Courseiva

Google Cloud Generative AI Leader Generative AI Leader (Generative AI Leader) — Questions 976–997

997 questions total · 14pages · All types, answers revealed

Take a mock examExam hub
Page 13

Page 14 of 14

976
MCQhard

A legal firm wants to automate contract analysis to extract key clauses and risks. They have 10,000 contracts in PDF format. The solution must handle varying layouts and be cost-effective. Which approach is BEST?

A.Use Document AI to convert PDFs to structured text, then use Vertex AI with a prompt that specifies clauses to extract
B.Build a retrieval-augmented generation (RAG) system in Vertex AI Agent Builder
C.Use the Gemini API directly with the raw PDF files as input
D.Fine-tune a Gemini model on 100 annotated contracts and run inference on all contracts
AnswerA

Document AI handles layout parsing, and the structured text is then fed to a GenAI model with a well-designed prompt. This combination is scalable and cost-effective.

Why this answer

Using Document AI to parse PDFs into text, then Vertex AI with a structured prompt for clause extraction combines robust document understanding with flexible GenAI. Fine-tuning on 100 contracts is insufficient for layout variation. Agent Builder is overkill.

Direct Gemini API on raw PDFs loses document structure.

Full explanation →
977
Multi-Selecthard

Which THREE are valid methods to reduce bias in generative AI outputs?

Select 3 answers
A.Using only English prompts
B.Increasing model size
C.Using a more diverse training dataset
D.Using safety filters
E.Applying prompt engineering to instruct the model to be fair
AnswersC, D, E

Diverse data reduces the risk of model learning biased patterns.

Why this answer

Option C is correct because training on a more diverse dataset reduces representational bias by exposing the model to a wider range of demographics, cultures, and perspectives. This directly mitigates the model's tendency to overrepresent majority groups or underrepresent minorities, which is a root cause of biased outputs in generative AI.

Exam trap

Google Cloud often tests the misconception that increasing model size or using a single language (like English) can solve bias, when in reality these actions can worsen bias by amplifying existing skews or introducing new cultural blind spots.

Full explanation →
978
MCQmedium

A data scientist is fine-tuning a large language model for a specialized domain using limited labeled data. To avoid catastrophic forgetting and reduce computational cost, which approach is recommended?

A.Using prompt engineering with in-context learning
B.Full fine-tuning of all model parameters
C.Training a new model from scratch on the domain data
D.Adapter-based fine-tuning using LoRA
AnswerD

LoRA is parameter-efficient, cost-effective, and reduces forgetting.

Why this answer

Adapter-based fine-tuning using LoRA (Low-Rank Adaptation) is recommended because it freezes the pre-trained model weights and injects trainable low-rank matrices into the transformer layers. This approach drastically reduces the number of parameters to update (often by 10,000x), lowering memory and compute requirements, while preserving the original knowledge to prevent catastrophic forgetting on limited domain data.

Exam trap

Cisco often tests the misconception that prompt engineering (Option A) is a form of fine-tuning, when in fact it is a zero-shot or few-shot inference technique that does not modify model parameters, making it unsuitable for persistent domain adaptation with limited labeled data.

How to eliminate wrong answers

Option A is wrong because prompt engineering with in-context learning does not update model weights, so it cannot adapt the model to a specialized domain with limited labeled data in a persistent manner; it relies on the model's existing knowledge and context window, which is insufficient for deep domain adaptation. Option B is wrong because full fine-tuning of all model parameters updates the entire model, which is computationally expensive and, with limited labeled data, risks catastrophic forgetting of the original pre-trained knowledge. Option C is wrong because training a new model from scratch on domain data requires a massive amount of labeled data and compute resources, defeating the purpose of leveraging a pre-trained LLM and reducing cost.

Full explanation →
979
MCQhard

A company is deploying a generative AI application that generates medical reports. They need to ensure the output is factual and minimizes hallucinations. Which approach is most effective?

A.Fine-tune the model with RLHF
B.Set the temperature to 0.0
C.Implement retrieval-augmented generation (RAG) with a curated knowledge base
D.Use prompt engineering to instruct the model to be accurate
AnswerC

RAG grounds outputs in retrieved facts, reducing hallucinations.

