NCA-GENM Question 203
Select 3You are tasked with designing a generative AI system that integrates text, images, and audio data. The system must adhere to ethical principles, prioritize energy efficiency, and ensure reliability. Which of the following steps should you take to meet these requirements?
- A
Implement model transparency by documenting how the system processes multimodal data.
- B
Use a large-scale model with the highest possible parameter count to maximize accuracy, regardless of energy consumption.
- C
Conduct bias assessments on the training data to ensure fairness in the model's outputs.
- D
Optimize the model's architecture and inference process to reduce energy consumption while maintaining performance.
- E
Limit the system's evaluation to synthetic datasets to save time and resources.
Show answer and explanation
Correct answers: A, C, D
Explanation
To create an ethical, energy-conscious, and reliable generative AI system, you must prioritize transparency, fairness, and energy efficiency. Transparency ensures the system is verifiable and trustworthy. Bias assessments on training data help mitigate unfair outputs, while optimizing the model reduces energy consumption without compromising performance. Neglecting these principles, such as by using overly large models or relying solely on synthetic datasets, can result in unethical or unreliable outcomes.
- A. Correct.
Implementing transparency ensures that users and stakeholders understand how the AI system processes multimodal data, making it more trustworthy and verifiable.
- B. Incorrect.
Using a large-scale model with the highest parameter count maximizes accuracy but disregards energy efficiency. This is not aligned with the goal of prioritizing energy-conscious design.
- C. Correct.
Bias assessments are critical for ensuring fairness in the AI system's outputs, which is a key ethical principle.
- D. Correct.
Optimizing the model's architecture and inference process helps reduce energy consumption while maintaining performance, aligning with energy-conscious and reliable system design.
- E. Incorrect.
Limiting evaluation to synthetic datasets compromises reliability and fairness, as real-world scenarios often contain complexities not captured in synthetic data.