Generative AI Leader Practice Test 39 — 15 Questions

Question 1

A data scientist is evaluating how to ground a generative AI model to reduce hallucinations when answering questions about a private knowledge base. Which TWO techniques are most suitable?

Accepted Answer

Retrieval-Augmented Generation (RAG). Option D, Retrieval-Augmented Generation (RAG), is correct because it grounds the model by retrieving relevant passages from the private knowledge base at query time and injecting them into the prompt, so answers are based on actual source documents rather than parametric memory, which directly reduces hallucinations. Option E, prompt engineering instructing the model to answer only from the provided context, is correct because it constrains generation to the supplied grounding text and is the standard companion technique used with RAG to keep responses faithful to retrieved content. Option A, using a larger model like Gemini Ultra, does not by itself ground the model in a private knowledge base and can still hallucinate about proprietary data. Option B, fine-tuning on the private knowledge base, can teach style and some facts but is costly, must be redone as data changes, and does not reliably prevent hallucinations or provide citations. Option C, increasing the temperature to 0.9, raises randomness and would make hallucinations more likely, not less.

Answer

Using a larger model like Gemini Ultra

Answer

Fine‑tuning on the private knowledge base

Answer

Increasing the temperature to 0.9

Question 2

A healthcare organization wants to use generative AI to draft patient education materials. They are concerned about the model generating incorrect medical information. Which combination of Google Cloud services should they use to ground the model's responses in trusted medical literature?

Accepted Answer

Use Retrieval-Augmented Generation (RAG) with Vertex AI Search and Gemini. Retrieval-Augmented Generation (RAG) with Vertex AI Search allows the model to retrieve and cite information from a curated corpus of trusted medical literature before generating responses. This grounds the output in verified sources, reducing the risk of hallucination or incorrect medical advice, while Gemini provides the generative capabilities.

Answer

Use Imagen to generate visual aids and combine with Gemini for text, then manually review

Answer

Fine-tune Gemini on trusted medical literature and deploy with Vertex AI Endpoints

Answer

Use only Gemini Pro with careful prompt engineering and system instructions

Question 3

A retail company is building a generative AI chatbot to assist customers with product recommendations and order tracking. The chatbot uses Vertex AI with Gemini 1.5 Pro, and the development team has implemented a Retrieval-Augmented Generation (RAG) pipeline using Vertex AI Search for grounding. The pipeline uses a vector store containing product descriptions and order history. During testing, the team observes that the chatbot sometimes provides incorrect order statuses—for example, claiming an order is 'shipped' when it is actually 'pending'. The team suspects the issue is related to how context is retrieved and used. The RAG pipeline currently retrieves the top 5 chunks based on cosine similarity from the vector store, and passes them as context to the model. The team is considering several changes to improve factual accuracy. Which single action would most effectively reduce hallucinations in this scenario?

Accepted Answer

Increase the similarity score threshold for retrieval to 0.85 to filter out less relevant chunks.. Increasing the similarity score threshold to 0.85 ensures that only highly relevant chunks are passed to the Gemini 1.5 Pro model, directly reducing the risk of the model generating responses based on irrelevant or low-confidence context. In a RAG pipeline using Vertex AI Search, low-similarity chunks can contain order statuses from different customers or products, leading to hallucinations like incorrect order statuses. Filtering out these less relevant chunks improves the factual grounding of the model's output.

Answer

Switch from Vertex AI Search to a different vector database like Pinecone.

Answer

Reduce the model temperature to 0.0 to make outputs more deterministic.

Answer

Increase the top-K retrieval value to 10 to provide more context to the model.

Question 4

A startup is building a customer support chatbot using Vertex AI and wants to ground responses in their product documentation to reduce hallucinations. Which approach should they use?

Accepted Answer

Enable Vertex AI Grounding with a custom enterprise data store containing the documentation.. Vertex AI Grounding with a custom enterprise data store is the correct approach because it allows the chatbot to retrieve and cite specific chunks from the product documentation in real time, directly reducing hallucinations by constraining responses to verified content. This method uses the underlying grounding service to query a vector-based data store (powered by Vertex AI Search) and append source references to the model's output, ensuring factual accuracy without retraining.

Answer

Use the Codey API for text generation.

Answer

Use the base model without any grounding to maximize flexibility.