Why this answer

Retrieval-Augmented Generation (RAG) is the most effective approach because it grounds the model's output in a curated, authoritative knowledge base of medical data. By retrieving relevant, verified documents at inference time, RAG directly reduces the model's reliance on its parametric memory, which is the primary source of hallucinations in generative AI. This is especially critical in high-stakes domains like medical reporting, where factual accuracy is paramount.

Exam trap

The trap here is that candidates often choose 'Set the temperature to 0.0' because they confuse reducing randomness with eliminating factual errors, but temperature only controls output variability, not the truthfulness of the model's internal knowledge.

How to eliminate wrong answers

Option A is wrong because RLHF (Reinforcement Learning from Human Feedback) optimizes the model for human preference alignment and helpfulness, but it does not provide a mechanism to retrieve or verify facts from an external source, so it cannot reliably prevent hallucinations in factual domains. Option B is wrong because setting temperature to 0.0 makes the model deterministic (always picking the highest-probability token), but it does not correct factual errors stored in the model's weights; the model can still confidently generate false information. Option D is wrong because prompt engineering instructs the model to be accurate, but it is a soft constraint that the model can easily override; without external grounding, the model has no way to verify its own output against a trusted source.

Full explanation →
980
MCQmedium

A global news agency is using a generative AI model to summarize breaking news articles in real-time. The model is deployed on Vertex AI across multiple regions (us-central1, europe-west4, asia-southeast1) for low latency worldwide. The agency has a Service Level Objective (SLO) of 99.9% availability and p99 latency under 2 seconds. Recently, during a major event, traffic spiked 10x, and the europe-west4 region experienced latency spikes over 5 seconds and some 503 errors. The team suspects the regional endpoint is under-provisioned. Which combination of actions should they take to meet the SLO consistently?

A.Enable the global endpoint feature in Vertex AI with automatic traffic splitting, and increase the minimum replicas for each regional endpoint
B.Increase the maximum replicas for the europe-west4 endpoint and reduce the min replicas in other regions
C.Implement Cloud CDN caching for common summaries and reduce the number of regions to two
D.Configure a global load balancer with a single Vertex AI endpoint and increase max replicas globally
AnswerA

Global endpoint distributes traffic and increases capacity; higher min replicas prevent cold starts during spikes.

Why this answer

Enabling global endpoint with automatic traffic splitting and increasing min replicas per region (option D) provides both failover and capacity. Simply increasing replicas in europe (A) doesn't help if traffic shifts. Global endpoint without min replicas (B) still risks cold starts.

Using Cloud CDN (C) is for static content, not model inference.

Full explanation →
981
MCQhard

A large enterprise runs a generative AI solution serving millions of daily inference requests. To reduce costs, they propose using serverless endpoints (Vertex AI Prediction) with a custom container, but they notice high latency during cold starts. Which strategy best addresses this problem while minimizing cost?

A.Set a minimum number of replicas to maintain a baseline of always-on instances.
B.Upgrade to GPU-accelerated machines for all replicas.
C.Implement client-side request batching to reduce the number of inference calls.
D.Use prewarmed containers by setting an idle timeout to keep instances alive.
AnswerA

B is correct because it eliminates cold starts for the baseline load, and autoscaling handles additional traffic.

Why this answer

Option A is correct because setting a minimum number of replicas ensures that a baseline of always-on instances is maintained, eliminating cold starts for the majority of requests. This directly addresses the latency spike caused by container initialization and model loading in serverless endpoints, while the cost impact is limited to the minimum replicas rather than scaling all instances.

Exam trap

Google Cloud often tests the misconception that prewarming via idle timeout is a configurable parameter in serverless ML services, but in Vertex AI Prediction, the idle timeout is fixed and not user-adjustable, making minimum replicas the correct approach.

