Google Professional Machine Learning Engineer Question 32
Select 4Google Cloud PlatformYour company needs to implement an AI-powered solution for real-time language translation in a customer support application. You are considering using Google's pre-trained foundational models and ML APIs instead of building a custom solution. Which of the following factors should you consider before making this decision?
- A
The availability of pre-trained models for the specific languages required by your application.
- B
The latency and performance requirements of the real-time translation feature.
- C
The ability to fine-tune the pre-trained model to meet domain-specific translation needs.
- D
The cost associated with using ML APIs for continuous, high-frequency translations.
- E
The need for managing and maintaining the underlying infrastructure for the translation model.
Show answer and explanation
Correct answers: A, B, C, D
Explanation
When using ML APIs or foundational models for building AI solutions, you need to evaluate factors such as the availability of pre-trained models for your use case, latency and performance for real-time applications, customization needs through fine-tuning, and the cost of API usage. These aspects ensure that the solution aligns with your application requirements and business constraints. However, managing infrastructure is not a concern when using ML APIs, as this is fully managed by Google Cloud.
- A. Correct.
The availability of pre-trained models for the required languages is critical because foundational models may not support niche or less commonly spoken languages.
- B. Correct.
Latency and performance are key considerations for real-time applications, as ML APIs must meet the speed requirements for a seamless user experience.
- C. Correct.
The ability to fine-tune pre-trained models is important if the translations need to be adapted for specific jargon, industry terms, or context.
- D. Correct.
Cost is a significant factor for high-frequency API usage since pre-trained APIs typically incur usage-based charges.
- E. Incorrect.
While ML APIs do not require infrastructure management, this option is incorrect in this case as the infrastructure for ML APIs is handled by Google Cloud, and it is not a direct consideration for using them.