CHAPTER 6 REVIEW

Basic Applications of Artificial Intelligence in Learning and Career

Digital Competency in Learning and Career Development

University of Economics and Law – Vietnam National University, Ho Chi Minh City

5 Topics AI Fundamentals → Applications → Ethics → Trends

Chapter 6 covers five main topics:

• Topic 1: Basic Introduction to Artificial Intelligence (AI)
• Topic 2: Popular AI Tools to Support Learning
• Topic 3: Applications of AI in Work
• Topic 4: Ethics and Limitations of Using AI
• Topic 5: Trends in AI Development and Impact on Future Careers
 TABLE OF CONTENTS

1. Chapter Overview
2. Topic 1 – Basic Introduction to AI
• Definition
• AI techniques (Supervised, Unsupervised, Reinforcement)
• Predictive AI vs Generative AI
• Real-world applications
3. Topic 2 – Popular AI Tools for Learning
• EdTech AI landscape
• Smart research assistants
• Writing & editing tools
• Language learning tools
• Productivity tools
• Prompt engineering (CO-STAR, Zero/Few/Chain-of-Thought)
4. Topic 3 – Applications of AI in Work
• Automation vs Augmentation
• AI in healthcare & education
• AI in marketing & content creation
5. Topic 4 – Ethics and Limitations of AI
• Bias & fairness
• Privacy & security
• AI hallucination
• Current limitations
6. Topic 5 – Trends and Future Careers
• Emerging trends
• AI and the future of work
• Career paths & skills
7. Key Comparison Tables & Summary Charts
8. Practice Questions
• Single-answer MCQ
• Multiple-answer MCQ
• Matching
• Fill-in-the-blank
9. Answer Key with Explanations
10. Rapid Review & Exam Reminders
 1 CHAPTER OVERVIEW

Main Focus of Chapter 6

Chapter 6 introduces students to the fundamentals and practical applications of Artificial Intelligence (AI) in
academic learning and professional career development. It bridges theoretical AI concepts with actionable
skills students can use immediately — from choosing the right AI tool for research to understanding the
ethical responsibilities of AI use.

Why This Chapter Matters

• AI is no longer a niche technical topic — it is embedded in every field: business, healthcare, education,
law, and more.
• Students who understand how to use AI tools effectively hold a significant competitive advantage in the
modern labor market.
• Critical awareness of AI limitations and ethics is essential to avoid academic integrity violations and
real-world harms.
• Emerging AI trends are actively reshaping career paths, requiring all professionals to adapt regardless
of their discipline.

Connection to Digital Competency

Chapter 6 directly maps to Domain 6 of Vietnam's TT02/2025 Digital Competency Framework: AI
Application and Emerging Technologies. It also aligns with UNESCO's emphasis on AI Literacy as a
tool to bridge the digital divide. Specifically, this chapter addresses:
• Understanding what AI is and how it works (Basic AI Literacy)
• Using AI tools to increase productivity, support learning, and develop skills
• Evaluating the socio-economic impact of AI
• Applying AI responsibly with ethical awareness
 2 TOPIC 1 – BASIC INTRODUCTION TO ARTIFICIAL INTELLIGENCE

2.1 Definition of Artificial Intelligence

Official Course Definition

Artificial Intelligence (AI) is a technology that enables computers to simulate human learning,
comprehension, problem solving, decision making, and creativity.

This definition emphasizes three things:

• Simulation: AI imitates human cognitive functions — it does not truly "understand" or "feel."
• Range of functions: Learning, comprehension, problem solving, decision making, and creativity are all
named — this is broader than just "computing."
• Technology: AI is a technology tool, not an autonomous agent with independent goals.

EXAM TRAP: AI "simulates" human capabilities — it does NOT truly understand or feel. Do not confuse
simulation with genuine intelligence or consciousness.

2.2 AI Techniques: Three Learning Approaches

The slides describe three fundamental AI/Machine Learning techniques:

Supervised Learning

Training a machine from labeled data. The algorithm learns by being shown input-output pairs and
uses them to predict future outputs.

Context/Examples: Email spam filters (labeled "spam" or "not spam"), image recognition, fraud
detection.

Unsupervised Learning

Analyzes and clusters unlabeled data sets; discovering hidden patterns in data without the need for
human intervention.

Context/Examples: Customer segmentation, topic modeling, anomaly detection in large datasets.

Reinforcement Learning

Mimics the trial-and-error learning process that humans use to achieve its goals. An agent receives
rewards or penalties based on its actions.

Context/Examples: Game-playing AI (chess, Go), autonomous vehicle navigation, robotics.

