NCA-GENL Question 196
Select 3You are tasked with designing a generative AI system to summarize legal documents. The system must prioritize ethical considerations, minimize energy consumption, and ensure reliable outputs. Which of the following steps should you implement to meet these requirements?
- A
Train the model using a dataset that includes diverse and unbiased legal cases to prevent discriminatory outputs.
- B
Deploy the model on edge devices to reduce latency and energy consumption during inference.
- C
Implement explainability features to allow users to understand why the model generated a specific summary.
- D
Use an overfitting-prone training pipeline for higher accuracy on the training dataset.
- E
Regularly audit the model's performance and energy usage to identify and rectify inefficiencies.
Show answer and explanation
Correct answers: A, C, E
Explanation
To create an ethical, energy-conscious, and reliable generative AI system, it is essential to focus on unbiased training data, include explainability features, and maintain performance audits. These steps align with the principles of responsible AI development and ensure the system's trustworthiness and efficiency.
- A. Correct.
Training the model on a diverse and unbiased dataset helps ensure ethical outcomes by avoiding biases in the AI system's outputs.
- B. Incorrect.
Deploying the model on edge devices can reduce latency, but it does not inherently minimize energy consumption, especially for resource-intensive models.
- C. Correct.
Implementing explainability features is crucial for ethical AI systems as it helps users understand and trust AI decisions, particularly in sensitive applications like legal document summarization.
- D. Incorrect.
Using an overfitting-prone training pipeline can lead to unreliable results on new data, compromising the trustworthiness of the AI system.
- E. Correct.
Regularly auditing the model ensures that it remains efficient and reliable over time, both in terms of accuracy and energy usage.