AI-102 Question 444
Single answerYou are designing an Azure AI solution that uses Azure OpenAI Service to generate natural language responses for a customer support chatbot. The chatbot needs to deliver contextually accurate responses while avoiding inappropriate or harmful language. Which configuration or approach would best ensure these requirements are met?
- A
Use the 'text-davinci-003' model and implement content filtering with Azure OpenAI moderation tools.
- B
Use the 'text-curie-001' model without additional filtering, as it is optimized for faster response times.
- C
Enable the 'temperature' parameter to 1.5 to provide highly creative and varied responses.
- D
Set up an Azure Logic App to manually review all chatbot responses before delivering them to users.
Show answer and explanation
Correct answer: A
Explanation
The 'text-davinci-003' model is highly capable of generating accurate and contextually relevant responses, which is critical for a customer support chatbot. Combining this with Azure OpenAI moderation tools ensures harmful or inappropriate content is filtered out, adhering to the requirements of the scenario. Other options either compromise response quality, introduce inefficiencies, or fail to address harmful language filtering.
- A. Correct.
This is the correct option because 'text-davinci-003' is a powerful model known for generating contextually accurate responses. Additionally, implementing Azure OpenAI moderation tools ensures that inappropriate or harmful content can be filtered out, meeting the requirements effectively.
- B. Incorrect.
While 'text-curie-001' is faster, it is less accurate and nuanced compared to 'text-davinci-003'. Additionally, not using filtering mechanisms risks delivering inappropriate language.
- C. Incorrect.
Setting the 'temperature' parameter to 1.5 increases creativity but reduces the control over the response accuracy, potentially leading to unpredictable or inappropriate outputs.
- D. Incorrect.
While manual review through an Azure Logic App provides control, it is impractical for real-time chatbot scenarios due to latency and operational inefficiency.