Answer

Fine-tune the model on the documentation and deploy.

Question 5

A multinational corporation deploys a generative AI chatbot across multiple regions. They need to comply with GDPR and local data residency requirements. Which THREE actions are necessary?

Accepted Answer

Store and process data only in approved geographic regions (data residency controls). Option C is correct because GDPR and local data residency laws require that personal data be stored and processed only within approved jurisdictions, so implementing geographic data residency controls (e.g., region-pinned storage and compute) is necessary to prevent cross-border transfers. Option D is correct because GDPR accountability and data minimization principles require tracking what personal data the chatbot processes and retaining it only as long as necessary, which is achieved through prompt/response logging with configurable retention and deletion policies. Option E is correct because GDPR Article 32 mandates appropriate technical measures including encryption of personal data at rest and in transit, and using customer-managed encryption keys (CMEK) gives the organization control over key lifecycle and revocation to meet stricter local requirements. Option A is not universally required because GDPR consent is only one of several lawful bases (e.g., contract, legitimate interest) and is not needed for every data collection. Option B is not necessary because anonymization is not required for all training data; pseudonymization or other safeguards may suffice, and fully anonymized data falls outside GDPR scope but is not a blanket obligation.

Answer

Obtain explicit consent from every user before collecting any data

Answer

Anonymize all training data before fine-tuning

Question 6

A startup is building a GenAI application and must decide between using a pre-built API (e.g., Vertex AI Gemini API) or fine-tuning a custom model. Which factor STRONGLY favors using the pre-built API?

Accepted Answer

The startup needs to launch quickly with minimal ML infrastructure and operational overhead. Using a pre-built API like Vertex AI Gemini API eliminates the need to manage ML infrastructure, handle model training, or operationalize a custom model. This allows the startup to integrate GenAI capabilities rapidly via simple API calls, focusing on application logic rather than the complexities of model deployment, scaling, and maintenance.

Answer

The application must process sensitive data that cannot leave the company's VPC

Answer

The startup has a large dataset of labeled examples and high compute budget

Answer

The application requires highly accurate, domain-specific terminology

Question 7

A financial institution is implementing a generative AI chatbot to handle customer inquiries. The institution must comply with regulatory requirements (e.g., GDPR, SOX) and ensure data privacy. Which TWO actions should the institution take?

Accepted Answer

Establish a Center of Excellence (CoE) for AI governance to oversee model deployment and monitoring.. Option A is correct because establishing a Center of Excellence (CoE) for AI governance provides the oversight, policy enforcement, and monitoring needed to keep a generative AI chatbot compliant with GDPR and SOX, covering model deployment, risk management, and accountability. Option D is correct because model validation and testing are essential to verify that chatbot outputs meet regulatory standards, detect bias or data leakage, and ensure ongoing compliance before and after deployment. Option B is incorrect because using Vertex AI without additional data governance controls fails to address GDPR and SOX privacy and audit requirements. Option C is incorrect because a pre-trained model without customization does not by itself satisfy regulatory compliance or data privacy obligations. Option E is incorrect because on-premises deployment alone does not guarantee compliance and may not be feasible or sufficient for the institution's regulatory and operational needs.

Answer

Use Vertex AI without additional data governance controls to simplify deployment.

Answer

Use a pre-trained model without customization to reduce development time.

Answer

Deploy the model on-premises only to keep data within local infrastructure.

Question 8

A company wants to use Google DeepMind's advances in protein structure prediction to accelerate drug discovery. Which DeepMind achievement is most relevant to this goal?

Accepted Answer

AlphaFold. AlphaFold solves protein structure prediction, a key enabler for drug discovery. AlphaGo is for board games, AlphaCode for programming, and WaveNet for audio.

Answer

WaveNet

Answer

AlphaCode

Answer

AlphaGo

Question 9

Refer to the exhibit. This is the IAM policy for a project containing a Vertex AI Agent Builder agent and a data store. The agent is unable to access the data store. What is the most likely cause?

Accepted Answer

The agent service account needs the data store viewer role. The agent service account must have the Data Store Viewer role (or equivalent permissions) to read data from the data store. Without this role, the agent cannot access the indexed content, even if the user has permissions. This is a common IAM misconfiguration in Vertex AI Agent Builder.

