Generative AI Leader Practice Test 37 — 15 Questions

Question 1

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?

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

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 2

An organization wants to use a generative model to automatically generate legal contracts. The model must produce clauses that are not only grammatically correct but also legally enforceable and consistent with current jurisdiction laws. Which combination of techniques best ensures legal compliance?

Accepted Answer

Fine-tune a model on a diverse set of enforceable contracts and incorporate an external compliance verifier that uses rule-based checks.. Fine-tuning on a diverse corpus of enforceable contracts teaches the model patterns of valid legal language across contexts, while an external rule-based compliance verifier checks generated clauses against jurisdiction-specific statutes and regulations. This hybrid approach combines generative fluency with deterministic legal validation, which is necessary because LLMs alone cannot guarantee current legal compliance. The verifier provides an auditable, updatable layer that can be revised as laws change.

Answer

Fine-tune a small model exclusively on legal contracts from a single jurisdiction and use it for generation.

Answer

Implement retrieval-augmented generation (RAG) with a vector database of all relevant laws.

Answer

Use a large instruction-tuned model with carefully engineered prompts describing jurisdiction details.

Question 3

A startup wants to generate concise summaries of long news articles using an LLM on Vertex AI. They prioritize low latency and cost. Which model choice is most appropriate?

Accepted Answer

Use Gemini 1.5 Flash, which is designed for high throughput and low cost.. Gemini 1.5 Flash is optimized for high-throughput, low-latency, and cost-efficient summarization tasks, making it the ideal choice for a startup that needs to process long news articles quickly without incurring high costs. It balances performance and economy, whereas Gemini 1.5 Pro prioritizes accuracy at higher latency and cost, and PaLM 2 Bison is less efficient for this use case.

Answer

Use Gemini 1.5 Pro for the highest accuracy.

Answer

Use PaLM 2 Bison, as it is the most economical.

Answer

Use Vertex AI Text Embeddings, since embeddings can generate summaries.

Question 4

A data scientist runs the above command to upload a model to Vertex AI Model Registry. The model is a TensorFlow 2.6 model trained on tabular data. After deployment to an endpoint, the prediction latency is higher than expected. What is the most likely cause?

Accepted Answer

The artifact URI points to a single file instead of a directory. When uploading a model to Vertex AI Model Registry, the artifact URI must point to a directory containing the SavedModel or model artifacts, not a single file. If the URI points to a single file, the model may not load properly or may serve with higher latency. For a TensorFlow 2.6 tabular model, GPU acceleration is generally not the primary latency bottleneck; tabular models are typically small and run efficiently on CPU. The container image choice is less likely to be the root cause than an incorrect artifact URI.

Answer

The model should be uploaded with a different display name

Answer

The container image used is CPU-only, but a GPU-accelerated image would improve latency

Answer

The model is uploaded to the wrong region

Question 5

A company is using Vertex AI Model Garden to discover and test various foundation models. They need a model that can generate code from natural language. Which model should they select?

Accepted Answer

Codey. Codey is Google's family of models specifically designed for code generation, including converting natural language descriptions into code. It is built on the PaLM 2 architecture and is optimized for tasks like code completion, code generation, and code chat, making it the correct choice for generating code from natural language.

Answer

Chirp

Answer

Med-PaLM

Answer

Imagen

Question 6

Which Google resource provides interactive visualizations and exercises to help AI practitioners understand concepts like fairness, interpretability, and privacy?

Accepted Answer

PAIR Explorables. PAIR (People + AI Research) Explorables is the correct answer because it is a Google resource specifically designed to provide interactive visualizations and hands-on exercises that help AI practitioners grasp complex concepts like fairness, interpretability, and privacy. Unlike static documentation or tools, Explorables allow users to manipulate parameters and see real-time effects, making abstract responsible AI principles tangible and actionable.

Answer

Model Cards

Answer

People + AI Guidebook

Answer

TensorFlow Fairness Indicators

Question 7

According to Google's AI Principles, AI systems should be 'accountable to people.' What does this principle primarily require?

Accepted Answer

Organizations should provide avenues for feedback and recourse when AI systems cause harm. Google's AI principle of being 'accountable to people' means that AI systems should be subject to human oversight, and organizations should provide mechanisms for feedback, appeal, and redress when AI causes harm. This includes having clear lines of responsibility and the ability to challenge AI decisions. Providing avenues for feedback and recourse is a direct manifestation of accountability.

Answer

AI systems should always have a human monitor present during operation

Answer

The AI system should be able to explain all its decisions to end users

Answer

AI systems should be designed to maximize shareholder value

Question 8

A data scientist is using the Gemini API to generate product descriptions for an e-commerce site. The descriptions are often too verbose and include speculative claims that are not in the product specifications. The scientist wants to reduce hallucinations and control the length of the output without retraining the model. What should they do?

Accepted Answer

Refine the prompt to be concise and include instructions to stick to facts and limit output to 50 words.. Refining the prompt to be concise and include explicit instructions to stick to facts and limit output to 50 words directly addresses both issues without retraining. Prompt engineering is the most effective technique for controlling output length and reducing hallucinations in the Gemini API, as it guides the model's behavior through natural language constraints rather than altering generation parameters. Note that few-shot examples (Option C) are also a form of prompt engineering and could help, but the question asks for the most direct approach: explicit natural-language instructions in the prompt are the simplest and most reliable way to enforce a strict word limit and factual grounding, whereas few-shot examples alone do not guarantee a specific output length.

Answer

Increase the max output token count to 2048 and decrease temperature to 0.1.

Answer

Add three few-shot examples of short, factual descriptions.

Answer

Set temperature to 0.0 and top_k to 1.