How to eliminate wrong answers

Option B is wrong because upgrading to GPU-accelerated machines increases cost significantly without solving cold start latency; GPUs primarily improve per-request throughput, not initialization time. Option C is wrong because client-side request batching reduces the number of inference calls but does not affect cold start latency; it may even increase perceived latency for individual requests. Option D is wrong because setting an idle timeout to keep instances alive is not a supported mechanism in Vertex AI Prediction; the service uses an internal keep-alive policy, and user-configurable idle timeouts are not available, making this option technically infeasible.

Full explanation →
982
MCQmedium

A healthcare company is developing a generative AI system to assist radiologists by highlighting potential abnormalities in X-ray images. They want to ensure the system's outputs are explainable and can be verified by medical professionals. Which combination of features should they use?

A.AutoML Tables and Vertex Explainable AI
B.Model Cards and Datasheets for Datasets
C.Vertex AI Model Registry and Feature Store
D.Grounding (citing sources) and chain-of-thought reasoning
AnswerD

Grounding provides source references; chain-of-thought shows reasoning steps, enabling radiologists to verify outputs.

Why this answer

Grounding (citing sources) and chain-of-thought reasoning provide explainability by showing the reasoning steps and source evidence. Confidence indicators help radiologists assess reliability. The other options either lack explainability or are less suitable for medical decision support.

Full explanation →
983
MCQeasy

A marketing agency uses generative AI to create social media posts. They need the output to be in a specific JSON format for downstream processing. Which prompt technique should they use?

A.Set the temperature to 0.0
B.Use a model fine-tuned for JSON generation
C.Include a system instruction to output JSON
D.Specify the desired JSON schema in the prompt and use few-shot examples
AnswerD

Explicit schema description plus few-shot examples reliably produce JSON output.

Why this answer

Option D is correct because explicitly specifying the desired JSON schema in the prompt combined with few-shot examples provides the most reliable way to enforce structured output from a generative AI model. This technique leverages the model's pattern-matching ability by showing concrete input-output pairs, which is more effective than vague instructions or parameter adjustments alone for achieving exact JSON formatting.

Exam trap

Cisco often tests the misconception that a simple parameter change (like temperature) or a vague instruction is sufficient to enforce structured output, when in reality explicit schema definition with examples is required for reliable JSON generation.

How to eliminate wrong answers

Option A is wrong because setting temperature to 0.0 only reduces randomness in token selection, making output more deterministic, but does not guarantee the model will output valid JSON or follow a specific schema—it can still produce malformed or non-JSON text. Option B is wrong because while a fine-tuned model may improve JSON generation, fine-tuning requires significant data and compute resources, and the question asks for a prompt technique, not a model customization approach. Option C is wrong because a system instruction to output JSON is a weak constraint—models often ignore or partially follow system instructions, especially for complex schemas, leading to missing fields, extra keys, or incorrect nesting.

Full explanation →
984
Multi-Selectmedium

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?

Select 2 answers
A.Establish a Center of Excellence (CoE) for AI governance to oversee model deployment and monitoring.
B.Use Vertex AI without additional data governance controls to simplify deployment.
C.Use a pre-trained model without customization to reduce development time.
D.Implement model validation and testing to ensure outputs meet regulatory standards.
E.Deploy the model on-premises only to keep data within local infrastructure.
AnswersA, D

A CoE provides centralized governance and best practices.

Why this answer

Options B and D are correct. B: implementing model validation and testing ensures the model behaves as expected and helps meet compliance requirements. D: establishing a Center of Excellence (CoE) for AI governance provides oversight and standardization.

Option A is wrong because using a pre-trained model without customization may not meet specific compliance needs. Option C is wrong because deploying on-premises only is not necessary and may limit scalability. Option E is wrong because Vertex AI without data governance would not satisfy regulatory demands.

Full explanation →
985
MCQhard

A company has been using an on-premises ML infrastructure for generative AI and wants to migrate to Google Cloud. They have a pipeline that fine-tunes a large language model weekly using a proprietary dataset. The migration must minimize downtime and data transfer costs. Which approach best addresses these requirements?

