Đề Thi FE AIG202c - SP26 - FE - RE

[adminadmin is verified member.]

Học kỳ

SP2026

Thời Gian

3/5/26

Loại tài liệu

FE

AIG202c SP26 FE RE

Câu 1: Multiple Choice

When testing to determine if you are interacting with a human or a computer, which of the following would be a poor or ineffective question to ask?

• A. Do an impression of Simon Cowell
• B. Tell me who your best friend reminds you of
• C. Who would you get to play yourself in a movie and why?
• D. Can you play chess?

Câu 2: Multiple Choice

Which of the following is a principle often included in Al ethics guidelines to ensure transparency?

• A. Explainability
• B. Profitability
• C. Complexity
• D. Confidentiality

Câu 3: Multiple Choice

Which of the following do you need to create in order to publish a notebook to your GitHub repository?

• A. Apps
• B. Profile
• C. Access token
• D. Login credential

Câu 4: Multiple Choice

What is the primary function of Neural Networks in Al?

• A. Mimic human brain functionalities
• B. Identify objects in images
• C. Translate languages
• D. Forecast financial trends

Câu 5: Multiple Choice

Which of the following would be a BAD question to ask, if you were testing to see if you were talking to a human or a computer?

• A. Do an impression of Simon Cowell
• B. Tell me who your best friend reminds you of
• C. Who would you get to play yourself in a movie and why?
• D. Can you play chess?

Câu 6: Multiple Choice

Which of the following is not a stage in the Al Project Cycle?

• A. Problem Scoping
• B. Data Exploration
• C. Data Acquisition
• D. Prototyping

Câu 7: Multiple Choice

What is the main function of chatbots in customer service?

• A. To replace human employees entirely
• B. To simulate human conversation and interact with users
• C. To perform financial transactions securely
• D. To design websites automatically without user input

Câu 8: Multiple Choice

Which field of Al involves interpreting visual information from images or videos?

• A. Natural Language Processing (NLP)
• B. Robotics
• C. Computer Vision
• D. Machine Learning

Câu 9: Multiple Choice

What type of learning involves Al systems learning from data without explicit programming?

• A. Supervised learning
• B. Reinforcement learning
• C. Unsupervised learning
• D. Deep learning

Câu 10: Multiple Choice

Advances in the field of Computer Vision make which of the following possible:

• A. Real-time transcription
• B. Detecting cancerous moles in skin images
• C. Detecting fraudulent transactions
• D. On-demand online tutors

Câu 11: Multiple Choice

Many people question whether an Al can make ethical decisions. What do we mean by ETHICS?

• A. The laws of particular country
• B. The rules of behavior imposed by a programmer or the Al creator
• C. Human emotions
• D. A set of moral principles

Câu 12: Multiple Choice

Which of the following tasks is least likely to be performed by generative Al models like LLMs?

• A. Predicting stock market trends
• B. Creating human-like text
• C. Summarizing large documents
• D. Answering questions based on text input

Câu 13: Multiple Choice

Which of these questions would "Siri" or "Alexa" not be able to answer?

• A. What's the weather like outside?
• B. Is there any traffic on the way to work?
• C. What am I having for my tea tonight
• D. What day is Christmas Day on this year?

Câu 14: Multiple Choice

Who coined the term "artificial intelligence"?

• A. Alan Turing
• B. John McCarthy
• C. Marvin Minsky
• D. Geoffrey Hinton

Câu 15: Multiple Choice

What is the primary goal of Artificial Intelligence (AI)?

• A. To replace human intelligence entirely
• B. To simulate human intelligence in machines
• C. To limit the capabilities of machines
• D. To create autonomous robots

Câu 16: Multiple Choice

You are head of an organization that has realized that it is slowly losing expert knowledge as older workers retire. Which of the following techniques will NOT help mitigate the problem?

• A. Deploy an Al chatbot on mobile devices to make the technology portable.
• B. Implement an Al system that maintains a knowledge base.
• C. Use Al technology like Watson Assistant, Watson Discovery, and Watson Text to Speech on mobile devices to make the information easy to access.
• D. Implement a secure monitoring system to ensure workers leaving the company do not take sensitive industry information with them.

Câu 17: Multiple Choice

Generative Al models can [blank] the training data to create unique content.

