Google Professional Machine Learning Engineer Question 33
Select 3Google Cloud PlatformYour team is building a customer support chatbot using Google Cloud's generative AI capabilities. The chatbot must understand user intent, provide accurate responses, and integrate seamlessly with your current CRM system. Which considerations should you prioritize when selecting between using Google Cloud's pre-trained ML APIs (like Dialogflow CX) versus fine-tuning a foundational model from Vertex AI?
- A
The complexity of the chatbot's conversational flows and domain-specific requirements.
- B
The latency and scalability requirements of the chatbot during peak usage.
- C
The availability of labeled training data specific to your organization's use case.
- D
The need for real-time integration with third-party systems like your CRM.
- E
The ability to customize the chatbot's response style and tone.
Show answer and explanation
Correct answers: A, C, E
Explanation
Selecting between pre-trained ML APIs and fine-tuning foundational models depends on the specific requirements of your AI solution. Pre-trained APIs like Dialogflow CX are quick to implement and provide robust general-purpose functionality, but they may not handle very complex or domain-specific needs as effectively as a fine-tuned model. Additionally, the availability of labeled training data and the need for deep customization in response style and tone are key factors that favor fine-tuning a foundational model.
- A. Correct.
If the chatbot requires very complex conversations or domain-specific knowledge, fine-tuning a foundational model might be better suited than using pre-trained APIs like Dialogflow CX, which are more generalized.
- B. Incorrect.
While latency and scalability are critical considerations, both pre-trained APIs and foundational models can be optimized for these, so this is not a distinguishing factor in the choice between the two.
- C. Correct.
If you have specific labeled training data, fine-tuning a foundational model becomes a more viable option to achieve greater accuracy and customization.
- D. Incorrect.
Real-time integration with third-party systems like CRMs is generally supported by both Dialogflow CX and fine-tuned foundational models, so this is not a differentiating factor.
- E. Correct.
If you require highly specific customization in the chatbot's tone or response style, fine-tuning a foundational model provides more flexibility than pre-trained APIs.