Course: Understanding Artificial Intelligence
Artificial Intelligence (AI) is the study and engineering of systems that perform tasks normally requiring human intelligence. In Chapter 1 we introduce AI’s core goals: problem-solving, perception, language understanding, and decision-making. The chapter distinguishes Narrow AI, which excels at single tasks, from the still-theoretical General AI that would match human versatility. We explain how AI systems combine data, algorithms, and computing power to learn patterns and produce predictions or actions. Machine learning and its deeper branch, deep learning, are presented as the principal methods that enable modern AI. The composition emphasizes real-world impacts: AI improves diagnostics in healthcare, optimizes logistics in transportation, detects fraud in finance, and personalizes education. Ethical questions appear throughout the chapter, including concerns about privacy, bias, and workforce change. Readers are invited to consider both the transformative potential and the limitations of AI systems. Historical context is summarized to show how ideas from the mid-20th century evolved into today’s data-driven methods. Practical examples—voice assistants, recommendation engines, and image classifiers—anchor theory to everyday experience. The chapter also introduces essential vocabulary and the basic AI workflow from data collection to deployment. Finally, it frames AI as a human-centered technology: its design, governance, and use must reflect social values and public accountability.Lesson: Chapter 1 — What is Artificial Intelligence (AI)?
Topic: Fundamental concepts, types, mechanisms, applications, and societal implications of AI
Learning Outcomes
- Define Artificial Intelligence and distinguish Narrow AI from General AI.
- Explain basic AI components: data, algorithms, computational resources.
- Summarize key AI techniques: machine learning and deep learning.
- Identify major application domains and primary societal/ethical concerns.
- Interpret a simple AI workflow from data collection to deployment.
Answer
After this lesson students will be able to state definitions, provide examples of Narrow AI, describe the role of data and algorithms, and discuss the ethical trade-offs in deploying AI systems.
Conceptual Explanation
| Concept | Short Explanation | Example |
|---|---|---|
| Artificial Intelligence | Engineering systems to perform tasks that normally require human cognitive abilities. | Automated diagnostic assistance in radiology |
| Narrow (Weak) AI | Systems specialized for a single or narrow set of tasks. | Voice assistant (Siri, Alexa) |
| General (Strong) AI | Theoretical systems with human-level general intelligence across domains. | Not yet realized — ongoing research |
| Machine Learning | Algorithms learn patterns from data to make predictions or decisions. | Recommendation engines (Netflix) |
| Deep Learning | Neural networks with many layers for modeling complex patterns. | Image recognition with convolutional neural networks |
| AI Workflow | Collect → Preprocess → Train → Validate → Deploy → Monitor → Retrain | Medical imaging pipeline (data to deployment) |
| Ethical Concerns | Privacy, bias, transparency, accountability, and displacement risks. | Algorithmic bias in lending decisions |
Answer
The table summarizes the foundational concepts and gives concrete examples for classroom discussion and follow-up tasks.
Bilingual (EN → বাংলা)
| English | বাংলা (Bangla) |
|---|---|
| Artificial Intelligence | কৃত্রিম বুদ্ধিমত্তা |
| Narrow AI | সীমিত কৃত্রিম বুদ্ধিমত্তা |
| General AI | সাধারণ কৃত্রিম বুদ্ধিমত্তা |
| Machine Learning | যন্ত্র শিখন |
| Deep Learning | ডীপ লার্নিং / গভীর শিখন |
| Data Collection | ডেটা সংগ্রহ |
| Bias | পক্ষপাত / পূর্বাগ্রহ |
Answer
Use these paired terms when explaining concepts to bilingual learners or preparing translated materials.
Important Notes
- High-quality data is as important as the algorithm; garbage-in leads to garbage-out.
- Narrow AI systems can be excellent at one task but may fail badly outside their scope.
- General AI remains hypothetical; avoid conflating current capabilities with science fiction.
- Explainability and monitoring are essential for safe, trustworthy deployments.
- Ethical review and stakeholder engagement are required for sensitive applications (e.g., healthcare, criminal justice).
Answer
Keep these notes visible when designing class activities or real-world assignments to promote responsible practice.
Exercises
True / False
- AI systems can always explain their own decisions.
