Generative AI Leader Practice Test 15 — 15 Questions

Question 1

Refer to the exhibit. A team has deployed a model to an endpoint with the configuration shown. They notice that during peak traffic, the endpoint frequently returns 429 (Too Many Requests) errors. Which action should they take to resolve this issue?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Increase MIN_REPLICA_COUNT to 5. The 429 (Too Many Requests) errors indicate the endpoint is receiving more concurrent requests than its current replica capacity can handle. Increasing MIN_REPLICA_COUNT raises the baseline number of serving replicas, so the endpoint has more capacity to absorb peak traffic without throttling. This directly addresses the throughput bottleneck rather than changing the instance shape or disabling scaling.

Answer

Change MACHINE_TYPE to n1-highmem-4

Answer

Decrease MAX_REPLICA_COUNT to 1

Answer

Disable autoscaling by setting MIN_REPLICA_COUNT equals MAX_REPLICA_COUNT

Question 2

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 3

What is the primary purpose of the temperature parameter when generating text with an LLM?

Accepted Answer

It controls the randomness or creativity of the output. Temperature controls the randomness of token selection. Higher temperature increases creativity, lower temperature makes outputs more deterministic.

Answer

It sets the maximum number of tokens in the response

Answer

It determines the number of most likely tokens considered at each step

Answer

It sets the cumulative probability threshold for token selection

Question 4

A music streaming service wants to use AI-generated playlists and artwork, but is concerned about potential copyright infringement. They plan to use a generative model that was trained on a large corpus of publicly available music and images. Which action is MOST important to mitigate IP risk?

Accepted Answer

Review the training data provenance and ensure it consists of properly licensed or public domain works. Reviewing the training data provenance and ensuring it consists of properly licensed or public domain works is the most important action to mitigate IP risk. This directly addresses the root cause: if the model was trained on copyrighted material without permission, the outputs could infringe. By verifying that the training data is legally usable, the service reduces the risk of generating infringing content.

Answer

Only use models hosted on Google Cloud, as Google assumes liability

Answer

Add a watermark to all generated content using SynthID

Answer

Ask the model to self-certify that its outputs are original

Question 5

Which TWO options are benefits of using Vertex AI Model Garden compared to using raw pre-trained models from external sources? (Choose two.)

Accepted Answer

Integration with Vertex AI tools like evaluation and monitoring. Option C is correct because Model Garden models are first-class Vertex AI resources, so they plug directly into Vertex AI Evaluation and Vertex AI Model Monitoring (including skew/drift detection) without custom glue code. Option D is correct because Model Garden provides one-click or API-driven deployment to Vertex AI Endpoints, which handles autoscaling, traffic splitting, and serving infrastructure that you would otherwise build yourself for an externally sourced model. Option A is not inherently true — Model Garden includes both open and partner/proprietary models, and pricing varies, so it is not a guaranteed cost benefit. Option B is not unique to Model Garden, since raw pre-trained models from external sources can also be fine-tuned. Option E is not guaranteed, as data privacy depends on the specific model's terms and your configuration, not on Model Garden itself.

Answer

Lower cost compared to using generic APIs

Answer

Ability to fine-tune models on custom data

Answer

Guaranteed data privacy and no data sharing

Question 6

A support team is using a generative AI model to answer customer questions from an internal knowledge base. They notice that when the same question is asked twice, the model sometimes gives different answers, and occasionally includes details not found in the knowledge base. They want to reduce variability and keep responses closer to the source content. Which action should they take?

Accepted Answer

Decrease the temperature parameter and ground responses with retrieved knowledge base content.. Reducing temperature makes sampling more deterministic, so identical prompts tend to yield consistent answers. Grounding the model with retrieved knowledge base passages supplies authoritative context and reduces hallucinated details. Used together, these changes directly target both observed issues: inconsistent responses and content not present in the source material.

Answer

Increase the temperature parameter to encourage more creative responses.

Answer

Switch to a larger model without changing the prompt or parameters.

Answer

Increase the maximum output tokens so the model has more room to explain its reasoning.

Question 7

A company wants to build a text-to-speech application for generating voiceovers in multiple languages. They need to use a pre-built Google API without training custom models. Which service should they use?

Accepted Answer

Cloud Text-to-Speech API. The Cloud Text-to-Speech API is a pre-built, fully managed Google API that converts text into natural-sounding speech in many languages and voices without requiring any custom model training. It directly matches the requirement for a no-training, multilingual TTS solution.

Answer

Vertex AI Text-to-Speech with a custom model

Answer

Gemini API with a custom prompt

Answer

Cloud Speech-to-Text API

Question 8

A team is evaluating whether to build a custom fine-tuned model or use a pre-built API for a document summarization task. Which TWO factors most strongly support using a pre-built API?