• A. Differentiate between
• B. Draw conclusions from
• C. Learn from
• D. Identify patterns and classify

Câu 18: Multiple Choice

An airplane training system that functions without the help of a trainer, is an example of:

• A. Weak Intelligence
• B. Narrow Intelligence
• C. Broad Intelligence
• D. Generalized Intelligence

Câu 19: Multiple Choice

Maria is a scientific visualizer. How can the image-generation capabilities of generative Al help her in her daily work?

• A. Help her convert text into audio using text-to-speech
• B. Help her create comprehensive artwork from a simple sketch
• C. Help write new code snippets based on functionality
• D. Help her generate synthetic datasets for more diversity

Câu 20: Multiple Choice

Which of these are the first three steps customers should take to get started on their Al Journey?

• A. Prove, adopt and scale
• B. Prove, adopt and design
• C. Prove, scale and design
• D. Adopt, scale and design

Câu 21: Multiple Choice

How is Watson currently aiding in solving business problems based on the provided scenarios?

• A. Gathering expertise learned from thousands of employees to reduce time spent on tracking down expert knowledge.
• B. Using data from sensors to detect moisture levels in the air, helping irrigation systems give water where it's needed most.
• C. Identifying the best routes based on traffic, temperature, and road conditions to transport bees for pollination.
• D. Analyzing financial data to predict market trends and stock performance accurately.

Câu 22: Multiple Choice

Which statement is true about Adobe Firefly?

• A. It only works offline for privacy and data safety
• B. It is trained on Adobe Stock and public domain content
• C. It supports only English for input text prompts
• D. It requires no installation and runs only on mobile apps

Câu 23: Multiple Choice

Which of the following is a benefit of having an integrated Al system in a call center environment?

• A. Reducing costs
• B. Eliminating the need for human agents.
• C. Decreasing agent satisfaction due to fear of job loss
• D. Increasing agent turnover

Câu 24: Multiple Choice

What is one of the potential future impacts of Al in the job market?

• A. Decrease in demand for skilled professionals
• B. Increased job opportunities across all sectors
• C. Shift in job roles, with some jobs becoming obsolete
• D. Unchanged job market dynamics

Câu 25: Multiple Choice (Select 2 options)

After you uploaded your document, and it has finished processing. You click on the "Search" tag, what are the two search option in which you can extract useful information extracted from your document?

• A. Use the English Language
• B. Use natural language
• C. Use the Discovery Query Language
• D. Use unnatural language

Câu 26: Multiple Choice

In the financial sector, what role does Bloomberg GPT play?

• A. Detecting network security fraud from various text sources
• B. Automating interview scheduling from various text sources
• C. Analyzing market sentiment from various text sources
• D. Generating medical images for rare conditions from various text sources

Câu 27: Multiple Choice

Using a generative Al tool, Emily wants to create an image of a zebra-striped cat with a purple hat. Identify the best prompt for this task.

• A. List cat breeds that look like zebras or have stripes.
• B. Draw an image of a zebra-striped cat with a purple hat.
• C. What's the difference between a zebra and a cat?
• D. Find a zebra-striped cat that wears a purple hat

Câu 28: Multiple Choice

What type of generative Al capability does a large language model primarily exhibit?

• A. Image generation
• B. Audio generation
• C. Text generation
• D. [Not listed in text, but typically Video]

Câu 29: Multiple Choice

What do Abu Dhabi National Oil Company (ADNOC) hope that the use of Al will help them achieve?

• A. To understand their business better and retain knowledge.
• B. To eliminate knowledge transfer.
• C. To differentiate their business into educating people about geology.
• D. A production target of 17 million barrels per day.

Câu 30: Multiple Choice

What is the primary goal of Al Safety?

• A. Ensuring Al systems are completely risk-free
• B. Minimizing potential harm caused by Al systems
• C. Achieving perfect accuracy in Al decision-making
• D. Avoiding all errors in Al algorithms

Câu 31: Multiple Choice

How can prompt engineering help mitigate biases in generative Al models?

• A. By ensuring that prompts are designed to encourage diverse and inclusive responses
• B. By ignoring prompts that may lead to biased outputs
• C. By training generative Al models using biased datasets
• D. By generating prompts automatically using Al algorithms

Câu 32: Multiple Choice

What role does human oversight play in prompt engineering for generative Al?