Question 9

A company is deploying a generative AI application that must meet HIPAA compliance. They need audit logging, data isolation, and a Business Associate Agreement (BAA). Which Gemini API access tier should they use?

Accepted Answer

Vertex AI. Vertex AI is the correct choice because it is the only Google Cloud platform that offers HIPAA-compliant services with support for audit logging, data isolation, and the ability to sign a Business Associate Agreement (BAA). Vertex AI provides enterprise-grade security controls, including VPC Service Controls, CMEK, and access transparency, which are required for protected health information (PHI) under HIPAA. The Gemini API accessed through Vertex AI inherits these compliance capabilities, making it suitable for regulated healthcare workloads.

Answer

Google AI Studio

Answer

Gemini Nano on-device

Answer

Cloud Vision API

Question 10

A machine learning engineer notices that a generative AI model consistently produces outputs that reinforce gender stereotypes when describing occupations. What is the MOST likely cause?

Accepted Answer

The training data is not representative of diverse populations. The most likely cause is that the training data is not representative of diverse populations. Generative AI models learn patterns, correlations, and biases directly from their training data; if the data over-represents certain demographics or occupations in stereotypical roles, the model will reproduce those associations. This is a well-documented failure mode in NLP models, where biased training data leads to biased outputs even when prompts are neutral.

Answer

The inference temperature is set too high

Answer

The model architecture is too small for the task

Answer

The prompt does not include enough context

Question 11

A financial institution is deploying a generative AI chatbot to provide investment advice. According to regulatory requirements, high-stakes AI decisions must have human review. Which setup BEST satisfies this requirement?

Accepted Answer

AI generates recommendations, but a human advisor must review and approve before any action, and can override the AI. It directly implements the regulatory requirement for human-in-the-loop (HITL) oversight in high-stakes AI decisions. In financial advisory contexts, regulations like the EU AI Act or SEC guidelines mandate that a qualified human advisor must review and approve AI-generated recommendations before any action is taken, ensuring accountability and the ability to override erroneous outputs.

Answer

Use a separate AI model to review the first model's outputs and flag issues

Answer

AI provides advice directly to the customer with a disclaimer that it is not financial advice

Answer

Allow the AI to execute trades automatically, with an audit log for later review

Question 12

A company wants to use Gemini for Google Workspace to improve productivity. They want to generate meeting summaries in Google Meet and draft email replies in Gmail. Which two Duet AI features should they enable? (Choose TWO)

Accepted Answer

Meet 'Take notes for me'. Option A, Meet 'Take notes for me', is correct because this Duet AI/Gemini for Google Workspace feature automatically captures and summarizes meeting content in Google Meet, directly fulfilling the requirement to generate meeting summaries. Option B, Gmail 'Help me write', is correct because it uses generative AI to draft and refine email replies in Gmail, matching the requirement to draft email replies. Option C, Docs 'Help me write', is not correct because it generates and refines text within Google Docs documents, not meeting summaries or Gmail replies. Option D, Sheets formula assistance, is not correct because it helps create and explain spreadsheet formulas in Google Sheets, which is unrelated to the stated use cases. Option E, Slides speaker notes generation, is not correct because it produces speaker notes for Google Slides presentations, not meeting summaries or email replies. Gmail Smart Compose is a long-standing predictive-text feature, not a Duet AI/Gemini for Google Workspace capability, so it does not satisfy the requirement.

Answer

Docs 'Help me write'

Answer

Sheets formula assistance

Answer

Slides speaker notes generation

Question 13

A developer wants to use the Gemini API to generate creative text. Which TWO parameters can they adjust to influence the output?

Accepted Answer

Top-k. Top-k (C) is correct because it limits sampling to the k most probable next tokens, directly shaping the randomness and creativity of the generated text. Temperature (E) is correct because it scales the logits before the softmax, controlling how deterministic or diverse the model's token choices are. Color space (A) is irrelevant since it pertains to image pixel encoding, not text generation parameters. Audio sample rate (B) applies to audio processing, not Gemini text output. Image size (D) affects image dimensions, which has no bearing on text generation creativity.

Answer

Color space

Answer

Audio sample rate

Answer

Image size

Question 14

A retail company is deploying a generative AI chatbot on Vertex AI to provide product recommendations. The chatbot uses a base foundation model with no fine-tuning. Users report that the chatbot sometimes gives offensive or insensitive responses. The team must quickly implement safety controls without modifying the model. They also want to reduce irrelevant off-topic answers. Which combination of techniques should they apply?

Accepted Answer

Enable Vertex AI Safety Filters and craft system instructions defining appropriate behavior.. Vertex AI Safety Filters provide out-of-the-box content moderation without modifying the model, and crafting system instructions (system-level prompts) can constrain the chatbot's behavior to stay on-topic and avoid offensive responses. This combination addresses both safety and relevance without requiring fine-tuning or altering model parameters.

Answer

Fine-tune the model on a curated dataset of safe retail conversations.

Answer

Set temperature to 0.0 and top_p to 0.1.

Answer

Provide 50 few-shot examples of safe interactions.

Question 15

A research team wants to leverage Google DeepMind's work to accelerate drug discovery. They are interested in using a model that predicts protein structures and another that can generate novel protein sequences with desired properties. Which TWO Google DeepMind achievements are most relevant? (Select 2 options.)

Accepted Answer

AlphaFold. AlphaFold (D) is correct because it is Google DeepMind's breakthrough model for predicting protein 3D structures from amino acid sequences, which directly accelerates drug discovery by enabling researchers to understand target proteins. AlphaProteo (E) is correct because it is DeepMind's AI system designed to generate novel protein sequences that bind to specific targets, effectively creating new proteins with desired therapeutic properties.

Answer

WaveNet

Answer

Gemini

Answer

AlphaCode