

Q44

AI in AML (Anti-Money Laundering) helps to:

a) Increase manual file checks
b) Monitor large volumes of transactions and detect suspicious activity ✓
c) Print KYC forms
d) Approve all transactions blindly

Answer: b) AI automates AML monitoring, reducing false positives and catching complex money laundering patterns.

Q45

Robo-advisors in wealth management use AI to:

a) Allocate funds randomly
b) Suggest personalized investment portfolios based on customer profile ✓
c) Store gold in lockers
d) Decide office timings

Answer: b) Robo-advisors use algorithms to recommend asset allocation based on risk appetite, goals, and financial profile.

Q46

AI helps in KYC by:

a) Ignoring identity documents
b) Automating face matching and document verification ✓
c) Deleting customer data
d) Printing forms

Answer: b) AI uses Computer Vision and OCR to match faces, verify ID documents, and enable Video KYC for faster onboarding.

Q47

In internal audit, AI can:

a) Increase manual sampling only
b) Analyse 100% of transactions and highlight unusual cases ✓
c) Stop all reporting
d) Remove compliance

Answer: b) Unlike manual audit which samples a small percentage, AI can analyse every transaction for continuous auditing.

Q48

RBI’s primary focus regarding AI in banking is:

a) Marketing campaigns
b) Explainability, fairness, and transparency of AI decisions ✓
c) Building large offices
d) Extending branch timings

Answer: b) RBI requires banks to ensure their AI models are explainable, fair, and transparent — especially in credit decisions.

Q49

NITI Aayog’s National Strategy for AI (#AIForAll) identified how many focus sectors?

a) 3
b) 5 ✓
c) 7
d) 10

Answer: b) 5 sectors Healthcare, Agriculture, Education, Smart Cities, and Smart Mobility. Released in 2018.

Q50

The IndiaAI Mission was approved with a budget of:

a) ₹5,000 crore
b) ₹7,500 crore
c) ₹10,372 crore ✓
d) ₹15,000 crore

Answer: c) ₹10,372 crore Approved by the Union Cabinet in March 2024. It has 7 pillars including Compute Capacity and Startup Financing.

Q51

The National AI Portal of India (INDIAai) was launched by:

a) NITI Aayog alone
b) MeitY and NASSCOM ✓
c) RBI
d) DRDO

Answer: b) MeitY and NASSCOM Launched in 2020 as a knowledge hub for AI resources, news, and events. Website: indiaai.gov.in

Q52

SUPACE in the context of AI in India refers to:

a) A military defence system
b) Supreme Court Portal for Assistance in Court Efficiency ✓
c) A banking chatbot
d) A weather prediction system

Answer: b) Launched in 2021, SUPACE uses AI to assist Supreme Court judges with legal research and case analysis.

Q53

India’s own multilingual AI model supporting 22+ Indian languages is:

a) ChatGPT India
b) Hanooman / BharatGPT ✓
c) Google India AI
d) Microsoft Bharat

Answer: b) Hanooman / BharatGPT Developed by IIT Bombay consortium in 2024 to make AI accessible in Indian regional languages.

Q54

The Digital Personal Data Protection Act (DPDPA) was passed by India in:

a) 2020
b) 2021
c) 2023 ✓
d) 2025

Answer: c) 2023 This law governs how AI systems collect and process personal data. It established the Data Protection Board of India.

Q55

AI in Indian agriculture is used for:

a) Only record-keeping
b) Crop disease detection, weather prediction, and soil analysis ✓
c) Only irrigation
d) Only fertilizer purchase

Answer: b) AI helps farmers identify crop diseases from photos, predict weather, and optimize crop yields.

Q56

Which area is AI MOST used in capital markets?

a) Physical cheque clearing
b) Algorithmic trading and risk analysis ✓
c) Locker allotment
d) Passbook printing

Answer: b) AI powers high-frequency trading, portfolio optimization, and real-time risk modelling in capital markets.

Q57

AI-based personalization in digital banking apps means:

a) Same offers for all customers
b) Customized product offers based on user behaviour and profile ✓
c) Removing all offers
d) Offline-only products

Answer: b) AI analyses spending patterns, preferences, and demographics to recommend relevant products to each user.

Q58

In risk management, AI helps by:

a) Hiding risk from regulators
b) Providing early warning signals based on data patterns ✓
c) Removing risk reporting
d) Ignoring non-performing assets

Answer: b) AI analyses financial data to identify stressed accounts early, enabling proactive risk management.