• A. Human oversight is not necessary for prompt engineering
• B. Human oversight is essential for designing effective prompts and evaluating model outputs
• C. Human oversight is only required during the training phase of generative Al models
• D. Human oversight is necessary for adjusting hyperparameters during model generation

Câu 33: Multiple Choice

Which of the following is the main purpose of using the user feedback loop?

• A. To iteratively refine text prompts based on the response generated by the LLM
• B. To generate responses with examples
• C. To provide explicit instructions to generate neutral responses
• D. To generate meaningful responses without needing prior training on specific prompts

Câu 34: Multiple Choice

Jennifer wants to request some useful information about a complex medical condition using a large language model. Which among the following techniques should she employ on the text prompt to ensure that the generated content is relevant, accurate, and precise for this specialized field?

• A. Framing
• B. Domain expertise
• C. Bias mitigation
• D. Task specification

Câu 35: Multiple Choice

Why is prompt engineering important in the development of generative Al models?

• A. Prompt engineering ensures that generative Al models generate output that is relevant and coherent
• B. Prompt engineering helps reduce the computational resources required for training generative Al models
• C. Prompt engineering allows generative Al models to generate output without any human intervention
• D. Prompt engineering is not important for generative Al models

Câu 36: Multiple Choice

Monica is a writer looking to improve her storytelling by using a large language model (LLM) for creative writing prompts. Which feature of a prompt engineering tool would be most beneficial for her to achieve this goal?

• A. Iteratively refining prompts by adjusting them based on the initial outputs of the LLM's story
• B. Providing feedback for prompts based on a given input.
• C. Suggesting different prompt formats (like questions, dialogues, or scenarios) tailored to the preferred storytelling style
• D. Offering a library of pre-defined story beginnings designed for particular genres and themes

Câu 37: Multiple Choice

What happens in a fallback dialog?

• A. Bot says goodbye
• B. Bot repeats the greeting
• C. Bot says it didn't understand
• D. Bot opens browser

Câu 38: Multiple Choice

How does incorporating the persona pattern dimension enhance prompt effectiveness? Select the best option.

• A. By increasing the number of words in the prompt
• B. By introducing a sense of character-driven context and perspective
• C. By including a series of questions through multiple prompts
• D. By focusing solely on generic content without historical or fictional references

Câu 39: Multiple Choice

Which of the following is NOT a common prompt engineering technique?

• A. Adapting prompts based on user feedback
• B. Using pre-trained language models without any prompts
• C. Adjusting the temperature parameter during model generation
• D. Designing prompts to elicit specific types of responses

Câu 40: Multiple Choice

What is the purpose of evaluating prompt effectiveness in prompt engineering for generative Al?

• A. To determine the computational resources required for generating model outputs
• B. To assess the relevance and coherence of model outputs generated from different prompts
• C. To randomly select prompts for generating model outputs
• D. To fine-tune prompts based on feedback from model outputs

Câu 41: Multiple Choice

How can prompt engineering help improve the performance of Al models?

• A. By generating random prompts for training data
• B. By ensuring that prompts are designed to elicit relevant and coherent responses
• C. By ignoring prompts and allowing the model to generate output autonomously
• D. By training models using only prompts that produce desired responses

Câu 42: Multiple Choice

Why might a generic prompt result in an average output?

• A. Because the model needs technical instructions
• B. Because it triggers the model's error handling mode
• C. Because it lacks specific details and context
• D. Because prompts must be written in code only

Câu 43: Multiple Choice

Which among the following statements is accurate about the Tree-of-Thought approach?

• A. The Tree-of-Thought approach only works for generating responses to marketing-related prompts.
• B. The Tree-of-Thought approach is less effective than the Chain-of-Thought approach for generative Al reasoning.
• C. The Tree-of-Thought approach eliminates the need for any prompt instructions or constraints.
• D. The Tree-of-Thought approach enables generative Al models to explore multiple paths simultaneously and assess potential outcomes.

Câu 44: Multiple Choice

How can prompt engineering be used to mitigate biases in Al models?

• A. By generating prompts automatically using Al algorithms
• B. By ensuring that prompts are designed to encourage diverse and inclusive responses
• C. By ignoring prompts that may lead to biased outputs
• D. By training models using biased datasets

Câu 45: Multiple Choice

Imagine you are a content developer working with LLMs, and you must ensure that the responses generated are indiscriminatory and neutral. Which among the following techniques would you employ with your text prompts to instruct the model appropriately?

