Ethical AI in practice
Bias, transparency, accountability, law and environmental footprint, with documented cases and concrete actions.
7 lessons · 1 quiz · 35 min · completion badge
- Explain where the biases of an AI system come from, with two documented cases.
- Request and read the documentation of a model and a dataset.
- Place the use of AI within the framework of the revised FADP and the AI Act.
- Estimate the footprint of a use and choose leaner options.
- Set clear accountability rules within a team.
- Who it is for
- Managers, lawyers, AI champions, teachers and anyone who decides how AI is used in an organisation.
- Prerequisites
- The course “What AI can do, and its limits”.
- Duration
- 35 min
- Level
- Advanced
What the course covers.
- 01Bias and fairness2 lessons
- Where bias comes from7 min
The biases of an AI system come from the data, from design choices and from how it is used. Two studies have shown this clearly.
- Measuring and correcting bias6 min
Fairness can be measured, but there are several definitions, sometimes incompatible with one another.
- Where bias comes from7 min
- 02Transparency and accountability2 lessons
- Documenting models and data3 min
Model cards and datasheets: the two documents to request before deploying a system.
- Who is accountable?3 min
A model cannot be accountable. An organisation must state who decides, who approves and who answers for the result.
- Documenting models and data3 min
- 03Law, privacy and frugality3 lessons
- The legal framework in Switzerland and Europe3 min
The revised FADP already applies to AI. The EU AI Act classifies uses by level of risk.
- The environmental footprint3 min
Training and using a model consumes energy. A few choices reduce this footprint significantly.
- The risks specific to large language models6 min
Corpus size, misleading fluency and concentration of power: the critiques that have shaped the field.
- The legal framework in Switzerland and Europe3 min
- ✓Final quiz6 questions