Generative AI Leader Google Cloud's Generative AI Offerings Practice Test 3 — 15 Questions

Question 1

An engineer is testing a generative AI application using the Gemini API. They receive a 400 error with message 'INVALID_ARGUMENT: text has been blocked.' What is the most likely cause?

Accepted Answer

The input text was flagged by a safety filter. The 400 error with 'INVALID_ARGUMENT: text has been blocked' indicates that the input text was rejected by Gemini's safety filters before any model processing occurred. These filters evaluate content against categories like hate speech, harassment, or sexually explicit material, and when triggered, they block the request at the API gateway level, returning this specific error message.

Answer

The specified Gemini model version does not exist

Answer

The API key is invalid

Answer

The API quota has been exceeded

Question 2

A retailer is building a product recommendation chatbot using Vertex AI Agent Builder. They want the agent to answer questions about product availability, prices, and promotions, but also to escalate to a human agent when the query is complex. What should they configure in Agent Builder?

Accepted Answer

Define an agent with a 'handoff to human' intent and configure the corresponding flow. In Vertex AI Agent Builder, you define intents to handle specific user requests. A 'handoff to human' intent, when triggered, initiates a configured flow that transfers the conversation to a human agent. This is the standard method for escalation. Option A is incorrect because playbooks define conversation steps but not directly the escalation trigger; they can be part of a flow but not the primary escalation mechanism. Option C is incorrect: while you could integrate a tool to call a human support API, Agent Builder provides a built-in handoff intent and flow, making custom tool integration unnecessary. Option D is incorrect: generative fallback only generates responses when the agent cannot match a query, it does not provide a structured escalation to a human.

Answer

Create a playbook with a step that transfers to a human via a webhook

Answer

Integrate a tool that calls a human support API when confidence is low

Answer

Use Vertex AI Agent Builder's generative fallback to automatically escalate

Question 3

A data scientist uses Vertex AI Model Evaluation to assess a fine-tuned model for sentiment analysis. The evaluation report shows high precision but low recall on the 'negative' class. What is the best course of action to improve recall without sacrificing too much precision?

Accepted Answer

Collect more labeled examples of negative sentiment and retrain. Collecting more labeled examples of negative sentiment and retraining addresses the root cause of low recall: insufficient or imbalanced training data for the negative class. This improves the model's ability to recognize negative sentiment without sacrificing precision, as the decision boundary is refined with more representative data. Option A (adjusting prediction threshold) can increase recall but typically at the cost of precision, contradicting the goal. Option B (switching model architecture) is excessive and may not fix data imbalance. Option D (using a larger pretrained model) does not specifically target recall on the negative class.

Answer

Adjust the prediction threshold for the negative class

Answer

Switch to a different model architecture (e.g., from BERT to RoBERTa)

Answer

Use a larger pretrained model from Model Garden

Question 4

A global e-commerce company uses Vertex AI Gemini API for real-time product description generation. They observe that sometimes the model generates text in a language other than the user's language, despite being prompted in English. They need to ensure output language consistency. Which approach is most effective?

Accepted Answer

Run a language detection model on the output and regenerate if not English. The most effective approach because Vertex AI Gemini API does not have a built-in language parameter or a safety filter that can block non-English text. Safety filters are limited to predefined harmful content categories. Therefore, implementing a language detection model to verify the output and regenerating if it is not in English is a practical, albeit post-processing, solution. This ensures consistency without relying on unavailable API features.

Answer

Set the language parameter in the generation config to 'en'

Answer

Fine-tune the model on a dataset of English-only product descriptions

Answer

Configure a safety filter that blocks non-English text

Question 5

A company is building a customer service chatbot using Vertex AI Agent Builder. The chatbot needs to answer questions based on a large internal knowledge base stored in a Cloud Storage bucket. The team wants to ensure the model can reference the latest documents without fine-tuning. Which configuration should they use?

Accepted Answer

Ground the model with a Vertex AI Search data store connected to the Cloud Storage bucket. Vertex AI Agent Builder can be grounded with a Vertex AI Search data store connected to Cloud Storage, allowing the model to retrieve the latest documents without fine-tuning. Option A is wrong because fine-tuning is not necessary and would require retraining to incorporate new documents, which is inefficient for a frequently updated knowledge base. Option B is wrong because a pre-built model without additional configuration would not have access to the custom knowledge base. Option C is wrong because while BigQuery can store data, using a BigQuery connector is not the recommended approach for grounding with documents in Cloud Storage; Vertex AI Search data store is designed for this purpose.