Q59

The world’s first comprehensive AI regulation law is the:

a) GDPR
b) EU AI Act ✓
c) DPDPA
d) CCPA

Answer: b) EU AI Act Passed by the European Union, it classifies AI systems by risk levels and sets rules for high-risk AI applications.

Q60

AI-based chatbots primarily help banks by reducing:

a) Digital reach
b) Customer satisfaction
c) Response time and operational costs ✓
d) Number of ATMs

Answer: c) Chatbots handle routine queries 24×7, reducing the load on call centres and cutting operational costs.

Q61

Project Insight by the Income Tax Department uses AI for:

a) Detecting tax evasion and suspicious financial transactions ✓
b) Issuing PAN cards
c) Printing tax receipts
d) Managing office staff

Answer: a) Project Insight uses data analytics and AI to identify tax evaders by analysing financial data from multiple sources.

Q62

AI can help in loan collection by:

a) Ignoring overdue accounts
b) Predicting default risk and suggesting recovery strategies ✓
c) Removing all contact data
d) Closing branches

Answer: b) Predictive analytics helps identify which accounts are most likely to default and recommends optimal recovery approaches.

Q63

How many pillars does the IndiaAI Mission have?

a) 3
b) 5
c) 7 ✓
d) 10

Answer: c) 7 pillars Compute Capacity, Innovation Centre, Datasets Platform, Application Development, FutureSkills, Startup Financing, Safe & Trusted AI.

Q64

AI-driven regulatory reporting can:

a) Delay submission timelines
b) Reduce reporting errors and provide faster analytics ✓
c) Increase paperwork
d) Stop digitization

Answer: b) AI automates data extraction, validation, and report generation, making regulatory compliance faster and more accurate.

Q65

Banks using AI must ensure:

a) Only speed, no fairness
b) Accuracy, fairness, transparency, and customer protection ✓
c) Only profit maximization
d) Only marketing campaigns

Answer: b) These are core regulatory expectations from RBI for AI adoption in banking.

Ethics, Risks & Latest Trends

Questions 66–85 • Responsible AI, deepfakes, GenAI, cybersecurity & regulation

Q66

A key regulatory concern with AI in banking is:

a) Colour of banking devices
b) Explainability and fairness of AI decisions ✓
c) Number of chairs in offices
d) Building architecture

Answer: b) Regulators want AI models to be transparent and non-discriminatory, especially for credit and risk decisions.

Q67

Data privacy in AI mainly refers to:

a) Sharing customer data freely
b) Protecting personal data and using it lawfully ✓
c) Deleting all data
d) Printing data on notice boards

Answer: b) AI systems must comply with data protection laws like DPDPA (India) and GDPR (EU).

Q68

“Human-in-the-Loop” (HITL) in AI decision systems means:

a) Humans are completely removed from the process
b) Human oversight is involved in key decisions ✓
c) Only machines decide everything
d) Only manual systems are used

Answer: b) HITL ensures that critical decisions (loans, legal judgments) have human review, reducing risk of automated errors.

Q69

One ethical principle in AI use is:

a) Opacity (hiding how AI works)
b) Accountability ✓
c) Ignoring user consent
d) Unlimited data sharing

Answer: b) Accountability Organizations must be responsible for outcomes of their AI systems — this is a core principle of Responsible AI.

Q70

AI systems should be periodically reviewed because:

a) Staff changes require new systems
b) Data patterns and business conditions change over time ✓
c) Buildings get renovated
d) Branch codes change

Answer: b) Model drift occurs as real-world patterns change. Regular review ensures AI remains accurate and relevant.

Q71

Which of the following is a major risk of using AI in lending?

a) Faster processing
b) Hidden algorithmic bias against certain groups ✓
c) Better documentation
d) Improved customer service

Answer: b) If the training data contains historical biases, the AI model will replicate and amplify those biases in lending decisions.

Q72

Model governance for AI in banks includes:

a) Only hardware management
b) Policies, validation, monitoring, and documentation of AI models ✓
c) Locker key management
d) Uniform design decisions

Answer: b) Model governance ensures AI models are properly tested, documented, monitored, and comply with regulations.