A.Use Vertex AI Pipelines to orchestrate the fine-tuning process, and use Vertex AI Managed Datasets to incrementally sync new data with BigQuery as the source.
B.Use AutoML to train a new model directly from the dataset without fine-tuning.
C.Deploy the existing pipeline on a Google Kubernetes Engine cluster and use Google Cloud Filestore for shared storage.
D.Use Cloud Storage Transfer Service to move all data to Cloud Storage, then set up a Vertex AI custom training job to run the fine-tuning.
AnswerA

C is correct because it allows incremental sync and automated pipeline execution with minimal disruption.

Why this answer

Vertex AI Pipelines provides a managed, serverless orchestration service that can run the weekly fine-tuning workflow with minimal operational overhead, while Vertex AI Managed Datasets can incrementally sync new data from BigQuery, reducing data transfer costs by avoiding full dataset copies. This combination minimizes downtime because the pipeline can be triggered on a schedule without manual intervention, and incremental syncs avoid re-transferring the entire proprietary dataset each week.

Exam trap

The trap here is that candidates often assume full data migration (e.g., Cloud Storage Transfer Service) is necessary, overlooking incremental sync capabilities of Vertex AI Managed Datasets with BigQuery, which directly addresses cost and downtime minimization.

How to eliminate wrong answers

Option B is wrong because AutoML is designed for training models from scratch on labeled data, not for fine-tuning an existing large language model, and it would require a full dataset transfer, increasing costs and downtime. Option C is wrong because deploying the existing pipeline on GKE with Filestore for shared storage does not address data transfer costs (Filestore still requires initial data migration) and introduces additional operational complexity for managing Kubernetes clusters, which does not minimize downtime compared to a managed service. Option D is wrong because using Cloud Storage Transfer Service to move all data to Cloud Storage incurs high initial data transfer costs and does not leverage incremental sync capabilities, and the custom training job setup lacks the orchestration and scheduling benefits of Vertex AI Pipelines, leading to more downtime during migration.

Full explanation →
986
MCQhard

An organization wants to ensure that when their employees use the Gemini API via Vertex AI, the grounding searches are restricted to internal company knowledge bases rather than the public web. Which feature should they enable?

A.Enterprise Data Governance on Vertex AI
B.VPC Service Controls
C.Google Search Grounding
D.Vertex AI Search (with private data indexing)
AnswerD

Vertex AI Search can index internal documents and be used as a grounding source for Gemini API calls.

Why this answer

Vertex AI Search allows organizations to index private, internal data sources (e.g., documents, databases) and use them as the grounding source for Gemini API queries. By enabling this feature, grounding searches are restricted to the indexed private knowledge base, ensuring no public web results are used. This directly meets the requirement to keep grounding internal.

Exam trap

The trap here is that candidates may confuse 'Google Search Grounding' (which uses the public web) with the ability to ground on private data, leading them to select Option C instead of recognizing that Vertex AI Search with private data indexing is the correct solution.

How to eliminate wrong answers

Option A is wrong because Enterprise Data Governance on Vertex AI provides data access controls and audit logging, but it does not restrict grounding sources for Gemini API queries. Option B is wrong because VPC Service Controls create a security perimeter around Google Cloud resources, but they do not control the grounding source used by the Gemini API. Option C is wrong because Google Search Grounding explicitly enables grounding via the public web, which is the opposite of the requirement to restrict to internal knowledge bases.

Full explanation →
987
MCQhard

A large enterprise is deploying a generative AI-powered code assistant for their developers. The solution uses Vertex AI with a fine-tuned Codey model. The security team requires that all prompts and responses be logged for audit purposes, but the logs must not contain sensitive information such as API keys or passwords. The operations team is concerned about high latency during peak usage. You need to design a solution that meets security requirements without compromising performance. Which approach should you take?

A.Use Cloud Audit Logs to capture all API calls to Vertex AI, but do not log the actual prompts and responses
B.Enable Vertex AI model monitoring with Cloud Logging, and configure a log sink with a custom exclusion filter to redact sensitive patterns before storing
C.Log all prompts and responses to Cloud Storage and use a Cloud DLP job to scan and redact sensitive data periodically
D.Implement a custom proxy that logs all requests after stripping sensitive data, then forward to the model
AnswerB

This ensures all interactions are logged but sensitive data is removed, meeting security without major performance impact.

