AI application distractors often blur automation with autonomy, classification with segmentation, encryption with ordinary data processing, and data bias with overfitting. Attempt all 12 MCQs before checking the explanations, then write the decisive rule behind each answer in one line. Reliable elimination matters more than isolated answer memorisation. Individual question links are not available for this set, so continue through the Skill Development Courses category or the AI for Placement learning module.
1. AI technologies and applications: six rules to separate first
Keep these distinctions beside you while solving. They connect search, data representation and system behaviour to the wider Coding and CS Fundamentals.
Concept | Decisive rule |
|---|---|
Automation and autonomy | Automation follows a fixed rule or schedule. Autonomy senses conditions and adapts. |
Recommendation | A recommendation system ranks items from user or item signals. |
Computer vision | Classification labels a whole image, detection localises objects, and segmentation labels pixels or regions. |
Encryption | Encryption transforms readable plaintext into protected ciphertext. |
A* search | A* evaluates |
Bias and overfitting | Data bias comes from unrepresentative data. Overfitting means fitting training detail so closely that the model fails to generalise. |
Consider a fixed traffic light that gives green for 30 seconds every cycle. It remains automated. An adaptive controller becomes autonomous when sensors count 40 vehicles and it changes green from 30 to 45 seconds.
Now take a 4 x 4 vision grid with 12 road pixels and 4 car pixels. Classification can assign one label, such as road scene, detection can draw one car box, and segmentation can assign a road or car label to every pixel.
2. AI application MCQs 1-3: autonomy, recommendations and heuristic search
Question 1
Which situation best represents an autonomous system rather than an automated system?
A. Washing machine running fixed program
B. Traffic light timer
C. Self-driving car adjusting speed automatically
D. Calculator performing addition
Correct answer: C. Self-driving car adjusting speed automatically.
The first, second and fourth options execute predefined instructions. A self-driving car reacts to sensor input: the timer stays at 30 seconds, while the adaptive controller sees 40 vehicles and changes to 45 seconds.
Question 2
Which AI application suggests products to users based on their previous behavior?
A. Recommendation system
सिफारिश प्रणाली
B. Operating system
ऑपरेटिंग सिस्टम
C. Compiler
कम्पाइलर
D. File system
फाइल सिस्टम
Correct answer: A. Recommendation system / सिफारिश प्रणाली.
Recommendation systems rank items from interaction signals. After a learner views 6 AI course pages, saves 2, and completes 1 introductory lesson, another AI lesson can rank above database content. The other systems process or store data but do not recommend.
Question 3
Which search algorithm uses heuristic function to estimate the cost to reach goal state?
A. Depth First Search
B. Breadth First Search
C. A* Search
D. Uniform Cost Search
Correct answer: C. A* Search.
A* uses f(n) = g(n) + h(n), where g is cost paid and h estimates cost remaining. Let S -> A cost 2, A -> G cost 7, S -> B cost 4, and B -> G cost 3: after S, A has g=2, h=5, f=7, while B has g=4, h=2, f=6. A* expands B and reaches G at 4 + 3 = 7; DFS and BFS follow structural order, while uniform-cost search uses g without h.
3. Computer vision MCQs 4-7: classification, self-driving systems and segmentation
Question 4
Which task in computer vision involves identifying objects in an image?
A. Image classification
छवि वर्गीकरण
B. Speech recognition
वाक् पहचान
C. Text summarization
पाठ सारांश
D. Data compression
डेटा संपीड़न
Correct answer: A. Image classification / छवि वर्गीकरण.
Image classification is the only computer-vision task among these choices. It may label the whole image car, while object detection locates cars with boxes. Speech recognition handles audio, summarisation handles language, and compression reduces representation size.
Question 5
Self-driving cars mainly use:
Self-Driving Cars
A. Computer Vision
कंप्यूटर विज़न
B. NLP
NLP
C. DBMS
DBMS
D. Operating Systems
ऑपरेटिंग सिस्टम
Correct answer: A. Computer Vision / कंप्यूटर विज़न.
Computer vision interprets camera frames to recognise lanes, vehicles, signals and pedestrians. A real self-driving stack may also use radar, lidar, localisation and planning, but computer vision is the relevant choice. NLP handles language, while DBMS and operating systems provide infrastructure, not visual perception.
Question 6
Computer Vision mainly deals with:
Computer Vision
A. Understanding images and videos
B. Managing databases
डेटाबेस प्रबंधन
C. Optimizing algorithms
एल्गोरिद्म का अनुकूलन
D. Managing computer networks
Correct answer: A. Understanding images and videos.