Q73

The fastest-growing use case of Generative AI in banks is:

a) Manual cheque verification
b) Automated document summarization and report drafting ✓
c) ATM repair
d) Cash counting

Answer: b) GenAI automatically summarizes lengthy documents, generates audit reports, and drafts customer communications.

Q74

AI helps detect deepfakes by:

a) Increasing video resolution
b) Analysing inconsistencies in facial movements, lighting, and audio ✓
c) Blocking all video content
d) Slowing down internet speed

Answer: b) AI-based deepfake detection tools look for subtle artifacts like unnatural blinking, lighting mismatches, and audio sync issues.

Q75

Responsible AI focuses on:

a) Only revenue growth
b) Ethical, safe, and fair use of AI systems ✓
c) Maximum automation without any control
d) Full data visibility to the public

Answer: b) Responsible AI ensures AI is developed and deployed ethically with fairness, transparency, and accountability.

Q76

AI can be used in cybersecurity to:

a) Only encrypt passwords
b) Detect threats, phishing emails, and network anomalies in real-time ✓
c) Only install antivirus
d) Only block websites

Answer: b) AI analyses millions of data points to identify threats, detect phishing attempts, and respond to attacks automatically.

Q77

Which technology powers Generative AI models like ChatGPT?

a) Simple linear regression
b) Deep Learning using Transformer neural networks ✓
c) Punch card systems
d) Basic if-else programming

Answer: b) ChatGPT and similar models use the Transformer architecture with deep neural networks trained on massive datasets.

Q78

The biggest opportunity area for AI in Indian banking in 2025–26 is:

a) AI-based fraud monitoring and compliance automation ✓
b) Passbook printing
c) Building more manual counters
d) Offline customer enrolment

Answer: a) With increasing digital transactions (UPI), AI-powered fraud detection and regulatory compliance are the highest-priority areas.

Q79

Cognitive AI refers to:

a) Manual reasoning
b) Machines mimicking human thinking and decision reasoning ✓
c) Physical robotic arms
d) Only image scanning

Answer: b) Cognitive AI simulates human thought processes — understanding context, drawing inferences, and making judgments.

Q80

Overall, the impact of AI on banking can be summarized as:

a) Only job losses
b) More efficient, data-driven, and personalized banking with new risks to manage ✓
c) No change in services
d) Only increased paperwork

Answer: b) AI transforms banking with better services and efficiency, but also introduces new risks that require proper governance.

Expected & High-Probability Questions (2025–2026)

Questions 81–100 • Tricky, latest, and most-likely-to-appear questions

Q81 Expected 2026

Consider the following statements about the IndiaAI Mission:
1. It was approved in March 2024 with a budget of ₹10,372 crore.
2. It has 5 pillars.
3. It includes building 10,000+ GPU compute infrastructure.

Which of the above statements is/are correct?

a) 1 and 3 only ✓
b) 2 and 3 only
c) 1 and 2 only
d) 1, 2, and 3

Answer: a) 1 and 3 only Statement 2 is wrong — IndiaAI Mission has 7 pillars (not 5). NITI Aayog’s AI strategy had 5 focus sectors.

Q82 Expected 2026

What is the critical regulatory expectation for AI-based lending?

a) Data privacy and explainable decision-making ✓
b) Unlimited automation
c) Zero documentation
d) Removal of all human involvement

Answer: a) Regulators (especially RBI) require banks to explain why an AI model approved or rejected a loan application.

Q83 Expected 2026

Which new AI trend helps handle highly unstructured data in banking?

a) Rule-based systems
b) Foundation models and Large Language Models (LLMs) ✓
c) Tele-banking
d) Fax processing

Answer: b) LLMs can process unstructured data like emails, contracts, and handwritten notes — a game-changer for banking operations.

Q84 Expected 2026

AI will help RBI in supervision by:

a) Manual inspections only
b) Real-time monitoring of financial data and risk indicators ✓
c) Removing all financial oversight
d) Reducing all regulation

Answer: b) RBI is increasingly using SupTech (Supervisory Technology) — AI tools to monitor banks’ data in real-time.

Q85 Expected 2026

Which of the following correctly matches AI companies with their products?

a) Google — ChatGPT, Microsoft — Gemini
b) OpenAI — Gemini, Google — Claude
c) OpenAI — ChatGPT, Google — Gemini, Anthropic — Claude ✓
d) Meta — ChatGPT, Apple — Gemini

Answer: c) OpenAI made ChatGPT/GPT-4, Google DeepMind created Gemini, and Anthropic developed Claude.