Comparison of the Three Learning Types

Technique Data Required Key Characteristic

Supervised Learning Labeled data (input + correct Learns from examples; predicts
output) known categories

Unsupervised Learning Unlabeled data (no pre-set Discovers hidden patterns without
answers) human guidance
 Reinforcement Learning No dataset; learns through Trial-and-error goal-achievement
feedback (rewards/penalties) mimicking human learning

EXAM TRAP: Unsupervised learning works WITHOUT labeled data and WITHOUT human intervention —
do not confuse it with supervised learning, which requires labeled examples.

2.3 Machine Learning vs. Deep Learning

The slides distinguish Machine Learning and Deep Learning as related but distinct concepts:

Machine Learning (ML) Deep Learning (DL)

Broader field; algorithms that learn from data A subset of ML that uses neural networks with many
layers

Can work with smaller, structured datasets Requires large amounts of data and computing
power

More interpretable; results easier to explain Functions more like a "black box"; less interpretable

Examples: decision trees, regression, random Examples: image recognition, NLP, speech
forests synthesis

TIP: Deep Learning is a SUBSET of Machine Learning. Machine Learning is a SUBSET of Artificial
Intelligence. Think of concentric circles: AI > ML > DL.

2.4 Predictive AI vs. Generative AI (GenAI)

The course explicitly contrasts two major AI categories:

Predictive AI Generative AI (GenAI)

Goal: Forecast future outcomes based on historical Goal: Create new content based on learned
data patterns

Weather forecasting Text generation (ChatGPT)

Stock market prediction Image creation (Midjourney)

Netflix recommendations Code generation (GitHub Copilot)

Spam detection filters Voice synthesis

KEY POINT: This is a HIGH-PRIORITY comparison for exams. Remember: Predictive = forecast existing
trends. Generative = create NEW content.

2.5 Real-World AI Applications

The slides list the following real-world applications of AI:

Application Brief Description

Energy Consumption Optimizing power usage in buildings and grids

Content Creation Generating articles, images, videos automatically

Chatbot / Customer Service AI-powered conversation for support and queries

Sales Forecasting Predicting future sales from historical data

 Stock Price Forecasting Predicting financial market movements

Sentiment Analysis Determining emotional tone in text (e.g., reviews)

Translation Automatic language translation (e.g., DeepL, Google Translate)

Weather Forecasting Predicting atmospheric conditions using AI models

Anomaly Detection Identifying unusual patterns in data (e.g., fraud)

Key figures associated with AI development mentioned in the course:

• Prof. Geoffrey Hinton: Pioneer of deep learning and neural networks
• Prof. Yoshua Bengio: Deep learning pioneer; AI safety researcher
• Prof. Yann LeCun: Convolutional neural network (CNN) pioneer; Chief AI Scientist at Meta
 3 TOPIC 2 – POPULAR AI TOOLS TO SUPPORT LEARNING

3.1 The EdTech AI Landscape (2025)

AI has specialized into niche categories for students. The course identifies four major areas:

Category Purpose Key Note

Research Finding sources, analyzing papers Always verify citations from AI

Writing Grammar, structure, ideation Never copy-paste without verification

Study Aids Flashcards, tutoring, summarization Use active recall methods

Creativity Slides, images, videos AI generates; human judges quality

KEY POINT: "Don't rely on just one tool. Build a personal AI stack." — The course recommends combining
multiple AI tools for different tasks.

3.2 Smart Research Assistants

Tool Description & Best Use

Perplexity AI Combines ChatGPT-like conversation with live web search and citations. Great
for initial fact-finding.

Consensus An academic search engine. Asks questions and gets answers based ONLY on
peer-reviewed papers. Ideal for academic research.

ChatPDF Upload any PDF (textbook, paper) and chat with it. Example: "Summarize the
methodology in 3 bullets."

The slides list five best practices for using research AI tools:
• Be Particular: Clearly state what you need.
• Give Context: Provide background knowledge.
• Employ Plain Language: Use simple, clear language.
• Limit the Scope: Keep the topic relevant and focused; keep it narrow.
• Repeat and Make Improvements: Iterate and refine your prompts.

3.3 Writing & Editing Tools

Tool Specialty

Grammarly Beyond spellcheck: checks tone, clarity, and engagement.

QuillBot Excellent for paraphrasing and improving vocabulary variety.

Hemingway Editor Improves readability by highlighting overly complex sentences.

Best Practices for AI Writing Tools (from slides)

• Use AI to critique your draft, not write it.

• Ask: "What arguments am I missing?"

• Never copy-paste generated text without verification (Academic Integrity).

3.4 Language Learning & Translation Tools
Tool Key Feature

DeepL Superior translation accuracy compared to Google Translate, especially for

formal academic text.

ELSA Speak Uses AI to analyze pronunciation down to the phoneme level and provides
corrective feedback.

Duolingo Uses "Spaced Repetition Systems" (SRS) powered by AI to optimize

vocabulary retention.

3.5 Productivity & Organization Tools

Tool Function

Notion AI Summarize lecture notes, generate to-do lists from text, and organize
databases.

[Link] Records lectures/meetings and provides automated transcript + summary.