Answer
False — many AI models (especially deep networks) are not inherently interpretable; explainability techniques are required.
- Narrow AI includes voice assistants and image classifiers.
Answer
True — these are task-specific systems.
- General AI currently exists and is widely deployed.
Answer
False — General AI is theoretical and not yet achieved.
- High-quality data reduces model bias completely.
Answer
False — good data reduces bias risk but does not eliminate structural or historical biases entirely.
- Machine learning often requires iterative validation and retraining.
Answer
True — models must be validated and updated as data distributions change.
Fill in the Blanks
- AI that is specialized for a single task is called ______.
Answer
Answer: Narrow AI (or Weak AI)
- The AI process that uses large neural networks with many layers is called ______.
Answer
Answer: Deep Learning
- Before training models we must perform ______ to remove errors and standardize inputs.
Answer
Answer: Data preprocessing (or cleaning)
Multiple Choice (MCQ)
- Which component is NOT part of a typical AI workflow?
- A. Data collection
- B. Model training
- C. Astrological forecasting
- D. Model deployment
Answer
Answer: C — Astrological forecasting is unrelated to a standard AI pipeline.
- Which method is best for image recognition?
- A. Convolutional Neural Networks (CNN)
- B. K-means clustering
- C. Linear regression
- D. Rule-based expert system
Answer
Answer: A — CNNs are specialized for spatial feature extraction in images.
- Which ethical concern is directly related to model outputs?
- A. Energy cost of data centers
- B. Algorithmic bias
- C. Physical security of servers
- D. Trademark registration
Answer
Answer: B — Algorithmic bias affects fairness of outputs.
Matching
Match the left column (term) with the right column (definition).
| Term | Definition |
|---|---|
| 1. Supervised Learning | A. Finds structure in unlabeled data |
| 2. Unsupervised Learning | B. Learns from labeled examples |
| 3. Reinforcement Learning | C. Learns via rewards and penalties |
| 4. Overfitting | D. Model fits training data too closely, poor generalization |
Answer
1 → B; 2 → A; 3 → C; 4 → D.
Descriptive
Explain, in a short paragraph, why data quality matters for AI fairness.
Answer
Data quality affects which patterns a model learns. If training data reflect historical biases or sampling errors, the model will reproduce and amplify those biases. Clean, representative, and well-documented datasets reduce unintended correlations and help ensure the model treats groups equitably; combined with evaluation metrics and human review, good data practices are central to fair AI systems.
Creative
Design a one-paragraph classroom activity that demonstrates the difference between Narrow AI and General AI.
Answer
Activity: Split students into pairs. One student programs a simple rule-based chatbot (or uses a prepared Narrow-AI demo) that answers only weather questions; the other student writes a short essay imagining tasks a General AI could perform across many domains (teaching, diagnosis, creative writing). Pairs then present limitations of the Narrow system and the broad capabilities hypothesized for General AI—this highlights specialization versus generality and grounds the discussion in observable constraints.
Activities, Vocabulary, Homework, Practice Task
Activities
- Hands-on: Run a simple classification demo (e.g., image vs. non-image) and log model errors.
- Discussion: Debate ethical scenarios (privacy vs. utility) in small groups.
Answer
Facilitators should provide starter code for demos and scenario prompts; guide reflection on outcomes and biases.
Vocabulary
- AI, Narrow AI, General AI, Machine Learning, Deep Learning, Dataset, Overfitting, Bias, Explainability
Answer
Ensure students can both define and use these terms in context; include a short quiz next lesson.
Homework
- Read a short article on AI in healthcare and write a 300-word reflection covering benefits and risks.
- List three real-world Narrow AI examples and describe their inputs and outputs.
Answer
Submissions will be evaluated for comprehension and critical thinking; emphasize citation of sources.
Practice Task
Collect a small dataset (10–20 rows) of a personal activity (e.g., daily study hours vs. sleep) and sketch how you would train a simple predictive model on it.
Answer
Describe data fields, how you would clean missing values, what feature(s) to use, and which simple model (e.g., linear regression) might fit—submit a one-page plan.
Instructor notes: emphasize ethics and hands-on learning; adapt examples to local languages and datasets to reinforce inclusivity.