Q86 Expected 2026

NITI Aayog’s Responsible AI framework emphasizes:

a) Speed only
b) Safety, equality, inclusivity, privacy, transparency, and accountability ✓
c) Profit maximization
d) Hardware upgrade

Answer: b) NITI Aayog released the Responsible AI approach papers in 2021 focusing on these principles for Indian context.

Q87 Expected 2026

“Agentic AI” refers to AI systems that can:

a) Only answer questions when asked
b) Set their own goals and work independently to complete multi-step tasks ✓
c) Only store information
d) Only translate languages

Answer: b) Agentic AI can plan, execute, and iterate on tasks autonomously — the next evolution beyond simple chatbots.

Q88 Expected 2026

Small Language Models (SLMs) are important because:

a) They are more expensive than large models
b) They can run on mobile devices without internet, ensuring faster and private AI ✓
c) They are less accurate but cheaper
d) They replace all cloud computing

Answer: b) SLMs like Microsoft Phi and Google Gemma bring AI to edge devices — phones, IoT — enabling offline, private AI use.

Q89 Expected 2026

Which Indian institution launched SUPACE for AI-assisted judicial work?

a) Supreme Court of India ✓
b) NITI Aayog
c) Law Ministry
d) High Court of Delhi

Answer: a) Supreme Court of India SUPACE was launched in 2021 by CJI SA Bobde to help judges with AI-powered legal research and case analysis.

Q90 Expected 2026

Consider the following statements:
1. ChatGPT was launched by Google in 2022.
2. The EU AI Act is the world’s first comprehensive AI regulation.
3. India’s DPDPA 2023 governs how AI processes personal data.

Which statements are correct?

a) 1 and 2 only
b) 2 and 3 only ✓
c) 1 and 3 only
d) All of the above

Answer: b) 2 and 3 only Statement 1 is wrong — ChatGPT was launched by OpenAI (not Google) in November 2022.

Q91 Expected 2026

AI-based phishing detection works by:

a) Blocking all emails
b) Using NLP to analyse email content, sender patterns, and URL structures ✓
c) Only checking file sizes
d) Scanning hardware

Answer: b) AI analyses language patterns, suspicious links, sender behaviour, and email headers to identify phishing attempts.

Q92 Expected 2026

Use of Generative AI in internal audit enables:

a) Ignoring irregularities
b) Automated summarization and exception highlighting ✓
c) Manual report writing
d) Removing compliance

Answer: b) GenAI can read thousands of audit documents, summarize findings, and highlight anomalies instantly.

Q93 Expected 2026

Model Explainability in AI is important because:

a) It helps customers and regulators understand why decisions were made ✓
b) It reduces automation
c) It restricts digital banking
d) It increases cash transactions

Answer: a) Especially important for lending — customers have the right to know why their loan was approved or rejected.

Q94 Expected 2026

GPT in ChatGPT stands for:

a) General Purpose Technology
b) Generative Pre-trained Transformer ✓
c) Global Processing Tool
d) Guided Protocol Transfer

Answer: b) Generative Pre-trained Transformer GPT models are pre-trained on massive text data and use the Transformer architecture to generate text.

Q95 Expected 2026

In customer analytics, AI helps banks to:

a) Offer generic services only
b) Personalize offers based on individual behaviour patterns ✓
c) Block digital channels
d) Remove mobile apps

Answer: b) AI analyses customer spending, saving, and browsing patterns to offer relevant products and services.

Q96 Expected 2026

An AI system used for credit approval without proper documentation can lead to:

a) Stronger compliance
b) Regulatory breaches and customer complaints ✓
c) Better manual control
d) Lower loan volumes

Answer: b) Regulators require AI models to be properly tested, documented, and validated before deployment.

Q97 Expected 2026

Which AI application helps Indian Railways with maintenance?

a) Manual checking of tracks
b) Predictive maintenance using sensor data and AI analytics ✓
c) Only painting coaches
d) Printing tickets

Answer: b) AI analyses sensor data from tracks and trains to predict equipment failures before they happen.

Q98 Expected 2026

The EU AI Act classifies AI systems based on:

a) Cost of development
b) Level of risk they pose to society ✓
c) Country of origin
d) Number of users

Answer: b) The EU AI Act categorizes AI into: Unacceptable Risk (banned), High Risk (regulated), Limited Risk, and Minimal Risk.