Why this answer

Option B is correct because it uses Vertex AI model monitoring with Cloud Logging to capture prompts and responses, then applies a custom exclusion filter with a log sink to redact sensitive patterns (e.g., API keys, passwords) in real time before logs are stored. This meets the security requirement for audit logging without sensitive data while avoiding the latency overhead of post-processing or a custom proxy, thus satisfying the operations team's performance concern.

Exam trap

Google Cloud often tests the misconception that post-processing redaction (e.g., Cloud DLP) or custom proxies are acceptable for real-time logging, when in fact native streaming redaction via log sinks is required to meet both security and performance constraints.

How to eliminate wrong answers

Option A is wrong because Cloud Audit Logs capture only administrative actions (e.g., model deployment) and not the actual prompts and responses, failing the audit requirement. Option C is wrong because logging all data to Cloud Storage and running a periodic Cloud DLP job introduces significant latency and potential exposure window between logging and redaction, violating the performance requirement. Option D is wrong because implementing a custom proxy adds network hop latency and operational overhead, degrading performance during peak usage, and does not leverage native Vertex AI logging capabilities.

Full explanation →
988
MCQeasy

A data scientist wants to compare the performance of three different foundation models for a text summarization task. They have a labeled dataset of summaries. Which Vertex AI tool should they use to perform this evaluation?

A.Vertex AI RAG Engine - Retrieval evaluation
B.Vertex AI Agent Builder - Agent evaluation
C.Model Garden - Model comparison view
D.Vertex AI Studio - Evaluation
AnswerD

Vertex AI Studio includes evaluation capabilities to compare model responses against ground truth data.

Why this answer

Vertex AI Studio provides a model evaluation interface where you can compare model outputs using automatic metrics. Model Garden is a model hub. Agent Builder is for building agents.

RAG Engine is for retrieval.

Full explanation →
989
MCQeasy

Refer to the exhibit. A developer has defined a dynamic action in the Vertex AI Agent Builder agent YAML. The agent is not triggering the action. What is the most likely issue?

A.The action name is misspelled
B.The endpoint returns a 4xx status
C.The endpoint is not HTTPS
D.The agent is not enabled
AnswerA

The action name must exactly match what the agent tries to invoke.

Why this answer

Option A is correct because the most common reason a dynamic action fails to trigger in Vertex AI Agent Builder is a mismatch between the action name defined in the agent YAML and the name referenced in the agent's configuration or invocation logic. The YAML parser is case-sensitive and requires exact string matching; even a minor typo or capitalization error will cause the agent to ignore the action entirely, as it cannot resolve the reference to the defined action block.

Exam trap

Cisco often tests the distinction between configuration errors that prevent triggering (like name mismatches) versus runtime errors that occur after triggering (like endpoint failures), leading candidates to incorrectly select endpoint-related options when the real issue is a static YAML parsing problem.

How to eliminate wrong answers

Option B is wrong because a 4xx status from the endpoint would cause the action to fail after being triggered, not prevent the action from being triggered at all; the agent would still attempt to invoke the action and log the error. Option C is wrong because Vertex AI Agent Builder does not require the endpoint to be HTTPS for dynamic actions; it supports HTTP endpoints, though HTTPS is recommended for security. Option D is wrong because if the agent were not enabled, the entire agent would not function, not just a single dynamic action; the agent would fail to deploy or respond to any queries.

Full explanation →
990
MCQhard

A regulated healthcare organization needs to use Google Cloud AI services for processing protected health information (PHI). They require a signed Business Associate Agreement (BAA) and HIPAA compliance. Which service provides these assurances?

A.Google AI Studio
B.Colab Enterprise
C.Gemini API directly
D.Vertex AI
AnswerD

Vertex AI can be covered under Google Cloud's HIPAA BAA.

Why this answer

Vertex AI is the correct choice because it is the only Google Cloud AI service that offers a signed Business Associate Agreement (BAA) and is explicitly covered under Google Cloud's HIPAA compliance framework. This allows regulated healthcare organizations to process protected health information (PHI) with the necessary contractual and security assurances.