(Check permission first!)

[Link] Breaks down overwhelming tasks into small, manageable steps automatically.

NotebookLM (Google) Grounded AI that creates podcasts, FAQs, and study guides solely from your
uploaded documents.

Complete AI Tool Summary by Category

Task Category AI Tools Mentioned Notes

Research / Homework ChatGPT, Google Bard, Bing AI, Semantic Always verify citations
Scholar, [Link], Perplexity, Consensus

Knowledge Summary Notion AI, Obsidian AI, [Link], Whisper SpeechToText available

Language Learning Duolingo, ELSA Speak, Google Translate, SRS method in Duolingo
DeepL

Writing & Editing Grammarly, QuillBot, Claude AI, Never copy-paste

Hemingway Editor

Presentation Canva AI, Microsoft Designer, DALL-E, Image generation included

Midjourney

Time Management Motion, [Link], Todoist AI, Notion AI, Task scheduling
[Link]

Collaboration Google Meet AI, Zoom AI, Miro AI, Figma AI Auto-transcription

IMPORTANT: [Link] records lectures — always CHECK PERMISSION from lecturers/instructors before
using. Academic integrity and consent matter.

3.6 Prompt Engineering

"Garbage In, Garbage Out." The quality of AI output depends entirely on the quality of your prompt. The
course introduces two prompt frameworks:
The CO-STAR Framework
CO-STAR Prompt Framework

• C — Context: "I am a 3rd year marketing student..."

• O — Output: "Create a table..."

• S — Style: "Professional and concise..."

• T — Tone: "Academic..."

• A — Action: "Compare X and Y..."

• R — Restrictions: "No jargon, under 200 words."

Three Prompting Techniques

Technique Definition Example

Zero-Shot Asking without examples. "Write a poem about AI."

Few-Shot Providing examples to guide "Convert these 3 informal

style. sentences to formal. Here is an
example..."

Chain of Thought Asking AI to explain its reasoning "Solve this math problem and
step by step. explain step-by-step how you got
there."

KEY POINT: CO-STAR and the three prompting techniques (Zero-Shot, Few-Shot, Chain of Thought) are
HIGH-PRIORITY exam topics. Know the definitions and examples.
 4 TOPIC 3 – APPLICATIONS OF AI IN WORK

4.1 AI in the Modern Workplace

The course describes AI's role in the workplace with these key points:
• AI functions as a tool for automation and decision support.
• Widely integrated into business, healthcare, education, and creative industries.
• AI augments human skills rather than fully replacing them.
• Specific functions include: Robotic Process Automation (RPA) for repetitive tasks, predictive analytics
for demand forecasting and customer insights, chatbots for customer service, and anomaly detection
for fraud prevention.

4.2 Automation vs. Augmentation — The KEY Distinction

Automation Augmentation

AI replaces the human for a task entirely. AI empowers the human to do better work.

Data entry Doctor + AI diagnosis → better accuracy

Assembly line inspection Designer + AI brainstorming → more creativity

Basic translation Writer + AI editor → stronger prose

"AI won't replace you. A person using AI will."

This quote from the slides encapsulates the course's core philosophy. The threat is not AI itself — it is
the skill gap between those who master AI tools and those who do not.

EXAM TRAP: Do not confuse Automation with Augmentation. Automation = full replacement for a specific
task. Augmentation = AI assists/enhances the human. Most real-world AI is Augmentation.

4.3 AI in Healthcare
• Medical imaging: AI for early disease detection (e.g., cancer screening).
• Drug discovery: AI accelerates molecular analysis.
• Personalized medicine: Tailoring treatments based on genetic data.
• Administrative support: Automating patient records and scheduling.

4.4 AI in Education & Creativity

• Adaptive learning platforms: Personalized student learning paths.
• Grading and feedback: Automated essay scoring, plagiarism detection.
• Language tools: Real-time translation, writing assistants.
• Creative applications: Music generation, design support, storytelling.

4.5 AI in Marketing & Content Creation

Application Description Example Tools
Copywriting Generating ad copy, blog posts, Jasper, [Link]
and social media captions at
scale.

Visual Design Creating mockups, logos, and Midjourney, Canva Magic Studio
marketing assets instantly.

Personalization Tailoring email marketing HubSpot AI, Salesforce Einstein

campaigns to individual user
behavior.
 5 TOPIC 4 – ETHICS AND LIMITATIONS OF USING AI

5.1 Why Ethics in AI Matters

• AI systems influence human lives and decisions at scale.
• Risks of bias, discrimination, and inequality if unchecked.
• Importance of aligning AI with human values and rights.
• Ethical frameworks ensure trustworthy and responsible AI.

5.2 Bias and Fairness in AI

The course identifies three sources of AI bias:
• Biased data
• Biased algorithms
• Biased human assumptions

Examples of AI bias in practice:

• Facial recognition systems that are less accurate for minority groups.
• Biased hiring algorithms that systematically disadvantage certain demographics.

