AI, machine learning and neural networks
Describe how AI systems collect and use data, compare learning approaches and evaluate applications, fairness and social impacts.
Section Component 1 · Computer Systems and Logic
Mapped to the Cambridge IGCSE Computer Science 0265 syllabus for exams in 2029. These original explanations and self-checks are independent study support, not Cambridge papers, official marks or grades. Practical programming is taught with worked code and trace tasks; an executable Python workspace and timed digital-exam simulation are not yet available.
Before you begin
- Use the 2029 syllabus for this course. Cambridge specifies Python 3.10 or higher for programming and digital exams for both components. The syllabus currently identifies this route for 2029 only; confirm later-year availability with Cambridge and your school.
Work on paper. Try each question before revealing help, and explain why your method works.
Quick readiness check
Before ai, machine learning and neural networks, which statement is the safest starting point?
What you will learn
- Describe AI features and common applications
- Explain supervised, unsupervised and reinforcement learning and neural-network structure
- Evaluate fairness, ethics, employment, safety and sustainability
AI uses data and rules to produce intelligent behaviour
Artificial intelligence is a branch of computer science focused on simulating intelligent behaviours. AI systems collect data, apply rules, process large datasets, reason and can learn or adapt. Machine learning is a way to improve behaviour from data or feedback rather than manually writing every decision rule.
Supervised learning trains from examples with known outcomes; unsupervised learning finds patterns in unlabelled data; reinforcement learning adjusts actions using rewards or penalties. A neural network has input, hidden and output layers; these layers transform inputs into a result, but the network's output still needs careful evaluation.
Applications create benefits and responsibilities
Applications include recommendation systems, natural language processing and large language models, generative images/video, computer vision and predictive analytics. AI may support robotics or automated systems, but output quality depends on the data, model and context.
Biased or unrepresentative data can lead to unfair results. Evaluation should consider privacy, transparency, misinformation, accountability, employment, safety, energy use and environmental sustainability. A reasoned judgment identifies a specific benefit, a specific risk and a control or condition for responsible use.
Pause and explain
Which habit best checks your understanding of ai, machine learning and neural networks?
Worked example
A model predicts whether a loan application should be reviewed. Its training data excluded some communities. Explain one concern.
Show the worked solution
- The examples may not represent all applicants.
- The model can learn patterns that disadvantage the omitted communities.
- The decision process should be tested across groups and reviewed by people before consequential use.
Answer Unrepresentative data can create unfair predictions; test for bias and include appropriate human review.
Try a topic test
Choose Easy, Normal or Difficult. Save your answers and open worked review after finishing.
Test this topic →From guided practice to a new situation
Write your answer and reasoning first. Use a hint only if you are stuck. The model solution is for self-checking; your written work is not automatically marked.
A music service recommends songs using listening history. Identify the AI application and one type of data it may use.
Give me a hint
Connect the system's output to a recommendation use case.
Compare my reasoning
- Suggesting songs is a recommendation system.
- It may use listening history, likes or skips as input data.
- The system processes patterns to rank songs likely to interest the listener.
Recommendation system using preference or listening data.
Look for these in your work
- Application identified
- Input data example
- Purpose stated
A neural network labels a picture. Describe the role of its input, hidden and output layers.
Give me a hint
Trace information from data entering the model through internal processing to a result.
Compare my reasoning
- The input layer receives image features or pixel data.
- Hidden layers transform the information using learned connections.
- The output layer produces a label or scores for possible classes.
Input receives data, hidden layers transform it, output provides the prediction.
Look for these in your work
- All three layers
- Direction of processing
- Output purpose
A recruitment model is trained on past hiring data that favoured one group. Explain a fairness risk and one way to evaluate it.
Give me a hint
Consider how historical patterns can be learned and compare outcomes across groups.
Compare my reasoning
- The model may reproduce past unequal selection patterns.
- Compare error rates or selection rates across relevant groups using suitable data.
- Review the features, data and decisions with people, then adjust the system and monitor it.
Test for group disparities and review the data and decisions; a model should not be assumed fair from overall accuracy alone.
Look for these in your work
- Bias mechanism
- Evaluation measure
- Review/mitigation
Common mistakes
- Saying AI always learns without data or rules.
- Confusing supervised and unsupervised learning.
- Assuming an accurate average result is fair for every group.
What can you explain now?
Which learning approach uses labelled examples with known outcomes?
Compare with the explanation
Supervised learning
The training examples include the desired output or label.
After trying it yourself, choose your next review. This is your self-assessment.
Your review choice appears on Today. Sign in to sync it across devices.
Make it stick
How do data, learning methods and neural-network layers contribute to AI, and what makes an application fair and responsible?
If you needed the solution, close it and solve the problem again from a blank page. A correct answer is stronger when you can explain the reason for each step.