Knowledge check
Evaluate your understanding of this unit
Questions span sustainability definitions, environmental levers, data practices, lifecycle integration and governance.

Action item: Knowledge check
Use the feedback blocks to self-check your answers.
Knowledge check: 8 questions
1. In ESG framing for AI, which pillar covers carbon intensity of training and inference?
- A. Social
- B. Governance only
- C. Liquidity
- D. Environmental
Correct Answer: D
Feedback: Energy and emissions are environmental indicators.
2. Which technique reduces inference cost by representing weights with fewer bits?
- A. Oversampling the majority class indefinitely
- B. Quantisation
- C. Deleting test sets
- D. Increasing model depth without evaluation
Correct Answer: B
Feedback: Lower-precision weights cut memory bandwidth and compute for many architectures.
3. Data minimisation primarily supports sustainability by:
- A. Guaranteeing higher model accuracy
- B. Removing the need for security controls
- C. Reducing storage, movement and recomputation of unnecessary data
- D. Forcing on-prem deployment
Correct Answer: C
Feedback: Less data churn lowers energy and operational risk.
4. Why is experiment tracking a sustainability practice?
- A. It reduces duplicated training runs and helps teams find prior results
- B. It replaces model validation
- C. It removes carbon accounting requirements
- D. It automatically chooses the best cloud region
Correct Answer: A
Feedback: Repeatability cuts wasted GPU hours.
5. Foodforecast-style demand forecasting mainly improves sustainability by:
- A. Increasing batch sizes
- B. Using only synthetic customers
- C. Deleting all weather data
- D. Aligning production with demand to cut waste
Correct Answer: D
Feedback: Waste reduction is the core sustainability lever in that pattern.
6. Which item is essential in a sustainability action plan?
- A. A slogan with no metrics
- B. Clear owners plus measurable indicators and review cadence
- C. A ban on all external data
- D. One-off GPU purchases without utilisation targets
Correct Answer: B
Feedback: Accountability requires measurement and governance rhythm.
7. Advocacy for sustainable AI is most effective when:
- A. Only engineers hear about it
- B. Sustainability is discussed only after a production incident
- C. Incentives, budgets and review gates align teams to efficiency and fairness outcomes
- D. Metrics are hidden from leadership
Correct Answer: C
Feedback: Culture follows incentives and process, not posters alone.
8. In healthcare ML governance, why is documentation of de-identification methods important?
- A. It supports risk assessment for re-identification and regulatory review
- B. It removes the need for encryption
- C. It guarantees zero bias
- D. It replaces patient consent in all cases
Correct Answer: A
Feedback: Methods and residual risk must be explainable to reviewers.