Solutions to AI bias:
• Diverse datasets
• Algorithmic auditing
• Fairness metrics

5.3 Privacy and Security Concerns

Three Dimensions of Privacy Risk (from slides)

• Personal Data: Free AI tools often use your chats to train their models. Do NOT upload sensitive
personal or financial information.

• Corporate Data: Companies block public ChatGPT to prevent intellectual property (IP) leakage.
Example: Samsung code leak incident.

• Regulation: GDPR (Europe) and emerging laws are trying to regulate how AI stores and processes
user data.

Other privacy & security concerns noted:

• Data-driven AI requires large personal datasets — creating surveillance risks.
• Regulatory responses include: GDPR, AI Act (EU), and emerging U.S. policies.

5.4 The AI Hallucination Problem

"Large Language Models are confident liars."

What is AI Hallucination?

AI generating plausible-sounding but completely incorrect information, citations, or facts.

 Why does it happen?

AI predicts the next probable word — it does NOT "know" facts. It prioritizes fluency over accuracy.

The Rule: Always verify AI claims with trusted primary sources.

KEY POINT: AI Hallucination is one of the MOST TESTED concepts in this chapter. Remember: AI sounds
confident but can be completely wrong. Always verify from primary sources.

5.5 Current Limitations of AI

Limitation Explanation

Narrow intelligence AI excels only in specific tasks; it cannot generalize the way humans do.

Lack of common sense AI cannot understand context and nuance the way humans naturally do.
reasoning

High resource demand Requires significant energy and data to train and run.

Ethical alignment challenge AI cannot fully grasp human values — it optimizes for measurable
objectives, not human well-being.

5.6 The "Collaborate, Not Depend" Principle

From the slides — the correct relationship with AI tools

• Double-check, triple-check AI-provided information.

• AI is only for support — not meant to be a replacement for mastering knowledge.

• Get into the habit of good prompting.

IMPORTANT: Academic Integrity Warning: Never submit AI-generated text as your own work without
verification and proper acknowledgment. AI writing without verification is a violation of academic integrity.
 6 TOPIC 5 – TRENDS IN AI DEVELOPMENT & IMPACT ON FUTURE
CAREERS

6.1 Emerging Trends in AI

The course identifies four key emerging AI trends:

Generative AI

Text, images, video, and music creation. AI that produces new content rather than merely analyzing
existing content.

Multimodal AI

Systems integrating text, audio, video, and sensor data. AI that can process and generate multiple
types of data simultaneously.

Agentic AI

Systems that can plan, reason, and act autonomously. AI that takes actions in the world, not just
generates responses.

AI Democratization

Low-code/no-code AI tools accessible to non-experts. Making AI usable by people without

programming knowledge.

6.2 AI and the Future of Work

• Job transformation rather than replacement — new roles emerge as routine tasks are automated.
• Routine tasks are automated, creating new human roles that require creativity, empathy, and judgment.
• Growth in demand for AI-related careers across all sectors.

6.3 Career Paths in AI

Career Track Roles Key Skills Required

Technical ML Engineer, Data Scientist, AI Programming, statistics,

Researcher mathematics

Interdisciplinary AI in Healthcare, Finance, Law, Domain expertise + AI literacy

Education

Ethics & Policy Regulation, Responsible AI Ethics, law, critical thinking, policy
Governance

6.4 Preparing for AI-Driven Futures

The course gives clear guidance for students preparing for an AI-shaped career:
• Emphasize lifelong learning and adaptability — the field changes constantly.
• Develop hybrid skills: technical knowledge combined with domain expertise.
• Strengthen soft skills: problem-solving, creativity, collaboration.
• Understanding AI is an advantage across ALL fields, not just technology.
6.5 Challenges and Opportunities
Challenges Opportunities

Job displacement in routine-task roles New industries and job categories created

Ethical dilemmas (bias, privacy, accountability) Enhanced productivity through human-AI

collaboration

Regulation gaps for emerging AI systems Human-AI collaboration for solving global issues
 7 KEY COMPARISON TABLES & SUMMARY CHARTS

Table 7.1 – Predictive AI vs. Generative AI (Quick Reference)

Predictive AI Generative AI (GenAI)

Goal: Forecast future outcomes from historical data Goal: Create new content from learned patterns

Weather forecasting Text generation (ChatGPT)

Stock market prediction Image creation (Midjourney)

Netflix recommendations Code generation (GitHub Copilot)

Spam detection filters Voice synthesis

Table 7.2 – AI Techniques: Three-Way Comparison

Feature Supervised Unsupervised

Data type Labeled (input + correct output) Unlabeled (no pre-set answers)

Human guidance Required (human labels the data) Not required (AI finds patterns
alone)

Output type Predicts known categories/values Discovers clusters or hidden

structures

Example use Spam detection, fraud detection Customer segmentation, topic

modeling

Feature Reinforcement Learning

Data type No dataset; environment interaction with reward/penalty

Human guidance Defines reward function; AI learns through trial and error

Output type Optimal action policy for goal achievement

Example use Game-playing AI, autonomous vehicles, robotics

Table 7.3 – CO-STAR Framework at a Glance

Letter Element

C Context — Who you are and your background

O Output — What format you want

S Style — Writing style requested

T Tone — Emotional register (academic, casual, etc.)

