Data: what AI learns from
Data is information, such as numbers, words, pictures and sounds. AI needs lots of data to learn, and the quality of the data decides how well it works.
📌 Key points
- Data is information: numbers, words, pictures, sounds, videos and more.
- Computers store all data as numbers. A photo, for example, is stored as numbers for the colour of each tiny dot.
- AI learns from data. Without data, a machine learning system has nothing to learn from.
- More data usually helps, but good data matters most: it should be correct, clear and include lots of different examples.
- If the data is wrong or messy, the AI learns the wrong things. People say "rubbish in, rubbish out".
- Data can be collected by sensors (cameras, microphones, thermometers), by people typing it in, or from things people do online.
- Some data is personal data: your name, address, photos, voice or what you search for. It must be kept private and safe.
- A collection of data organised together is called a dataset.
📖 Key words
- Data
- Information such as numbers, words, pictures and sounds.
- Dataset
- A collection of data organised together, for example 10,000 labelled photos.
- Sensor
- A device that collects data from the world, such as a camera, microphone or thermometer.
- Personal data
- Information about a person, such as their name, address, photos or voice.
- Data quality
- How correct, clear and complete the data is.
✏️ Worked example
A class wants to build an AI that tells apples from oranges in photos. What data do they need?
- They need many photos of apples and many photos of oranges, each with the right name (a label).
- The photos should be varied: red and green apples, big and small fruit, different lighting and backgrounds.
- They should check the data. An orange labelled "apple" by mistake would teach the AI something wrong.
- Answer: lots of correct, labelled and varied photos. Better data gives a better AI.
Remember "rubbish in, rubbish out": an AI can only be as good as the data it learns from.
Test yourself8 questions from a bank of 20. A new mix every time.
Start the quiz →