Accepted Answer

The team lacks machine learning expertise to fine-tune a model. A pre-built API eliminates the need for in-house machine learning expertise, as the team can simply call the API endpoint without managing model training, hyperparameter tuning, or infrastructure. This is especially critical when the team lacks the specialized skills required for fine-tuning, such as data preparation, gradient computation, and evaluation on domain-specific metrics like ROUGE.

Answer

The summarization domain is very specialized and requires unique terminology

Answer

The expected usage volume is millions of API calls per day

Answer

The company has strict data residency requirements that prevent sending data to an API

Question 9

Which THREE of the following are generative AI modalities supported by Google Cloud services?

Accepted Answer

Image generation. Option A (Image generation) is correct because Google Cloud's Vertex AI supports image generation through models such as Imagen, which produce images from text prompts. Option B (Code generation) is correct because Vertex AI offers Codey and Gemini models that generate, complete, and explain code. Option C (Text generation) is correct because Vertex AI's PaLM, Gemini, and other large language models generate and summarize text. Option E (Speech generation) is also a genuine generative AI capability on Google Cloud via Vertex AI Text-to-Speech (Chirp voices), but since the question asks for THREE, the intended core modalities are text, code, and image. Option D (Tabular data generation) is not a core generative AI modality offered by Google Cloud's generative AI services.

Answer

Tabular data generation

Answer

Speech generation

Question 10

A research team is fine-tuning a large language model on a dataset containing personal data of EU citizens. They must comply with GDPR. Which measure is ESSENTIAL?

Accepted Answer

Obtain explicit consent from individuals for using their data in fine-tuning. Under GDPR, processing personal data requires a lawful basis; explicit consent is essential when no other basis (e.g., legitimate interest) clearly applies, especially for fine-tuning where data is used to train a model that may memorize and regenerate personal information. Without consent, the processing is unlawful, exposing the team to significant fines and regulatory action. This is a foundational requirement that overrides technical measures like storage location or documentation.

Answer

Ensure the training data is stored in a specific geographic region

Answer

Use a Model Card to document the training data

Answer

Apply safety filters to the model outputs

Question 11

Which Google Cloud AI service is specifically designed for extracting structured data from scanned documents, such as invoices and receipts?

Accepted Answer

Document AI. Document AI is the correct answer because it is purpose-built for understanding and extracting structured data from unstructured documents like invoices, receipts, and forms. It uses specialized processors (e.g., the Invoice Parser or Expense Parser) that combine optical character recognition (OCR) with natural language understanding and machine learning models trained on document layouts, enabling it to output structured fields such as vendor name, total amount, and line items.

Answer

Natural Language AI

Answer

Translation AI

Answer

Vision AI

Question 12

What is the primary purpose of the 'top-p' (nucleus sampling) parameter in text generation?

Accepted Answer

To define a cumulative probability threshold for token selection. The top-p parameter, also known as nucleus sampling, selects the smallest set of tokens whose cumulative probability exceeds the threshold p (e.g., 0.9). This dynamically adapts the candidate pool based on the model's confidence, allowing more diverse tokens when the distribution is flat and fewer when it is peaked. It directly implements a cumulative probability cutoff, not a fixed count or scaling factor.

Answer

To set a fixed number of highest probability tokens to consider

Answer

To limit the maximum number of tokens in the output

Answer

To control the randomness of the output by scaling logits

Question 13

Which TWO methods are most effective for improving factual accuracy in a language model's responses? (Choose two.)

Accepted Answer

Use prompt engineering to instruct the model to rely on provided facts.. Prompt engineering can explicitly instruct the model to base its responses on provided facts, reducing reliance on parametric knowledge that may be outdated or incorrect. By including directives like 'Use only the information in the following text' or 'Answer based solely on the provided context,' the model is guided to prioritize given facts over its internal training data, which improves factual accuracy in the output.

Answer

Decrease the temperature to make responses more deterministic.

Answer

Increase top-k sampling to consider a wider range of tokens.

Answer

Replace the model with a smaller, more focused model.

Question 14

A multinational corporation uses Vertex AI to fine-tune a language model on proprietary customer data. They want to ensure that the fine-tuned model does not inadvertently memorize and regurgitate sensitive customer information. Which approach is most effective?

Accepted Answer

Apply differential privacy techniques during fine-tuning. Differential privacy during training limits memorization by adding noise, making it much harder for the model to leak sensitive data from the training set.

Answer

Regularly audit the model by prompting it with known sensitive phrases

Answer

Use a smaller model to reduce memorization capacity

Answer

Train the model with a higher learning rate to reduce overfitting

Question 15

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