A Action — What you want AI to do

R Restrictions — Limits (word count, no jargon, etc.)

Table 7.4 – AI Limitations Quick Reference

 Limitation Exam-Relevant Detail

Narrow intelligence Excels only in SPECIFIC tasks; cannot generalize

No common sense Lacks contextual human reasoning

High resource demand Requires significant energy and data

Ethical alignment Cannot fully grasp human values

Hallucination Generates plausible but incorrect information; prioritizes fluency over

accuracy

Table 7.5 – Emerging AI Trends Quick Reference

Trend Core Idea

Generative AI Creates text, images, video, music

Multimodal AI Processes and integrates multiple data types simultaneously

Agentic AI Plans, reasons, and acts autonomously

AI Democratization Low-code/no-code tools make AI accessible to non-experts

 8 PRACTICE QUESTIONS

PART A – Single-Answer Multiple Choice

Choose the ONE best answer for each question.

Q1. According to the course, Artificial Intelligence (AI) is defined as a technology that enables
computers to simulate human:
A) Language only
B) Learning, comprehension, problem solving, decision making, and creativity
C) Physical movement and sensory perception
D) Internet browsing and data storage

Q2. Which AI learning technique analyzes and clusters unlabeled data sets to discover hidden
patterns WITHOUT human intervention?
A) Supervised Learning
B) Reinforcement Learning
C) Unsupervised Learning
D) Deep Learning

Q3. Which AI learning technique is described as mimicking the 'trial-and-error learning

process that humans use to achieve goals'?
A) Supervised Learning
B) Unsupervised Learning
C) Deep Learning
D) Reinforcement Learning

Q4. Supervised Learning trains a machine from which type of data?

A) Unlabeled data
B) Labeled data
C) Real-time sensor data
D) Randomly generated data

Q5. Which of the following is an example of Generative AI (GenAI)?

A) Netflix recommendation system
B) Spam detection filter
C) Stock market prediction model
D) Text generation using ChatGPT

Q6. Which of the following is an example of Predictive AI?

A) Image creation using Midjourney
B) Text generation using ChatGPT
C) Weather forecasting based on historical data
D) Voice synthesis

Q7. The AI academic search engine mentioned in the course that answers questions based
ONLY on peer-reviewed papers is:
A) Perplexity AI
B) Consensus
 C) ChatPDF
D) Semantic Scholar

Q8. In the CO-STAR prompt framework, the letter 'R' stands for:
A) Reasoning
B) Role
C) Restrictions
D) Result

Q9. Which prompting technique involves asking AI to explain its reasoning step by step?
A) Zero-Shot
B) Few-Shot
C) Chain of Thought
D) One-Shot

Q10. The quote 'AI won't replace you. A person using AI will.' relates to which concept?
A) AI Bias
B) Automation vs. Augmentation
C) AI Hallucination
D) AI Democratization

Q11. According to the course, AI hallucination occurs because AI:

A) Deliberately creates false information
B) Has too little training data
C) Prioritizes fluency over accuracy by predicting the next probable word
D) Cannot process text longer than a certain length

Q12. Which regulation is mentioned in the course as a response to AI privacy concerns in

Europe?
A) HIPAA
B) GDPR
C) CCPA
D) FERPA

Q13. Which of the following is NOT listed as a current limitation of AI according to the course?
A) Narrow intelligence
B) Lack of common sense reasoning
C) Too many correct answers too quickly
D) High resource demand

Q14. The course describes 'Agentic AI' as AI systems that can:

A) Generate images and music
B) Process only text data
C) Plan, reason, and act autonomously
D) Translate between multiple languages

Q15. The AI tool '[Link]' is described as:

A) An academic search engine for peer-reviewed papers
B) A tool that breaks down overwhelming tasks into small, manageable steps
C) An AI-powered pronunciation coach
 D) A plagiarism detection system

Q16. Which language learning tool uses 'Spaced Repetition Systems (SRS)' powered by AI?
A) DeepL
B) ELSA Speak
C) Duolingo
D) Google Translate

Q17. AI Democratization, as described in the course, refers to:

A) Governments controlling AI development
B) AI being available in all countries equally
C) Low-code/no-code AI tools accessible to non-experts
D) Making AI ethics guidelines public

Q18. Which of the following CORRECTLY describes 'Multimodal AI'?

