[Practice Exams] AWS Certified AI Practitioner-Video(Part-2)

Leverage transfer learning to fine-tune a pre-trained model on limited medical image data, boosting accuracy when data is scarce.

On demand pricing offers the most flexibility on Amazon Bedrock, charging only for usage and supporting cross-region inference for text generation, embeddings, and image generation tasks.

Use Amazon Web Service Queue to build a customizable, scalable AI powered chatbot that handles natural language processing and real-time interactions across websites and mobile apps, with easy AWS integrations.

Identify the right AWS storage solution for scalable, durable access to large, diverse data like documents, images, and videos, and explain how Amazon S3 integrates with other services.

Identify how Amazon Q in QuickSight enables natural language queries and instant visualizations, accelerating insights with a generative AI assistant.

Use Amazon Bedrock Guardrails to implement content moderation that detects and filters inappropriate language during user interactions, defining policies and triggering alerts to maintain a safe, professional environment.

Learn how SageMaker model monitor continuously evaluates production model predictions against a predefined baseline to detect data and model quality drift and trigger corrective actions.

Compare error metrics for regression models and penalize large deviations. Root mean squared error provides the best balance of penalty for large errors and interpretability.

Analyze the legal and ethical risks of using code generated by large language models, focusing on plagiarism and copyright infringement, attribution, and licensing.

Reduce the number of output tokens to lower costs when using a token-based generative AI model, aiming for concise responses without significantly impacting performance or accuracy.

Identify the most suitable model for generating both descriptive text and corresponding images from a single prompt by choosing a multimodal model, ensuring cohesive outputs.

Explore why large language models sometimes produce factually incorrect or fabricated outputs, a phenomenon known as hallucination, and examine the reasoning behind selecting the correct term from options.

Discover how Amazon OpenSearch supports AI-driven recommendations with k nearest neighbor search and semantic search to capture context and deliver relevant results.

discover how to detect bias and data imbalances with SageMaker Clarify to ensure fairness in your dataset and model, measuring bias before and after training.

Explain how to create a comprehensive model document with performance, intended use, and limitations using the SageMaker model card.

Explore how AWS Artifact provides on-demand access to security and compliance reports, including SOC reports, PCI, DSS reports, and select online agreements for compliance tasks.

Discover AWS services that host and manage large language models, focusing on SageMaker Jumpstart and Bedrock as the correct options for LM deployment.

Explore AWS services that support vector embeddings and similarity search for retrieval augmented generation, highlighting OpenSearch and RDS for PostgreSQL with the PG vector extension for efficient Rag implementations.

Learn how instruction fine tuning trains a foundation model to understand and follow specific directions, enabling an AI teaching assistant to provide guided, contextually accurate responses.

Develop interpretability to reveal the inner workings of a machine learning model and its decision-making process, delivering transparent, explainable outputs.

This lecture explains retrieval augmented generation, enabling a large language model to access external knowledge during inference with minimal development effort.

Generative AI can generate biased or unsuitable content for product descriptions, requiring human review to ensure content meets the company's standards.

Learn how few-shot prompting with well-crafted product descriptions enables personalized content for e-commerce with minimal extra development, outperforming fine-tuning, custom models, or rule-based systems.

Identify ml pipeline steps that increase features and alter algorithm behavior, focusing on feature engineering and hyperparameter tuning as the correct actions.

Identify self-supervised learning as the technique that generates labels from input data without explicit labeled examples.

Identify the machine learning sequence: data collection, data preprocessing, model training, and model evaluation, and understand how these steps flow toward evaluating model performance with sage maker ground truth labels.

Identify how generative AI techniques enhance Amazon Queues business web application workflow by combining retrieval augmented generation with large language models for natural language processing and external data retrieval.

Explain where Amazon Q developer can be used, showing access via both the Amazon web service management console and IDE, enabling flexible development options.