NCA-GENL Question 197
Select 3An organization is developing a generative AI model for summarizing large legal documents. They aim to ensure the system is ethical, energy-conscious, and reliable. Which of the following approaches would best align with these goals?
- A
Implementing model compression techniques to reduce energy consumption during inference.
- B
Ensuring training data includes diverse and unbiased legal cases to minimize bias in generated summaries.
- C
Deploying the model without interpretability mechanisms to prioritize performance.
- D
Regularly assessing the model's outputs for factual accuracy and using human feedback to improve reliability.
- E
Training the model on high-performance GPUs without monitoring carbon emissions.
Show answer and explanation
Correct answers: A, B, D
Explanation
Ethical, energy-conscious, and reliable AI systems require careful consideration of energy efficiency, unbiased training data, and output reliability. Model compression reduces energy consumption, diverse training data minimizes bias, and human feedback ensures reliability. Ignoring interpretability or environmental impact contradicts these principles, making such practices unsuitable for developing responsible generative AI systems.
- A. Correct.
Implementing model compression techniques is an effective way to reduce the energy consumption of AI systems, aligning with energy-conscious development practices.
- B. Correct.
Ensuring diverse and unbiased training data helps create ethical AI systems by reducing potential biases in outputs, which is critical for summarizing legal documents fairly.
- C. Incorrect.
Neglecting interpretability mechanisms prioritizes performance over reliability and transparency, which is not aligned with ethical AI development.
- D. Correct.
Regularly assessing outputs for factual accuracy and incorporating human feedback enhances the reliability of the AI system.
- E. Incorrect.
Training without monitoring carbon emissions ignores energy-conscious practices, which is a critical consideration for sustainable AI development.