A) AI that only works with text data
B) AI that integrates text, audio, video, and sensor data
C) AI that generates multiple languages simultaneously
D) AI that uses multiple algorithms in sequence

Q19. The Samsung code leak incident is mentioned in the course as an example of:
A) AI hallucination causing data errors
B) Corporate data leakage risk when using public AI tools like ChatGPT
C) An AI system producing biased output
D) A reinforcement learning failure in production

Q20. According to the course, which career track in AI involves roles such as 'regulation' and
'responsible AI governance'?
A) Technical careers
B) Interdisciplinary careers
C) Ethics and Policy careers
D) Creative AI careers

PART B – Multiple-Answer Questions

Choose ALL answers that apply. More than one answer may be correct.

Q21. Which of the following are listed as REAL-WORLD applications of AI in the course?
(Select ALL that apply)
A) Sentiment Analysis
B) Personalized Music Composition
C) Weather Forecasting
D) Anomaly Detection
E) Stock Price Forecasting
F) Language Translation

Q22. Which of the following AI tools are categorized as 'Productivity & Organization' tools in
the course? (Select ALL that apply)
A) Notion AI
B) Grammarly
 C) [Link]
D) [Link]
E) NotebookLM (Google)
F) DeepL

Q23. Which of the following correctly describe AI 'Augmentation' (as opposed to Automation)
in the workplace? (Select ALL that apply)
A) Doctor + AI diagnosis = better accuracy
B) Automated data entry replacing human typists
C) Designer + AI brainstorming = more creative output
D) Assembly line inspection done entirely by robots
E) Writer + AI editor = stronger prose
F) Basic translation fully automated by AI

Q24. Which of the following are listed as sources of AI Bias according to the course? (Select
ALL that apply)
A) Biased data
B) Inadequate computing power
C) Biased algorithms
D) Biased human assumptions
E) Low internet speed
F) Insufficient user testing

Q25. Which of the following are identified as EMERGING AI TRENDS in the course? (Select
ALL that apply)
A) Generative AI
B) Narrow AI specialization
C) Multimodal AI
D) Agentic AI
E) AI Democratization
F) Automated Assembly

PART C – Matching Questions

Match each item in Column A with the correct item in Column B.

Q26. Match each AI tool in Column A with its PRIMARY function in Column B.
Column A Column B

1. Perplexity AI A. Breaks overwhelming tasks into small

manageable steps

2. Consensus B. AI pronunciation coach that analyzes down to the

phoneme level

3. ChatPDF C. Chat with an uploaded PDF document

4. ELSA Speak D. Academic search engine answering only from

peer-reviewed papers

5. [Link] E. Combines conversational AI with live web search

and citations
Your answers: 1-___, 2-___, 3-___, 4-___, 5-___

Q27. Match each AI concept in Column A with its correct definition in Column B.
Column A Column B

1. Supervised Learning A. AI that forecasts future outcomes based on

historical data

2. Unsupervised Learning B. Training from labeled data to predict known

categories

3. Reinforcement Learning C. AI that creates new content (text, images, code)

from learned patterns

4. Generative AI D. Mimics trial-and-error learning to achieve goals

via rewards/penalties

5. Predictive AI E. Discovers hidden patterns in unlabeled data

without human intervention
Your answers: 1-___, 2-___, 3-___, 4-___, 5-___

Q28. Match each CO-STAR framework letter in Column A with its meaning in Column B.
Column A Column B

1. C (Context) A. The desired emotional register (e.g., academic,

professional)

2. O (Output) B. Limits imposed on the response (e.g., under 200

words, no jargon)

3. S (Style) C. Who you are and your background information

4. T (Tone) D. The writing style you want the AI to adopt

5. R (Restrictions) E. The format you want the AI to produce

Your answers: 1-___, 2-___, 3-___, 4-___, 5-___

PART D – Fill-in-the-Blank
Fill in the blank with the exact word or short phrase from the course materials.

Q29. The three main AI learning techniques described in the course are: Supervised Learning,
Unsupervised Learning, and ______ Learning.
Answer: _______________________________________

Q30. "Large Language Models are ______ liars." — This quote from the course illustrates the
AI hallucination problem.
Answer: _______________________________________

Q31. The CO-STAR prompt framework stands for: Context, Output, Style, Tone, Action, and
______.
Answer: _______________________________________

Q32. According to the course, the quality of AI output depends entirely on the quality of your
______. "Garbage In, Garbage Out."
Answer: _______________________________________

Q33. The AI trend where systems can plan, reason, and act autonomously is called ______ AI.
Answer: _______________________________________
 9 ANSWER KEY WITH EXPLANATIONS

PART A – Single-Answer MCQ

Q1: B — AI is defined as simulating human LEARNING, COMPREHENSION, PROBLEM SOLVING, DECISION