Q99 Expected 2026

Virtual assistants in banking can:

a) Replace ATM machines
b) Provide instant customer support and reduce service costs ✓
c) Replace database storage
d) Print cheques

Answer: b) AI-powered virtual assistants handle routine queries, account information, and transactions 24×7.

Q100 Expected 2026

India’s approach to AI regulation as of 2025 is:

a) Complete ban on AI
b) No guidelines at all
c) Sector-specific guidelines without a single comprehensive AI law ✓
d) Copying the EU AI Act entirely

Answer: c) India has sector-specific approaches (RBI for banking, MeitY for IT) rather than one comprehensive AI law like the EU AI Act.

⚡ 15. Quick Revision — One-Liners & Key Facts

🎯 Must-Remember One-Liners for Exams

Father of AI	John McCarthy (coined the term in 1956 at Dartmouth Conference)
Turing Test	Proposed by Alan Turing in 1950 to test machine intelligence
First Chatbot	ELIZA (1966) by Joseph Weizenbaum, MIT
AI ⊃ ML ⊃ DL	AI is broadest → ML is subset → Deep Learning is subset of ML
Narrow AI vs AGI	All current AI is Narrow (Weak). AGI (Strong) doesn’t exist yet.
Deep Blue	IBM chess AI that beat Kasparov in 1997
AlphaGo	Google DeepMind AI that beat Go champion in 2016
Transformer	Neural network architecture (2017) powering all modern LLMs
ChatGPT	OpenAI’s GenAI chatbot launched November 2022
GPT Full Form	Generative Pre-trained Transformer
LLM Full Form	Large Language Model
RLHF Full Form	Reinforcement Learning from Human Feedback
NLP Full Form	Natural Language Processing
IndiaAI Mission	₹10,372 crore, 7 pillars, approved March 2024
NITI Aayog AI Strategy	#AIForAll (2018), 5 focus sectors
INDIAai Portal	AI knowledge hub by MeitY + NASSCOM (2020)
SUPACE	Supreme Court AI portal for legal research (2021)
BharatGPT/Hanooman	India’s multilingual LLM, 22+ languages, IIT Bombay
DPDPA	Digital Personal Data Protection Act, 2023 (India)
EU AI Act	World’s first comprehensive AI law (risk-based classification)
SBI Chatbot	SIA (SBI Intelligent Assistant)
HDFC Chatbot	EVA (Electronic Virtual Assistant)
Overfitting	Model works on training data but fails on new data
Model Drift	AI accuracy declines as data patterns change over time
Black Box	AI model with unexplainable internal logic
AI Hallucination	AI generating confidently wrong information
Deepfake	AI-generated fake video/audio (cybersecurity threat)
HITL	Human-in-the-Loop — human oversight in AI decisions
Federated Learning	Training AI on distributed data without centralizing (privacy)
Edge AI	AI on local devices (phones, IoT) — fast, private

📌 Final Exam Checklist — Have You Revised?

✅ Father of AI → John McCarthy (1956)
✅ Turing Test → Alan Turing (1950)
✅ AI ⊃ ML ⊃ DL hierarchy
✅ 3 types of ML → Supervised, Unsupervised, Reinforcement
✅ Narrow AI vs AGI vs Super AI
✅ IndiaAI Mission → ₹10,372 crore, 7 pillars
✅ NITI Aayog → #AIForAll, 5 focus sectors
✅ Key banking chatbots → SIA (SBI), EVA (HDFC), iPal (ICICI)
✅ AI Ethics → Transparency, Fairness, Accountability, Privacy, Safety
✅ GenAI models → ChatGPT (OpenAI), Gemini (Google), Claude (Anthropic)
✅ India’s own LLM → BharatGPT / Hanooman (22+ languages)
✅ DPDPA 2023 → India’s data protection law
✅ EU AI Act → World’s first AI regulation law
✅ SUPACE → Supreme Court AI portal
✅ Deepfakes, Hallucination, Model Drift → Key risk concepts

📝 Disclaimer: This study material is prepared for educational and competitive exam preparation purposes. While every effort has been made to ensure accuracy, candidates are advised to verify facts from official sources (NITI Aayog, MeitY, RBI) for the most current information. AI is a rapidly evolving field — some facts and figures may change. Last updated: May 2026.