AWS Certified AI Practitioner Practice Exam

How can a company lower the monthly cost when using a few-shot prompting model hosted on Amazon Bedrock?

1. Customize the model by using fine-tuning

2. Decrease the number of tokens in the prompt

3. Increase the number of tokens in the prompt

4. Use Provisioned Throughput

The correct answer is: Decrease the number of tokens in the prompt

A company can reduce monthly costs when using a few-shot prompting model hosted on Amazon Bedrock by decreasing the number of tokens in the prompt. This approach is effective because costs associated with API calls or model utilization typically depend on the amount of data processed, including the number of tokens in the prompts sent to the model. By lowering the number of tokens, the company minimizes the data that the model has to handle during each request, thus reducing the overall computational resources utilized and ultimately leading to lower expenses. This direct relationship between the size of prompts and costs reflects on how cloud-based AI services bill their users. On the other hand, fine-tuning a model, while potentially beneficial for improving its performance on specific tasks, does not directly address the immediate operational cost associated with usage. Additionally, increasing the number of tokens in the prompt would lead to increased costs due to handling more data. Lastly, using Provisioned Throughput affects capacity and performance rather than directly corresponding to cost reduction in the context of token usage.