Exam trap

The trap here is that candidates may assume the Gemini API alone is HIPAA-compliant because it is a Google service, but in reality, only when accessed through Vertex AI (which provides the BAA and enterprise controls) does it meet healthcare regulatory requirements.

How to eliminate wrong answers

Option A is wrong because Google AI Studio is a free, web-based prototyping tool that does not offer a BAA or HIPAA compliance, and it is not designed for production workloads involving PHI. Option B is wrong because Colab Enterprise is a managed notebook environment that, while it can use Google Cloud resources, does not itself provide a BAA or HIPAA compliance for PHI processing. Option C is wrong because the Gemini API directly, when accessed via the standard API endpoint, does not include a BAA or HIPAA compliance; only when used through Vertex AI (which wraps the Gemini API) does the BAA and HIPAA coverage apply.

Full explanation →
991
MCQmedium

A research team at a university is exploring protein folding predictions. They need access to state-of-the-art models and large-scale computational resources at no cost for academic use. Which Google Cloud offering should they use?

A.Colab notebooks with AlphaFold
B.Google Cloud TPU v5e pods via Cloud TPU API
C.Vertex AI Model Registry with AlphaFold as a pre-trained model
D.BigQuery ML with a custom protein folding model
AnswerA

AlphaFold notebooks are available on Colab, providing free GPU resources for protein folding predictions in an academic setting.

Why this answer

Google DeepMind's AlphaFold provides cutting-edge protein structure predictions. Colab offers free GPU notebooks, and AlphaFold is available as a Colab notebook for academic use.

Full explanation →
992
MCQeasy

Which Google DeepMind achievement is recognized for predicting protein structures and advancing drug discovery?

A.AlphaFold
B.AlphaGo
C.Gemini
D.AlphaCode
AnswerA

AlphaFold predicts protein 3D structures from amino acid sequences, revolutionizing biology and drug discovery.

Why this answer

AlphaFold is the correct answer because it is a groundbreaking AI system developed by Google DeepMind that predicts the 3D structure of proteins from their amino acid sequences with atomic-level accuracy. This achievement has revolutionized computational biology by solving a 50-year-old grand challenge in molecular biology, enabling significant advances in drug discovery, enzyme design, and understanding disease mechanisms.

Exam trap

Cisco often tests the distinction between domain-specific AI achievements (like AlphaFold for biology) versus general-purpose AI models (like Gemini) or game-playing AI (like AlphaGo), so candidates may confuse the application area of each DeepMind project.

How to eliminate wrong answers

Option B is wrong because AlphaGo is an AI program that mastered the board game Go using deep reinforcement learning and Monte Carlo tree search, not protein structure prediction. Option C is wrong because Gemini is a multimodal large language model family designed for text, image, audio, and video understanding, not for structural biology or drug discovery. Option D is wrong because AlphaCode is an AI system for competitive programming that generates code solutions to algorithmic problems, not for predicting protein structures.

Full explanation →
993
MCQeasy

Which Google service provides free access to Jupyter notebooks with GPU support for prototyping ML models?

A.Colab
B.Kaggle Notebooks
C.Vertex AI Workbench
D.BigQuery Studio
AnswerA

Colab offers free notebooks with GPU support.

Why this answer

Google Colab is a free notebook service with GPU support, commonly used for prototyping.

Full explanation →
994
MCQeasy

A company wants to generate images for slide decks using Gemini in Google Slides. Which Gemini feature in Google Slides should they use?

A.Speaker notes generation
B.Slide layout optimization
C.Image generation from text prompts
D.Summation of slide content
AnswerC

Gemini in Slides can generate images based on text prompts, ideal for creating visuals.

Why this answer

Gemini in Slides can generate images from text prompts, allowing users to create visuals directly within the presentation.

Full explanation →
995
MCQeasy

What is the primary purpose of the transformer architecture in large language models (LLMs)?

A.To generate images from text descriptions
B.To enable parallel processing of tokens and capture long-range dependencies through self-attention
C.To convert text into numerical embeddings for downstream tasks
D.To store and retrieve information from a vector database
AnswerB

Self-attention allows each token to attend to all others, enabling parallelism and long-range context.

