Revise Artificial Intelligence for Computer Science 0478 (O Level) — revision notes and instant AI marking. Free to start.
📖 Revision notes · preview
Cambridge IGCSE Computer Science · 0478
Artificial Intelligence
Big idea: AI is a system that can learn from data, decide what to do with it,
and act on its own — and the two exam-tested ways it does this are through
pre-programmed "rulebooks" called expert systems, and self-improving pattern-finders called machine learning.
Summary — everything in this chapter, in one glance
AI = a machine that simulates human-like intelligent behaviour by learning, deciding, and acting autonomously.
Two types: Weak/Narrow AI (one job only) vs Strong AI / AGI (any intellectual task, like a human).
AI has three shared characteristics: it collects data, uses rules to process that data, and has an ability to reason and change its own rules/data.
An expert system mimics a human expert's knowledge to solve problems — built from four parts: knowledge base, rule base, inference engine, interface.
Machine learning is one method of achieving AI — it trains a system on data so it improves accuracy over time by spotting patterns.
AI, and each of its sub-technologies, comes with trade-offs: efficiency and accuracy vs job losses, bias, cost, and lack of "human touch."
1. What Is Artificial Intelligence?
Forget the sci-fi robot image for a second. On the exam, AI is defined as a machine
that can simulate intelligent behaviours similar to a human. The key word is simulate —
it's not actually thinking like a person, it's just really good at behaving as if it were.
Cambridge wants you to describe AI using three specific abilities. Think of them as the "AI checklist":
Learn — it can acquire new information (from data it's given, or data it collects itself).
Decide — it can analyse that information and make choices based on it.
Act autonomously — it can take action without a human telling it what to do each time.
Common mistake
In the exam, an answer like "AI is a smart robot" or "AI is a computer that thinks" earns
zero marks. Examiners are marking against the spec wording. Always anchor your
answer to the three terms: learns, decides, acts autonomously.
Weak AI vs Strong AI
There are two categories of AI you need to know, and they sit on a scale of "how much can it actually do":
Weak AI (Narrow AI)
Strong AI (AGI)
Designed to perform one specific task or a small set of tasks (e.g. a chess engine, a spam filter, a voice assistant that only answers questions).
Designed to perform any intellectual task a human can do — full general intelligence (AGI = Artificial General Intelligence).
This is every AI that exists today — even something as impressive as a self-driving car is still "narrow," because it can't also write you a poem.
This is currently theoretical — it doesn't fully exist yet. It's the "AI can do literally anything a human brain can" version.
Quick way to remember
Weak = one trick pony. Strong = the whole circus. A calculator app that recognises
handwriting is weak AI — brilliant at one thing, useless at anything else. A theoretical AI that
could learn medicine one day and write a novel the next would be strong AI.
Advantages & Disadvantages of AI
Advantages
Disadvantages
Increased efficiency
Job losses
Increased accuracy
Potential for biased decision making
Scalability
Ethical concerns over its use
Practice Question
Q1. Explain why a self-driving car is classified as "weak AI" even though it performs a highly complex task. [2]
Practice Question
Q2. Give one advantage and one disadvantage of using AI in a business. [2]
2. The Three Characteristics of AI
Every AI system, no matter what job it does, shares three common characteristics.
This is one of the most commonly tested definitions in this chapter, so it's worth locking in exactly.
[ COLLECTION OF DATA ]
|
v
[ RULES FOR USING DATA ] ---- processes data using
| algorithms to make
v decisions/predictions
[ ABILITY TO REASON ] ---- can CHANGE its own
rules AND its own data
Characteristic
What it actually means
Collection of data
AI systems need large amounts of data to perform their tasks — without data, there's nothing to learn from.
Rules for using data
The data is processed using rules or algorithms that let the system make decisions and predictions from it.
Ability to reason
The system can change its own rules and its own data — this is what separates AI from a normal program that just follows fixed instructions forever.
Why "ability to reason" is the big one
A normal computer program follows a rulebook it was given and never deviates. An AI system's
defining trick is that it can rewrite its own rulebook based on what it learns from
new data. That self-updating loop is what makes it "intelligent" rather than just "automated."
Exam gold — model answer format
When asked "Describe the characteristics of AI" (a very common 3-mark question), structure your
answer as: "It collects data. It uses rules/algorithms to process that data and make decisions.
It has the ability to reason and can change its own rules/data — for example, learning from mistakes
so it doesn't repeat them." That last part about learning from mistakes is explicitly accepted
in the mark scheme.
Practice Question
Q1. State the three characteristics shared by all AI systems. [3]
Practice Question — Worked Example (from official mark scheme)
Q2. Describe the characteristics of AI. [3]
3. Expert Systems
An expert system is AI's way of putting a human specialist "in a box." It
mimics human knowledge and experience and uses that combination to
solve problems or answer questions — without needing an actual human expert
to be present every time.
