AI in Everyday Life
AI you already use
Everyday AI is often invisible. Search engines rank results, streaming services recommend content, maps predict traffic and reroute, banks score transactions for fraud, email filters classify spam, and voice assistants convert speech to text and infer intent. Predictive text, face-based phone unlocking, smart-home automation, and customer-service chatbots are also common. These systems usually combine data collection, a model that scores or classifies, and an interface that acts on the score.
The benefits are convenience, speed, personalisation, and around-the-clock automation. The trade-offs include privacy (behavioural data may be collected), dependence on data quality, filter bubbles where recommendations narrow exposure, and errors that scale quickly. A recommendation is not a neutral fact: it is a prediction engineered to optimise an objective, such as engagement or sales.
Traps: AI is not confined to robots or chatbots; “the algorithm” is not automatically fair or accurate; personalisation can reduce rather than broaden choice; and a map or assistant only knows what its data sources provide. Questions often test whether you can identify the data input, the model's function, the output, and the human or business decision that follows.
Worked example. A music app notices you skip a genre and suggests similar artists. Input: listening and skip data. Model: predicts likelihood you will keep listening. Output: ranked recommendations. Human decision: whether to press play. Rule of thumb: describe any everyday AI as data in → prediction/classification → interface action → human choice. That chain exposes both usefulness and risk.