Artificial Intelligence (AI) – IGCSE 0478 Topic 10
1. What is AI?
Artificial Intelligence is the branch of computer science that aims to make computers or software behave in ways that normally require human intelligence. In the IGCSE syllabus this means that a program can:
- Acquire knowledge or learn from data.
- Reason, make decisions or solve problems using that knowledge.
- Perform a specific type of task without being explicitly programmed for every possible situation.
2. Key characteristics of AI systems (IGCSE level)
- Knowledge acquisition – the system can be supplied with facts, rules or examples.
- Reasoning / decision‑making – it can combine the knowledge to reach conclusions.
- Adaptability – it can be updated (new rules, new data) to improve performance.
- Automation of routine or hazardous tasks – the system can work continuously and without fatigue.
3. AI approaches covered in the Cambridge IGCSE syllabus
3.1 Expert Systems
Expert systems are rule‑based programmes that mimic the decision‑making of a human expert in a narrow domain (e.g., medical diagnosis, fault detection).
- Knowledge base – stores factual information about the domain.
- Rule base – a set of IF‑THEN rules that link facts to conclusions.
- Inference engine – applies the rules to the facts in the knowledge base and derives new facts (the “reasoning” part).
Example – Simple medical diagnosis system
IF temperature > 38°C AND sore_throat = true THEN diagnosis = "flu"
IF temperature ≤ 38°C AND sore_throat = true THEN diagnosis = "cold"
IF cough = true AND fever = false THEN diagnosis = "allergy"
The inference engine checks the patient’s symptoms against the rules and returns the most likely diagnosis.
3.2 Simple Machine Learning (ML)
Machine learning enables a computer to improve its performance on a task by learning from examples rather than by being given explicit rules.
- Training data – a collection of examples that contain inputs and the correct output (the “answer”).
- Model (hypothesis) – the mathematical function that the algorithm builds from the training data.
- Prediction / classification – using the model to produce an output for new, unseen inputs.
Example – Very simple spam‑filter classifier
Training data:
Email 1: “win money now” → spam
Email 2: “meeting agenda attached” → not spam
Email 3: “free offer today” → spam
…
The algorithm counts how often each word occurs in spam versus non‑spam messages. When a new email arrives it adds up the “spam‑indicative” scores and decides whether the mail is spam.
3.3 Other AI techniques mentioned in the syllabus
- Search algorithms – e.g., breadth‑first search used in simple game‑playing programmes.
- Decision‑tree classifiers – a visual way of representing IF‑THEN rules that can be derived from data (a bridge between expert systems and ML).
- Neural‑network basics – only a conceptual understanding is required (neurons, weighted connections, training by adjusting weights).
4. Comparison of Expert Systems and Simple Machine Learning
| Aspect |
Expert System |
Simple Machine Learning |
| How knowledge is obtained |
Explicitly programmed by a human expert (IF‑THEN rules) |
Automatically extracted from training data |
| Typical components |
Knowledge base, rule base, inference engine |
Training data, model, prediction step |
| Explainability |
High – the reasoning path is visible in the rules |
Variable – simple models (e.g., linear) are clear; complex models are less transparent |
| Adaptability to new information |
Requires manual addition/modification of rules |
Can be retrained with new data |
| Typical IGCSE use‑cases |
Medical diagnosis, fault detection, simple tutoring systems |
Spam filtering, basic price prediction, handwritten‑digit recognition |
5. Benefits and limitations of AI (relevant to IGCSE)
- Automation – routine, repetitive or hazardous tasks can be performed without human fatigue.
- Speed and consistency – decisions are made quickly and in the same way each time.
- Personalisation – systems can adapt to individual users (e.g., intelligent tutoring).
- Limitations
- Requires high‑quality data or expert knowledge.
- May produce biased or unexpected results if the input data are unrepresentative.
- Complex models can be difficult to interpret.
6. Ethical and societal considerations
- Data privacy – personal information used for training must be stored securely and used with consent.
- Bias and fairness – if training data contain prejudices, the AI may replicate them.
- Accountability – it must be clear who is responsible for an AI’s decision (the developer, the user or the organisation).
- Impact on employment – automation can replace some jobs while creating new ones that require different skills.
7. Exam‑relevant skills for AI
- Flow‑chart symbols – decision (diamond), process (rectangle), input/output (parallelogram), start/stop (oval).
- Pseudocode conventions –
IF … THEN … ELSE … ENDIF, FOR i = 1 TO n, WHILE … DO … ENDWHILE.
- Trace tables – record the values of variables after each step to check correctness.
- Test data selection – use boundary values, typical values and extreme cases to validate an AI programme.
- Validation vs. verification
- Verification: Does the programme do what the design says?
- Validation: Does the programme solve the real‑world problem?
8. Key points to remember
- AI is about enabling computers to perform tasks that normally need human intelligence.
- For the IGCSE the focus is on expert systems (knowledge base + rule base + inference engine) and simple machine‑learning (training data → model → prediction).
- Both approaches have strengths and weaknesses; the choice depends on the availability of expert knowledge versus large, reliable data sets.
- Understanding the ethical implications and being able to explain how an AI makes a decision are essential exam topics.