Common real-world examples:
Equipment troubleshooting and repair (like a car diagnostic system that tells a mechanic what's wrong)
Technical support (chatbots that walk you through fixing your Wi-Fi)
Medical diagnosis systems (systems that suggest a likely illness based on symptoms)
The four parts of an expert system
This is the single most exam-tested list in this whole chapter. Learn all four, in order:
USER
|
v
[ INTERFACE ] <--- lets the user interact & give input
|
v
[ INFERENCE ENGINE ] --- applies rules from the Rule Base
| ^ to facts from the Knowledge Base
v |
[ RULE BASE ] <----> [ KNOWLEDGE BASE ]
(the logic) (the facts)
Part
What it does
Knowledge base
A database of facts used to generate rules that solve problems and make decisions.
Rule base
A set of rules or logic used to apply the knowledge in the knowledge base to specific problems.
Inference engine
The program that actually applies the rules in the rule base to the facts in the knowledge base to solve problems. This is the "thinking" part.
Interface
The way for users to interact with the system and provide input.
Model answer template — memorise this sentence
"The knowledge base holds facts, the rule base holds logic, the
inference engine applies rules, and the interface allows user interaction."
Advantages & Disadvantages of expert systems
Advantages
Disadvantages
Consistent results
Only as good as the data entered
Faster responses
Responses can be cold and lack human emotion
Can store large amounts of data
Requires training to use correctly
Unbiased
Common mistake
Students often mix up the knowledge base (facts) with the rule base
(logic). Easy way to keep them apart: knowledge base = "what is true," rule base = "what to do
about it." The inference engine is the bridge that connects the two.
Practice Question
Q1. A hospital uses an expert system to help diagnose patients. Identify the four components of this expert system and state what each one does. [4]
Practice Question
Q2. Give one disadvantage of using an expert system for medical diagnosis instead of a human doctor. [1]
4. Machine Learning
This is the part students most often get wrong conceptually, so read this carefully:
Machine learning (ML) is not a separate thing from AI — it's
one method that can help achieve AI. Think of AI as the destination
("a system that learns, decides, and acts") and machine learning as one of the vehicles
that can get you there.
The core idea: by giving a machine data so that it can
"learn over time," it helps train the machine or software to
perform a task and improve its accuracy and efficiency —
all without being explicitly reprogrammed by a human for every new situation.
It does this by using algorithms to analyse data and identify patterns or relationships
within it.
Relationship to remember
Artificial Intelligence (the goal) ← achieved via ← Machine Learning (one method) ← powered by ← Data + Algorithms
Advantages & Disadvantages of machine learning
Advantages
Disadvantages
Reduces the need for manual work, saving time and effort
Needs vast amounts of quality data to perform well
Can detect patterns and make predictions more accurately than humans in many cases
Requires high processing power and resources
Continuously improves performance as it processes more data
Exam phrasing to avoid
Never write "AI and machine learning are different things" — this is factually wrong for the
spec and will lose you marks. Always phrase it as: "Machine learning is a method used to
train AI systems using data."
Practice Question
Q1. Explain the relationship between artificial intelligence and machine learning. [2]
Practice Question
Q2. State one disadvantage of using machine learning. [1]
What To Memorise
Artificial Intelligence (AI)
A machine that simulates intelligent human behaviour by learning, deciding, and acting autonomously.
Weak / Narrow AI
Designed to perform one specific task or set of tasks. All AI in existence today.
Strong AI / AGI
Designed to perform any intellectual task a human can — currently theoretical.
Three characteristics of AI
Collection of data → Rules for using data → Ability to reason (can change its own rules and data).
Expert system
A system that mimics human knowledge and experience to solve problems or answer questions.
Knowledge base
A database of facts used to generate rules for solving problems and making decisions.
Rule base
A set of rules/logic used to apply the knowledge to specific problems.
Inference engine
The program that applies the rule base to the knowledge base to solve problems.
Interface
The way for users to interact with the expert system and provide input.
Machine learning
One method used to achieve AI; trains a system on data using algorithms to spot patterns/relationships and improve accuracy over time.
Concepts Checklist
Exam Tips
Use the exact spec vocabulary
Vague answers like "AI is a smart robot" earn zero marks. Always use: learns, decides,
acts autonomously for AI in general, and collects data, rules for using data,
ability to reason for AI characteristics.
Four-part questions need four parts
Any question about expert systems worth 4 marks is almost certainly asking for
knowledge base, rule base, inference engine, interface — in that specific
four-part structure. Missing one part loses a guaranteed mark.
Don't separate ML from AI
A very common trap is writing about machine learning as if it's a rival technology to AI.
It isn't — it's a method used to train AI systems. Examiners specifically
penalise answers that treat them as unrelated.
Don't confuse knowledge base and rule base
Knowledge base = the facts. Rule base = the logic applied to
those facts. If a question asks "what stores the facts," the answer is knowledge base — not
rule base.
What examiners reward
Full marks come from naming the correct term and explaining what it does in your own
words — a bare list of keywords with no explanation often only earns partial credit on
"describe" or "explain" questions.