Why this answer

The transformer architecture's primary purpose is to enable parallel processing of all tokens in a sequence while capturing long-range dependencies through its self-attention mechanism. Unlike recurrent neural networks (RNNs) that process tokens sequentially, transformers compute attention scores between every pair of tokens simultaneously, allowing the model to weigh the relevance of distant tokens without the vanishing gradient problem. This parallelization and global context capture are the foundational innovations that make large language models (LLMs) scalable and effective for tasks like text generation and understanding.

Exam trap

The trap here is that candidates confuse the transformer's core innovation (parallel self-attention for sequence modeling) with auxiliary tasks like embedding generation or retrieval, which are separate components in an LLM pipeline, leading them to pick C or D as plausible but incorrect answers.

How to eliminate wrong answers

Option A is wrong because generating images from text descriptions is the domain of multimodal models like DALL·E or Stable Diffusion, which use diffusion or GAN architectures, not the transformer's primary purpose. Option C is wrong because converting text into numerical embeddings is a preprocessing step (e.g., tokenization and embedding layers) that occurs before the transformer processes tokens, but the transformer's core role is to model relationships between those embeddings via self-attention, not merely to create embeddings. Option D is wrong because storing and retrieving information from a vector database is a retrieval-augmented generation (RAG) technique that supplements LLMs with external knowledge; the transformer architecture itself does not inherently function as a database.

Full explanation →
996
MCQmedium

A company deploying a generative AI assistant wants to allow users to override the AI's suggestions before final actions are taken. Which design pattern does this represent?

A.Human-on-the-loop
B.Human-in-the-loop
C.Human-out-of-the-loop
D.Automated decision-making
AnswerB

Human-in-the-loop requires human approval before action, with override capability.

Why this answer

Option B is correct because the Human-in-the-loop (HITL) pattern ensures that a human reviews and can override the AI's suggestions before any final action is executed. This design explicitly maintains human oversight over critical decisions, preventing fully automated actions that could lead to harmful or unintended outcomes.

Exam trap

Cisco often tests the distinction between 'Human-in-the-loop' and 'Human-on-the-loop' by describing scenarios where the human either actively approves actions (in-the-loop) versus passively monitors (on-the-loop), so candidates must carefully note whether the human must intervene before or after the action is taken.

How to eliminate wrong answers

Option A is wrong because Human-on-the-loop implies the human monitors the AI's actions but does not actively intervene in real-time; the system acts autonomously unless the human steps in after the fact. Option C is wrong because Human-out-of-the-loop removes human oversight entirely, allowing the AI to make and execute decisions without any human review or override capability. Option D is wrong because Automated decision-making is a broad category that does not specify any human involvement; it could include fully autonomous systems, which contradicts the requirement for user override before final actions.

Full explanation →
997
MCQhard

An e-commerce company uses a generative AI model to generate product descriptions. They observe that descriptions for high-end products use more sophisticated language compared to budget products, potentially reinforcing class stereotypes. What is the most likely cause, and what should they do to mitigate it?

A.The training data reflects real-world associations; fine-tune with a balanced dataset that includes diverse product descriptions across price ranges
B.The safety filters are too aggressive; relax them
C.The temperature parameter is set too low; increase it to introduce more randomness
D.The model architecture is biased; switch to a different base model
AnswerA

Fine-tuning with balanced data reduces stereotype reinforcement.

Why this answer

The bias stems from training data correlations. Fine-tuning on balanced, diverse data can reduce stereotypical associations. The other options either do not address the root cause or are less effective.

Full explanation →
Page 13

Page 14 of 14

All pages

1234567891011121314

Practice Generative AI Leader by domain

Target a specific domain to shore up weak areas.

Fundamentals of Generative AIBusiness Strategies for Generative AI SolutionsGenerative AI Concepts and TechnologiesGoogle AI Ecosystem and StrategyResponsible AI and Data GovernanceGoogle Cloud's Generative AI OfferingsTechniques to Improve Generative AI Model OutputApplying Generative AI in Business
See all domains with question counts →
Google Cloud Generative AI Leader Generative AI Leader Generative AI Leader Questions 976–997 | Page 14/14 | Courseiva