AI ethics
AI brings benefits but also raises questions of fairness, privacy, truth, ownership, jobs and the environment. People decide how AI is built and used.
📌 Key points
- Bias: if training data is unbalanced or reflects past unfairness, AI can treat some groups unfairly, e.g. a face system that works less well for some skin tones.
- Privacy: AI often uses lots of personal data. People should know and agree how their data is collected and used, and many countries have data protection laws.
- Deepfakes are realistic fake videos, pictures or voices made with AI. They can spread misinformation or hurt people, so check the source before believing or sharing.
- Copyright: generative AI is trained on work made by people, such as writing, art and music. Questions about permission, credit and payment are still being debated.
- Jobs: AI can take over some tasks, change many jobs and create new ones. People may need to learn new skills.
- Energy: training and running big models in data centres uses a lot of electricity, and often water for cooling.
- Transparency and accountability: people should know when AI is being used and why it decided something. Humans, not the AI, are responsible for its decisions.
📖 Key words
- Bias
- When a system unfairly favours or disadvantages some people or groups.
- Privacy
- The right to keep your personal information to yourself and control how it is used.
- Deepfake
- A realistic fake video, picture or voice made with AI.
- Misinformation
- False or misleading information that spreads, sometimes by accident.
- Copyright
- The legal right that protects a person's creative work, such as books, music and art.
- Data centre
- A building full of computers that store data and run online services, including AI.
- Accountability
- Being responsible for decisions and their effects, and able to explain them.
✏️ Worked example
A company trains an AI to pick job applicants using ten years of its past hiring decisions. In the past, it mostly hired men. What could go wrong?
- The AI learns patterns from the past decisions, including the unfair pattern of mostly hiring men.
- It may give lower scores to women, even when they are just as qualified. This is bias.
- The AI is not "choosing" to be unfair. It is copying the pattern in its data.
- Fixes: check results for different groups, balance or correct the data, remove unfair clues, and keep people reviewing the decisions.
Ask three questions about any AI: Is it fair to everyone? Whose data does it use? Who is responsible if it goes wrong?
Test yourself8 questions from a bank of 19. A new mix every time.
Start the quiz →