Cambridge O Level Computer Science · Syllabus 2210 · Automated and Emerging Technologies
Machine Learning
What is Machine Learning?
An area of artificial intelligence in which a program has the ability to automatically adapt its own processes and/or data, building an internal model from example data and using it to classify, predict or decide.
This definition is part of the Automated and Emerging Technologies chapter in Cambridge O Level Computer Science.
Machine Learning in context
An automated system is a system in which sensors, a microprocessor and — where the task requires them — actuators work together so that a process is monitored, and often controlled, with limited or no continuous human intervention. Robotics is a branch of computer science that covers the design, construction and operation of robots — machines that have a mechanical structure, electrical components and a program that can be changed. Artificial intelligence is a branch of computer science concerned with the simulation of intelligent behaviours by computers, and at this level it means two things only: expert systems and machine learning. The three ideas overlap in real products, but they are three separate tests, and an answer that treats them as one loses marks.
Artificial intelligence is a branch of computer science dealing with the simulation of intelligent behaviours by computers. Its main characteristics are the collection of data, the rules for using that data, the ability to reason, and — in the syllabus's own careful wording — it can include the ability to learn and adapt. At this level the examinable scope is limited to two kinds of system: expert systems and machine learning.
Machine learning is when a program has the ability to automatically adapt its own processes and/or data. That is the syllabus definition, and the key words are automatically and its own. Instead of a person writing down every rule, relevant data is supplied, the program identifies useful patterns in it, and it builds or adjusts an internal model. It then uses that model to classify, predict or decide about new inputs, and the model can be checked and adjusted again in the light of how well it performs.
Almost every mark lost in this topic comes from merging two things that the syllabus keeps apart. These three tables are the separations that matter: sensor against microprocessor against actuator, automated system against robot against AI, and expert system against machine learning. If you can reproduce all three from memory, you can answer any comparison question Topic 6 can set.
Common mistakes with Machine Learning
- M19. “An expert system and machine learning are basically the same thing.” Why wrong They differ in where the decision-making comes from: an expert system applies rules a person wrote down; machine learning applies a model the program derived from data. Say instead “In an expert system a specialist writes the rules into the rule base. In machine learning the program finds patterns in example data and builds the model itself.”
- M22. “Machine learning means the program rewrites its own code — and more training data always makes it better.” Why wrong What changes is the model — stored values, weightings and structures. The source code is unchanged. And more data helps only if it is accurate, relevant and representative; more of the wrong data makes a model more confidently wrong. Say instead “The program automatically adapts its own processes and/or data. Adding data helps only if the new data represents the cases the model will actually meet.”
Examiner tips on Machine Learning
- The scope limit is a gift, not a restriction. The syllabus says AI here is limited to expert systems and machine learning. Writing three paragraphs about neural networks or chatbots in an answer about AI components does not gain credit and uses time you need elsewhere. Anything in this chapter marked Beyond the syllabus is context only.
Questions students ask about Machine Learning
Does machine learning mean the program rewrites its own code?
No. What it adapts is its model — stored values, weightings, thresholds and structures that the unchanged program consults when handling a new input. The syllabus wording is that the program “automatically adapts its own processes and/or data”, and that is exactly what it means.