Artificial intelligence describes systems that perform tasks normally associated with human intelligence. This content sits outside the 2013 computing programme of study, which predates it, but is included as modern context.
Artificial intelligence describes systems designed to do things normally associated with human intelligence. Machine learning finds patterns in data and uses them to make predictions.
Training data is the set of examples used to teach a system, and labelled data pairs each example with its correct category. What is learned is called a model, and using it on new data is inference.
Good training data is representative, varied and accurately labelled. Test data is kept back to check the model on examples it has not seen, and accuracy measures how often it is right.
A model reflects the data it learned from. Under-representing a group produces systematic errors for that group, which is bias.
Transparency means the way a decision was reached can be understood, and human oversight means people check important decisions. Accountability stays with the people and organisations using a system. A confident but false output is often called a hallucination.
Collecting personal data without meaningful consent is a privacy concern, whatever the system is used for.
Natural language processing works with human language, powering chatbots and speech recognition, which converts spoken words to text.
Computer vision interprets images and video, including object detection and facial recognition, which is affected by poor lighting. Classification assigns data to named categories, recommendation systems suggest content, and generative systems produce new content.
Back to Artificial Intelligence practice ยท View all Artificial Intelligence questions