• A. Bias mitigation
• B. Few-shot prompting
• C. Zero-shot prompting
• D. Contextual guidance

Câu 46: Multiple Choice

What does Al Accountability entail in the context of ethical Al?

• A. The responsibility of Al developers and users for Al actions
• B. The secrecy of Al algorithms
• C. The avoidance of Al in decision-making processes
• D. The reliability of Al systems

Câu 47: Multiple Choice

Why is it important for chatbots to handle digressions effectively?

• A. Digressions help improve the user experience by providing additional information
• B. Digressions can lead to misunderstandings and dissatisfaction among users
• C. Digressions increase the complexity of the conversation and challenge the chatbot's capabilities
• D. Digressions are essential for training machine learning models

Câu 48: Multiple Choice

In chatbot development, what do 'intents' typically represent?

• A. User interface design principles
• B. The purpose or goal behind user inputs
• C. Database management strategies
• D. Networking protocols

Câu 49: Multiple Choice

Which of the following is NOT an example of an intent for a weather chatbot?

• A. GetForecast
• B. FindLocation
• C. GreetUser
• D. Provide Temperature

Câu 50: Multiple Choice

What are "entities" in the context of artificial intelligence and chatbots?

• A. The personalities assigned to chatbots
• B. The actions taken by chatbots to achieve user goals
• C. Specific pieces of information within user input that the chatbot needs to extract
• D. The language used by chatbots to communicate with users

Câu 51: Multiple Choice

We should try and translate this human concept of playing a game: "Winning and Losing' to a computer. The algorithm used in tic-tac-toe is:

• A. MAX
• B. MINI
• C. MAXMINI
• D. MINIMAX

Câu 52: Multiple Choice

What impact do well-implemented 'digressions' have on chatbot conversations?

• A. Limiting conversational options for users
• B. Enhancing natural conversation and user engagement
• C. Increasing response times
• D. Reducing chatbot functionality

Câu 53: Multiple Choice

Which of the following is NOT a common strategy for managing digressions in chatbots?

• A. Redirecting the conversation back to the original topic
• B. Ignoring the user's input and continuing with a predefined script
• C. Acknowledging the digression and exploring the new topic briefly
• D. Terminating the conversation abruptly

Câu 54: Multiple Choice

How are 'slots' used in the context of chatbot interactions?

• A. To identify network vulnerabilities
• B. To manage and capture specific pieces of information from users
• C. To optimize website speed
• D. To conduct financial transactions

Câu 55: Multiple Choice

What is the primary goal of 'digressions' in chatbot interactions, particularly concerning user experience?

• A. To streamline conversations by avoiding diversions
• B. To enforce strict conversational rules
• C. To introduce spontaneous and engaging conversation elements
• D. To restrict conversational freedom for users

Câu 56: Multiple Choice

How do chatbots use context variables and slots to maintain context during a conversation?

• A. By ignoring the user's input
• B. By storing information about the user's preferences and previous interactions
• C. By generating random responses
• D. By terminating the conversation after a set number of interactions

Câu 57: Multiple Choice

How do chatbots use dialog management to guide conversations with users?

• A. By ignoring user input
• B. By following a predefined sequence of interactions based on the user's responses
• C. By generating random responses
• D. By terminating the conversation after a set number of interactions

Câu 58: Multiple Choice

How are 'intents' utilized within Al-powered chatbots?

• A. Assessing hardware specifications
• B. Identifying the user's intentions or goals in conversations
• C. Enhancing graphic design elements
• D. Automating backend server configurations

Câu 59: Multiple Choice

How are "slots" used in natural language understanding for chatbots?

• A. Slots are used to store the user's input during a conversation
• B. Slots are used to represent specific pieces of information within user input that the chatbot needs to extract
• C. Slots are used to determine the overall purpose of the user's input
• D. Slots are used to classify the user's emotions during a conversation

Câu 60: Multiple Choice

How does "contextual understanding" enhance the effectiveness of a chatbot's dialog?

• A. Contextual understanding enables chatbots to speak multiple languages fluently.
• B. Contextual understanding allows chatbots to remember past interactions and use that information to guide current conversations
• C. Contextual understanding helps chatbots generate random responses to user queries.
• D. Contextual understanding enables chatbots to ignore user input that is irrelevant to the current conversation.