AWS Certified AI Practitioner Practice Exam

What should the accounting firm include when developing and deploying the large language model (LLM) to avoid potential harms?

1. Include fairness metrics for model evaluation

2. Adjust the temperature parameter of the model

3. Modify the training data to mitigate bias

4. Avoid overfitting on the training data

The correct answer is: Include fairness metrics for model evaluation

Incorporating fairness metrics for model evaluation is crucial when developing and deploying large language models (LLMs) to avoid potential harms. Fairness metrics allow the accounting firm to assess how well the model treats different demographic groups, ensuring that it does not exhibit biased behavior or produce discriminatory outcomes. This is particularly important in fields like finance and accounting, where decisions made by automated systems can have significant implications for individuals and companies. By systematically evaluating the model's performance against these fairness metrics, the firm can identify areas where the model might perpetuate existing biases or generate harmful output, enabling them to take corrective actions. This proactive approach helps to build trust in the model, ensures compliance with ethical standards, and reduces the risk of reputational damage due to unfair or biased decision-making. The other options, while they address aspects of model training and performance, do not focus specifically on the broader ethical and social implications that fairness metrics capture. Adjusting the temperature parameter pertains to generating outputs with varying randomness and creativity rather than addressing bias. Modifying the training data to mitigate bias could be beneficial, but it is a more indirect approach that would need to be complemented by ongoing evaluation using fairness metrics. Avoiding overfitting, while important for generalization, does not inherently address