MAKING, and CREATIVITY — all five components are in the definition.
Q2: C — Unsupervised Learning = unlabeled data + no human intervention. A (Supervised) = labeled data. B
(Reinforcement) = trial-and-error rewards.
Q3: D — Reinforcement Learning 'mimics the trial-and-error learning process' — this specific phrase is from the
slides.
Q4: B — Supervised = LABELED data (input + correct output pairs).
Q5: D — ChatGPT text generation = Generative AI (creates NEW content). A, B, C = Predictive AI (forecasts
from historical data).
Q6: C — Weather forecasting from historical data = Predictive AI (forecasts outcomes). D (voice synthesis) =
Generative AI.
Q7: B — Consensus answers questions based ONLY on peer-reviewed papers. Perplexity uses live web
search. ChatPDF = upload a PDF.
Q8: C — CO-STAR: C=Context, O=Output, S=Style, T=Tone, A=Action, R=RESTRICTIONS.
Q9: C — Chain of Thought = asking AI to explain step-by-step reasoning. Zero-Shot = no examples. Few-Shot =
examples provided.
Q10: B — 'AI won't replace you; a person using AI will' = Automation vs. Augmentation concept.
Q11: C — Hallucination = AI predicts next probable word, prioritizing FLUENCY OVER ACCURACY. It does not
'know' facts.
Q12: B — GDPR (General Data Protection Regulation) = European AI/data privacy regulation. HIPAA = US
healthcare. CCPA = California.
Q13: C — C is NOT a listed limitation. The 4 limitations are: Narrow intelligence, No common sense, High
resource demand, Ethical alignment challenge.
Q14: C — Agentic AI = 'systems that can plan, reason, and act autonomously.' Generative = creates content.
Q15: B — [Link] = breaks down overwhelming tasks into small, manageable steps.
Q16: C — Duolingo uses SRS (Spaced Repetition Systems). DeepL = translation. ELSA = pronunciation.
Q17: C — AI Democratization = low-code/no-code AI tools accessible to non-experts.
Q18: B — Multimodal AI = integrates TEXT, AUDIO, VIDEO, and SENSOR DATA simultaneously.
Q19: B — Samsung code leak = example of CORPORATE DATA leakage when using public AI tools.
Q20: C — Ethics and Policy careers = regulation, responsible AI governance. Technical = ML engineer, data
scientist.

PART B – Multiple-Answer MCQ

Q21: Correct = A, C, D, E, F
B (Personalized Music Composition) is NOT in the slide's list of applications. The listed applications are:
Energy Consumption, Content Creation, Chatbot/Customer Service, Sales Forecasting, Stock Price
Forecasting, Sentiment Analysis, Translation, Weather Forecasting, Anomaly Detection.

Q22: Correct = A, C, D, E
 B (Grammarly) = Writing & Editing. F (DeepL) = Language Learning/Translation. Productivity &
Organization tools: Notion AI, [Link], [Link], NotebookLM.

Q23: Correct = A, C, E
B, D, F are examples of AUTOMATION (full replacement). A, C, E are AUGMENTATION (AI enhances
human performance): Doctor+AI, Designer+AI, Writer+AI.

Q24: Correct = A, C, D
The course lists THREE bias sources: biased data, biased algorithms, and biased human assumptions.
B, E, F are not listed.

Q25: Correct = A, C, D, E
The 4 emerging trends: Generative AI, Multimodal AI, Agentic AI, AI Democratization. B (Narrow AI) is a
current limitation, not a trend. F (Automated Assembly) is not listed.

PART C – Matching Answers

Q26: 1-E, 2-D, 3-C, 4-B, 5-A

Perplexity=live web search (E); Consensus=peer-reviewed only (D); ChatPDF=upload & chat (C);
ELSA=phoneme pronunciation (B); Goblin=task breakdown (A).

Q27: 1-B, 2-E, 3-D, 4-C, 5-A

Supervised=labeled data (B); Unsupervised=hidden patterns no labels (E); Reinforcement=trial-and-error
rewards (D); Generative=creates content (C); Predictive=forecasts outcomes (A).

Q28: 1-C, 2-E, 3-D, 4-A, 5-B

Context=background (C); Output=format (E); Style=writing style (D); Tone=emotional register (A);
Restrictions=limits (B).

PART D – Fill-in-the-Blank Answers

Q29: "Reinforcement" — The three techniques: Supervised, Unsupervised, and REINFORCEMENT Learning.
Q30: "confident" — Exact quote from slides: 'Large Language Models are confident liars.' — highlights the
hallucination problem.
Q31: "Restrictions" — CO-STAR = Context, Output, Style, Tone, Action, RESTRICTIONS.
Q32: "prompt" — 'Garbage In, Garbage Out' = prompt quality determines output quality.
Q33: "Agentic" — Agentic AI = plans, reasons, and acts AUTONOMOUSLY. Not to be confused with
Generative (creates content) or Multimodal (multiple data types).
 10 RAPID REVIEW & EXAM REMINDERS

10.1 The Most Important Chapter 6 Points

Must-Know Definitions

• AI: Technology enabling computers to simulate human learning, comprehension, problem solving,
decision making, and creativity.