Answer

The user needs more permissions

Answer

The agent needs a bigger quota

Answer

The data store is not in the same region

Question 10

Which THREE of the following are key considerations when deploying a generative AI model in a production environment with strict latency requirements? (Choose three.)

Accepted Answer

Implement speculative decoding to generate candidate tokens with a smaller draft model and verify with the large model.. Option B is correct because speculative decoding uses a smaller, faster draft model to propose multiple candidate tokens that the larger target model then verifies in parallel, reducing the number of expensive forward passes and lowering end-to-end latency while preserving output quality. Option C is correct because quantizing weights and activations to lower precision such as int8 shrinks memory bandwidth demands and enables faster matrix multiplications on supported hardware, directly cutting per-token inference time. Option D is correct because caching the key-value tensors from prior decoding steps avoids recomputing attention over the entire sequence for each new token, which is essential for keeping autoregressive generation latency low. Option A is not appropriate because the largest model variant increases compute and memory cost, worsening latency rather than meeting strict requirements. Option E is not appropriate because increasing batch size improves throughput and GPU utilization but typically raises per-request latency, which conflicts with strict latency goals.

Answer

Deploy the largest model variant available to ensure highest quality.

Answer

Increase the inference batch size to maximize GPU utilization.

Question 11

A company develops a generative AI model for resume screening. They discover that the model is rejecting candidates from certain demographic groups disproportionately. Which step should they take first to address unfair bias?

Accepted Answer

Audit the training data for demographic representativeness. Before attempting to fix bias, it is essential to evaluate the training data for representativeness. Bias often originates from imbalanced or unrepresentative training data.

Answer

Reduce the model's complexity to avoid overfitting to biased patterns

Answer

Apply adversarial debiasing to the model

Answer

Collect more data from all demographic groups equally

Question 12

A developer wants to build a RAG application using Vertex AI. Which vector database is natively integrated with Vertex AI for storing embeddings?

Accepted Answer

Vertex AI Vector Search. Vertex AI Vector Search is the native vector database integrated with Vertex AI for storing and querying embeddings. It is purpose-built for high-dimensional vector similarity search, enabling efficient retrieval in RAG applications without requiring external infrastructure.

Answer

Firestore

Answer

Cloud SQL

Answer

Bigtable

Question 13

A company is rolling out a generative AI tool to employees. To ensure successful adoption, they plan to provide training and identify early adopters. Which change management practice is MOST critical early in the rollout?

Accepted Answer

Identify and train AI champions who can support their teams. Identifying AI champions who can advocate and help peers is crucial for organic adoption. Training all employees at once is less effective. Monitoring usage is important but later. Setting up a help desk is reactive.

Answer

Set up a dedicated help desk for AI tool issues

Answer

Conduct mandatory training for all employees before launch

Answer

Monitor usage metrics for the first month

Question 14

A company is using a generative AI model to screen job applications. They want to ensure compliance with Google's AI Principle of avoiding unfair bias. Which practice is most effective in mitigating bias during the screening process?

Accepted Answer

Audit the training data for demographic representativeness and evaluate the model using fairness metrics. Auditing training data for demographic representativeness and evaluating the model using fairness metrics directly addresses the root causes of bias in generative AI systems. This practice aligns with Google's AI Principle of avoiding unfair bias by proactively identifying and mitigating imbalances in the data and measuring the model's performance across demographic groups using metrics like demographic parity or equal opportunity.

Answer

Use a pre-trained model without modification

Answer

Remove all demographic information from the resumes before processing

Answer

Only allow human reviewers to see the top 10% of candidates

Question 15

During a RAG pipeline implementation, the retrieval system frequently returns irrelevant documents, causing the generator to produce incorrect answers. Which change is most likely to improve the relevance of retrieved documents?

Accepted Answer

Add a re-ranking step using a cross-encoder model to refine the top results.. Adding a re-ranking step using a cross-encoder model refines the initial retrieval results by scoring each document against the query more accurately, thus improving relevance. Cross-encoders consider the interaction between query and document, unlike bi-encoders used in vector search, leading to better ranking. The other options are less likely to directly improve relevance.

Answer

Increase the number of documents retrieved from the vector store.

Answer

Use a different embedding model with higher vector dimension.

Answer

Decrease the chunk size of documents to reduce noise.