AWA AIF-C01 - Domain 2 - Fundamentals of Generative AI

AI that creates new content such as text, images, audio, video, or code based on patterns learned from data.

A large, broadly trained model that can be adapted for many downstream tasks and applications.

A foundation model specialized in understanding and generating human language, typically trained on very large text datasets.

A neural network architecture that uses attention mechanisms to efficiently process relationships between elements in sequential data.

A technique that allows a model to determine which parts of the input are most relevant when producing an output.

A unit of text processed by an LLM. A token may represent a word, part of a word, punctuation, or other text fragment.

The process of breaking input text into tokens that a model can process.

A pricing approach where inference cost is based on the number of input and/or output tokens processed.

The maximum amount of information, measured in tokens, that a model can consider within a single request.

Numerical representations of data that capture semantic meaning and relationships so similar concepts can be compared mathematically.

An ordered collection of numerical values used to represent data such as text, images, or other objects in a mathematical space.

A numerical vector representing the semantic characteristics of an item, commonly used for similarity searches.

Splitting larger documents or content into smaller pieces so they can be efficiently processed, embedded, retrieved, or supplied to a model.

Designing and refining prompts to guide a generative AI model toward useful, accurate, and consistent outputs.

Designing the information and context supplied to a model so it has the relevant information needed to perform a task effectively.

A model capable of working with multiple types of data, such as text, images, audio, or video.

A generative model commonly used for creating images, audio, and other content by learning to reverse a process that progressively adds noise to data.

The initial large-scale training of a foundation model on broad datasets so it learns general patterns and capabilities.

Additional training of an existing model on a more specific dataset to adapt its behavior or performance for a particular task or domain.

Choosing an appropriate model based on capabilities, performance, cost, latency, compliance, and other application requirements.

Measuring how well a model performs against defined requirements, benchmarks, or evaluation datasets.

Making a trained model available for use by an application or users through an inference endpoint or service.

Information collected from model outputs, users, or evaluation processes that can be used to improve an AI application or model.

A model-generated response that appears plausible but contains information that is incorrect, fabricated, or unsupported by available evidence.

The property that can cause a model to produce different outputs for the same or similar input, depending on model and inference settings.

The degree to which humans can understand how a model arrives at its outputs or decisions.

The amount of time between submitting a request and receiving a model response.

The level of computational and architectural sophistication of a model, which can affect capability, cost, speed, and resource requirements.

The ability of GenAI systems to handle varied tasks, inputs, or domains without requiring a separate model for every task.

AI designed to interact with users through natural-language dialogue, maintaining relevant context across an interaction.

Using GenAI to create new text, images, audio, video, code, or other forms of digital content.

An AI application that helps users perform tasks by understanding natural-language requests and generating useful responses or actions.

AI systems that can reason about goals, use tools, interact with external systems, maintain context or memory, and take actions.

A system in which multiple AI agents collaborate, communicate, or specialize in different tasks to accomplish a larger objective.

An open protocol for connecting AI applications or agents with external tools, data sources, and systems through a standardized interface.

An agent's ability to invoke external tools, APIs, databases, or services to obtain information or perform actions beyond the model itself.

Coordinating multiple steps, tools, models, agents, or processes so they work together to accomplish a larger task.

A fully managed AWS service providing access to foundation models from multiple providers through APIs for building generative AI applications.

AWS's managed machine learning service for building, training, customizing, deploying, and operating machine learning models.

A SageMaker capability providing access to pre-trained models, foundation models, and solution templates that can accelerate AI/ML development.