Answer

Fine-tune a model on the knowledge base documents

Answer

Use a pre-built model with no additional configuration

Answer

Store the documents in BigQuery and use a BigQuery connector

Question 6

A developer is using Vertex AI's Generative AI Studio to prototype a text summarization model. The initial results are too verbose. What is the most efficient way to adjust the output length without retraining?

Accepted Answer

Modify the prompt with specific length instructions and adjust model parameters. The most efficient way to adjust output length without retraining is to modify the prompt with specific length instructions (e.g., 'Summarize in one sentence') and adjust model parameters like max output tokens, temperature, and top_p in Vertex AI's Generative AI Studio. Option A is wrong because switching to a smaller base model like BERT would require retraining or different model architecture, and BERT is not primarily a text generation model for summarization. Option B is wrong because adding a separate classifier increases complexity and does not directly control the model's output length. Option C is wrong because fine-tuning requires additional training data and computational resources, which is not the most efficient approach compared to prompt and parameter adjustments.

Answer

Switch to a smaller base model like BERT

Answer

Use a separate classifier to filter long responses

Answer

Fine-tune the model with a dataset of concise summaries

Question 7

A startup wants to embed generative AI features into their mobile app but has limited ML expertise. Which Google Cloud service is best suited for rapid integration with no ML training?

Accepted Answer

Gemini API. The Gemini API provides direct, no-code access to Google's most capable generative AI models via a simple REST API, requiring zero ML training or infrastructure setup. This makes it the fastest path for a startup with limited ML expertise to embed generative AI features like text generation, summarization, or chat into a mobile app.

Answer

Vertex AI Model Garden

Answer

Vertex AI Agent Builder

Answer

Cloud Run with a custom container

Question 8

A company is deploying a large language model on Vertex AI for real-time inference. They observe high latency and want to optimize. They have already enabled model caching. What next step should they take to reduce latency?

Accepted Answer

Apply model quantization to reduce precision. Model quantization reduces the precision of the model's weights (e.g., from FP32 to INT8), which decreases memory footprint and accelerates computation on the hardware, directly lowering inference latency. Since Vertex AI already has model caching enabled, quantization is the next logical optimization step to reduce latency without requiring additional infrastructure changes.

Answer

Add more GPUs to the prediction endpoint

Answer

Use a larger, more accurate model variant

Answer

Increase the batch size for inference requests

Question 9

A marketing team wants to use Vertex AI to generate ad copy. They need the model to follow a specific tone and style. What is the best approach?

Accepted Answer

Provide few-shot examples in the prompt and adjust temperature. Few-shot prompting with adjusted temperature is the best approach because it directly controls the model's output style and tone without requiring additional infrastructure or training. By providing a few examples of desired ad copy in the prompt, the model learns the specific tone and style through in-context learning, while temperature tuning (e.g., 0.2 for deterministic output) ensures consistency. This is the most efficient and cost-effective method for immediate, controllable generation.

Answer

Use Vertex AI Grounding to retrieve style guides

Answer

Fine-tune the model on a dataset of past ad copy

Answer

Enable safety filters to enforce brand guidelines

Question 10

A company is building a generative AI application that must adhere to strict data residency regulations. Which TWO Google Cloud features can help ensure that data does not leave a specific geographic region?

Accepted Answer

Using the regional endpoint for Vertex AI. To ensure data residency, you can use the regional endpoint for Vertex AI (A) which ensures that API calls and data processing stay within the specified region. Additionally, deploying models on dedicated VMs in a specific region (E) ensures that compute resources and data do not leave that region. Option B (Vertex AI Model Caching) does not enforce residency as caching may use regional resources but does not guarantee data stays within a region. Option C (Global load balancer with Cloud Armor) is a network security and load balancing service that operates globally. Option D (Cloud CDN) caches content at edge locations globally, which would move data outside the region.

Answer

Vertex AI Model Caching

Answer

Global load balancer with Cloud Armor

Answer

Cloud CDN for content delivery

Question 11

A data scientist is using Vertex AI's Generative AI Studio to experiment with prompt designs. Which THREE features are available in the studio?

Accepted Answer

Grounding configuration. Vertex AI Generative AI Studio includes a grounding configuration feature that allows you to connect prompts to external data sources (e.g., Vertex AI Search, BigQuery, or private datasets) to ground responses in factual, up-to-date information, reducing hallucinations. This is a core capability for enterprise use cases requiring retrieval-augmented generation (RAG).