Computer vision extracts information from images and video frames. A 10-second video sampled at 30 frames per second contains 10 x 30 = 300 frames to analyse. Database management, algorithm optimisation and network management solve different problems.
Question 7
Which computer vision technique divides an image into multiple meaningful regions?
A. Image Classification
B. Object Recognition
C. Image Segmentation
D. Feature Matching
Correct answer: C. Image Segmentation.
Segmentation assigns region or pixel labels instead of one whole-image label. In the 4 x 4 grid, rows R R R R, R C C R, R C C R, R R R R contain 4 + 2 + 2 + 4 = 12 road pixels and 2 + 2 = 4 car pixels. Here R means road and C means car; object recognition identifies an object, while feature matching compares keypoints.
4. AI cybersecurity MCQs 8-11: encryption, access and system protection
Question 8
Which cybersecurity technique converts data into a secure format?
A. Encryption
एन्क्रिप्शन
B. Sorting
सॉर्टिंग
C. Searching
खोज
D. Parsing
पार्सिंग
Correct answer: A. Encryption / एन्क्रिप्शन.
Encryption transforms readable plaintext into ciphertext that requires the correct key to recover. Authenticated encryption takes the 14-character plaintext student_id=042, a key and unique nonce, then returns ciphertext plus an authentication tag; authorised decryption restores student_id=042. Sorting, searching and parsing do not provide confidentiality.
Question 9
Which of the following protects AI systems from malicious attacks?
A. Encryption
एन्क्रिप्शन
B. Sorting
सॉर्टिंग
C. Compression
संपीड़न
D. Compilation
कम्पाइलेशन
Correct answer: A. Encryption / एन्क्रिप्शन.
Encryption is the only security control here, protecting data in storage and transit. It cannot stop every attack, so AI systems also need authentication, authorisation, input validation and monitoring. Sorting, compression and compilation solve processing problems, not access control.
Question 10
Why is cybersecurity important in AI systems?
A. To protect data from unauthorized access
B. To increase processing speed
C. To reduce memory usage
D. To simplify algorithms
Correct answer: A. To protect data from unauthorized access.
Cybersecurity controls who can read or alter training data, models and outputs. For example, 100 students may read their own results, 5 authorised teachers may update them, and an unauthenticated visitor gets no access. Speed, memory use and simplicity are engineering goals, not cybersecurity.
Question 11
Cybersecurity in AI primarily aims to:
A. Improve model training speed
B. Protect AI systems and data from malicious attacks
C. Reduce model size
D. Eliminate bias in data
Correct answer: B. Protect AI systems and data from malicious attacks.
The aim is to protect an AI service, model and data from unauthorised use, tampering, theft or disruption. Training speed and model size concern efficiency, while data bias concerns representation and fairness. Security and fairness matter, but address different failure modes.
5. Responsible AI MCQ 12: recognise data bias without calling it overfitting
Question 12
A teacher uses an AI model trained only on urban students’ data. When applied to rural students, results are inaccurate. This problem is called:
A. Overfitting
B. Bias in data
C. Encryption error
D. Data compression
Correct answer: B. Bias in data.
The training data excludes rural students, so it is unrepresentative. Train on 1,000 urban and 0 rural records, then test on 100 urban and 100 rural: 92 / 100 = 92% urban accuracy and 61 / 100 = 61% rural accuracy, a 92 - 61 = 31 percentage-point gap. Overfitting can hurt generalisation, but under-representation is decisive here; encryption and compression are unrelated.
6. AI MCQ review: map every question to its decisive rule
Questions | Rule to write after checking |
|---|---|
1-3 |
|
4-7 | Label the visual task as classification, detection or segmentation. |
8-11 | Separate confidentiality and access control from speed or storage. |
12 | Compare representation across groups before blaming overfitting. |
For a missed definition, rewrite the six-row rule table. For a missed scenario, underline the input and required output. For the A* or bias examples, recompute f or the group accuracies without the choices. After one short revision interval, retry only missed questions.
For AI within a wider syllabus, use the UGC NET Computer Science hub. These questions support revision, but are not presented as UGC NET previous-year questions.
7. AI technologies and applications MCQs: the short version and next step
Autonomy adapts to sensed conditions.
Recommendation systems rank from signals.
Computer vision interprets images and videos.
Encryption and access control protect AI assets.
A* uses a heuristic, while responsible AI also requires representative data.
Retry Questions 3, 7 and 12 from the worked values alone. For structured learning in AI, machine learning, generative AI and practical applications, continue with Artificial Intelligence (AI).
Make a six-column sheet headed autonomy, recommendation, vision, security, search, bias. Put all 12 question numbers under one heading, write each decisive rule, and repeat only rows that took over 60 seconds.