• Supervised Learning: Trains from LABELED data.

• Unsupervised Learning: Discovers hidden patterns in UNLABELED data WITHOUT human

intervention.

• Reinforcement Learning: Trial-and-error learning via rewards and penalties.

• Predictive AI: Forecasts future outcomes from historical data.

• Generative AI (GenAI): Creates NEW content (text, images, code, voice).

• AI Hallucination: AI generates plausible-sounding but factually incorrect information by prioritizing

fluency over accuracy.

• Agentic AI: Can plan, reason, and act autonomously.

• AI Democratization: Low-code/no-code tools make AI accessible to non-experts.

• CO-STAR: Context, Output, Style, Tone, Action, Restrictions.

10.2 Key Terms to Memorize

Term One-Line Definition

Machine Learning Field of AI where systems learn from data

Deep Learning Subset of ML using multi-layer neural networks

Supervised Learning Learns from labeled data (input + correct output)

Unsupervised Learning Finds patterns in unlabeled data; no human guidance

Reinforcement Learning Trial-and-error via rewards/penalties

Generative AI Creates new content: text, images, code, audio

Predictive AI Forecasts outcomes from historical data

Hallucination AI confidently generating incorrect information

CO-STAR Prompt framework: Context, Output, Style, Tone, Action, Restrictions

Zero-Shot Prompting without examples

Few-Shot Prompting with examples to guide style

Chain of Thought Asking AI to explain reasoning step by step

Augmentation AI enhances human capability; human remains in control

Automation AI fully replaces human for a specific task

 Agentic AI AI that plans, reasons, and acts autonomously

Multimodal AI Processes/generates text, audio, video, sensors

AI Democratization No-code/low-code tools for non-technical users

SRS (Spaced Repetition) AI-optimized learning method; used by Duolingo

GDPR European data protection law affecting AI use

Narrow Intelligence AI excels only in specific tasks; cannot generalize

10.3 Common Mistakes Students Make

EXAM TRAP: Confusing Supervised and Unsupervised: Supervised = LABELED data (human-guided).
Unsupervised = UNLABELED data (no human needed).

EXAM TRAP: Mixing Predictive and Generative AI: Predictive = forecasts from existing data. Generative
= creates NEW content.

EXAM TRAP: Thinking AI "knows" facts: AI does NOT know facts — it predicts probable next words.
This is why hallucination occurs.

EXAM TRAP: Confusing Augmentation and Automation: Augmentation = AI assists human. Automation
= AI replaces human for the task.

EXAM TRAP: Mixing up ML and Deep Learning: Deep Learning is a SUBSET of Machine Learning. ML is
a SUBSET of AI.

EXAM TRAP: Listing ChatGPT as Predictive AI: ChatGPT is GENERATIVE AI. It CREATES text, it does
not forecast existing trends.

EXAM TRAP: Confusing CO-STAR with CREATE: CO-STAR (Chapter 6): Context, Output, Style, Tone,
Action, Restrictions. Different framework than CREATE from Chapter 3.

EXAM TRAP: Saying AI completely replaces human jobs: The course says job TRANSFORMATION,
not replacement. New roles emerge as routine tasks are automated.

10.4 Short Last-Minute Revision Reminders

✓ AI = simulates human learning, comprehension, problem solving, decision making, creativity.
✓ Three AI techniques: Supervised (labeled), Unsupervised (unlabeled, no human), Reinforcement
(trial-and-error).
✓ Predictive AI = forecast | Generative AI = create new content.
✓ AI Hallucination = fluency over accuracy; always verify with primary sources.
✓ Limitations of AI: Narrow intelligence, no common sense, high resource demand, ethical alignment
challenge.
✓ CO-STAR = Context, Output, Style, Tone, Action, Restrictions.
✓ Prompting techniques: Zero-Shot (no examples) | Few-Shot (with examples) | Chain of Thought
(step-by-step reasoning).
✓ "Garbage In, Garbage Out" — prompt quality determines AI output quality.
✓ "AI won't replace you. A person using AI will." — Automation vs. Augmentation.
✓ Four emerging trends: Generative AI | Multimodal AI | Agentic AI | AI Democratization.
✓ Career tracks in AI: Technical | Interdisciplinary | Ethics & Policy.
✓ Hybrid skills needed: technical + domain expertise + soft skills.
✓ GDPR (Europe) + AI Act (EU) = key regulations for AI privacy.
✓ Samsung leak incident = why companies block public AI tools (corporate data risk).
✓ Best practice: Collaborate with AI, do not depend on it. Double-check everything.

Good luck on your exam!

University of Economics and Law — Vietnam National University, Ho Chi Minh City