Answer

Automated hyperparameter tuning

Answer

A/B testing of multiple prompt versions

Question 12

A large enterprise is deploying a multi-modal generative AI application that processes customer support emails (text) and attached screenshots (images). They need to run inference on over 10,000 requests per minute with strict latency requirements (p99 < 500ms). They have already selected Gemini 1.5 Pro as the model and deployed it on Vertex AI using a GPU-based endpoint with autoscaling. During testing, they observe that the p99 latency spikes to over 2 seconds during peak traffic. The application is stateless and requests are independent. The team has access to Cloud Observability and can modify the deployment configuration. Which course of action should the team take to meet the latency requirements while minimizing cost?

Accepted Answer

Enable Vertex AI Model Caching and deploy the endpoint on a managed instance group with larger GPU nodes (e.g., A100 40GB). Vertex AI Model Caching reduces latency by caching the model's weights in GPU memory, eliminating the need to reload them for each request. Deploying on larger GPU nodes (A100 40GB) provides higher memory bandwidth and compute capacity, which directly addresses the p99 latency spike by ensuring the model can process more requests per second without queueing. This combination minimizes cost because it optimizes existing GPU utilization rather than simply adding more replicas, which would increase cost without fixing the root cause of latency.

Answer

Increase the maximum number of replicas in the autoscaling configuration to handle spikes

Answer

Use preemptible VMs for the endpoint to get priority scheduling

Answer

Switch to a CPU-based n2-standard instance to reduce GPU contention

Question 13

A fintech startup is building a generative AI application that generates personalized investment advice based on user profiles and market data. They are using Vertex AI Agent Builder to create an agent that retrieves information from a BigQuery table containing user data and from a real-time market data API. The agent needs to ensure that responses comply with financial regulations, meaning the model must not give specific stock recommendations unless the user explicitly requests them after disclaimers. The team has implemented grounding with both sources. During testing, the agent sometimes spontaneously suggests buying a particular stock without being asked, which could lead to regulatory issues. The team wants to enforce strict control over the agent's behavior. What should the team do?

Accepted Answer

Implement a custom system instruction that explicitly prohibits unsolicited stock recommendations and requires a disclaimer before any advice. System instructions in Vertex AI Agent Builder allow you to define strict behavioral rules that the agent must follow, such as prohibiting unsolicited stock recommendations and requiring a disclaimer before any advice. This directly addresses the regulatory compliance issue by enforcing a policy at the agent's instruction layer, which overrides any learned or grounded behavior. Unlike other options, this approach provides explicit, enforceable control without altering data sources or model training.

Answer

Increase the safety filter sensitivity to block any financial recommendations

Answer

Add more historical data to the BigQuery table to improve grounding accuracy

Answer

Fine-tune the model on a dataset of compliant conversations

Question 14

A retail company is building a chatbot for customer service. They need the model to generate product descriptions based on a catalog but also answer questions about store policies. The team wants to minimize latency and cost while maintaining high accuracy. Which Google Cloud generative AI offering should they use?

Accepted Answer

Vertex AI Gemini API. The Vertex AI Gemini API provides a single, unified multimodal model capable of both generating product descriptions from a catalog and answering policy questions, while offering optimized latency and cost through features like context caching and streaming. Unlike specialized models, Gemini's architecture handles diverse natural language tasks without requiring multiple endpoints, reducing operational overhead and inference time.

Answer

Vertex AI Model Garden with PaLM 2

Answer

Vertex AI Imagen

Answer

Vertex AI Codey APIs

Question 15

A startup wants to deploy a custom-tuned large language model for real-time inference on Vertex AI. They need the lowest possible latency for end users. What deployment strategy should they choose?

Accepted Answer

Deploy the tuned model to a Vertex AI endpoint with GPU acceleration and autoscaling.. Deploying a custom-tuned model to a Vertex AI endpoint with GPU acceleration and autoscaling provides the lowest possible latency for real-time inference. GPU acceleration enables parallel processing of inference requests, while autoscaling ensures sufficient compute resources are available to handle traffic spikes without cold starts. This combination minimizes both compute and network latency, which is critical for real-time user-facing applications.

Answer

Use Vertex AI Model Garden to deploy the base PaLM 2 model.

Answer

Wrap the model in a Cloud Function and invoke via HTTP.

Answer

Use Vertex AI Batch Prediction to process requests in batches.