📚 Glossary of AI, ML terms and AWS services referenced in the AWS Certified AI Practitioner (AIF-C01) Exam Guide
A
Accuracy : A model performance metric that measures the proportion of correct predictions made by a machine learning model.
AI (Artificial Intelligence) : Technology that enables machines to perform tasks that typically require human intelligence.
Algorithm : A set of rules or instructions used to solve problems or perform computations in machine learning models.
Amazon A2I (Amazon Augmented AI) : AWS service that makes it easy to build workflows for human review of machine learning predictions.
Amazon Bedrock : AWS service that provides access to foundation models from leading AI companies through a single API.
Amazon CloudFront : AWS content delivery network service that delivers data, videos, applications, and APIs globally.
Amazon CloudWatch : AWS monitoring and observability service for AWS cloud resources and applications.
Amazon Comprehend : AWS natural language processing service that uses machine learning to find insights and relationships in text.
Amazon DocumentDB : AWS managed document database service that is compatible with MongoDB.
Amazon DynamoDB : AWS fully managed NoSQL database service.
Amazon EC2 : AWS Elastic Compute Cloud service that provides scalable computing capacity.
Amazon ECS : Amazon Elastic Container Service for running containerized applications.
Amazon EKS : Amazon Elastic Kubernetes Service for running Kubernetes applications.
Amazon ElastiCache : AWS in-memory caching service.
Amazon EMR : AWS big data platform for processing large amounts of data using open source tools.
Amazon Fraud Detector : AWS service that uses machine learning to identify potentially fraudulent activities.
Amazon Kendra : AWS intelligent search service powered by machine learning.
Amazon Lex : AWS service for building conversational interfaces using voice and text.
Amazon Macie : AWS security service that uses machine learning to discover, classify, and protect sensitive data.
Amazon MemoryDB : AWS Redis-compatible, durable, in-memory database service.
Amazon Neptune : AWS fully managed graph database service.
Amazon OpenSearch Service : AWS managed service for search and analytics engines.
Amazon Personalize : AWS machine learning service for creating personalized recommendations.
Amazon Polly : AWS text-to-speech service that uses advanced deep learning technologies.
Amazon Q : AWS generative AI assistant for business use.
Amazon QuickSight : AWS business intelligence service for creating visualizations and dashboards.
Amazon RDS : Amazon Relational Database Service for managing relational databases.
Amazon Redshift : AWS data warehouse service for analytics.
Amazon Rekognition : AWS computer vision service that analyzes images and videos.
Amazon S3 : Amazon Simple Storage Service for object storage.
Amazon S3 Glacier : AWS long-term archival storage service.
Amazon SageMaker : AWS fully managed service for building, training, and deploying machine learning models.
Amazon SageMaker Clarify : Feature of SageMaker that helps detect bias in machine learning models.
Amazon SageMaker Data Wrangler : Feature of SageMaker for data preparation and feature engineering.
Amazon SageMaker Feature Store : Centralized repository for machine learning features.
Amazon SageMaker JumpStart : Hub for pre-built machine learning solutions and models.
Amazon SageMaker Model Cards : Documentation tool for machine learning models that provides transparency.
Amazon SageMaker Model Monitor : Service for monitoring machine learning models in production.
Amazon Textract : AWS service that extracts text and data from documents.
Amazon Transcribe : AWS speech-to-text service.
Amazon Translate : AWS neural machine translation service.
Amazon VPC : Amazon Virtual Private Cloud for creating isolated cloud resources.
Area Under the ROC Curve (AUC) : Performance metric that measures the ability of a binary classifier to distinguish between classes.
AWS Artifact : AWS service that provides access to compliance reports and agreements.
AWS Audit Manager : AWS service that helps automate audit preparation.
AWS Budgets : AWS service for setting custom cost and usage budgets.
AWS CloudTrail : AWS service that records API calls and account activity.
AWS Config : AWS service that tracks resource configurations and changes.
AWS Cost Explorer : AWS tool for visualizing and managing cloud costs.
AWS Data Exchange : AWS service for finding and subscribing to third-party data.
AWS Glue : AWS serverless data integration service.
AWS Glue DataBrew : AWS visual data preparation service.
AWS IAM (Identity and Access Management) : AWS service for managing access to AWS resources.
AWS Key Management Service (AWS KMS) : AWS service for creating and managing encryption keys.
AWS Lake Formation : AWS service for building secure data lakes.
AWS PrivateLink : AWS service for private connectivity to AWS services.
AWS Secrets Manager : AWS service for managing secrets and credentials.
AWS Trusted Advisor : AWS service that provides recommendations for cost optimization, security, and performance.
AWS Well-Architected Tool : AWS service for reviewing and improving cloud architectures.
B
Batch Inferencing : Type of machine learning inference where predictions are made on large datasets in batches rather than real-time.
BERTScore : Evaluation metric for text generation that leverages pre-trained contextual embeddings.
Bias : Systematic errors in machine learning models that can lead to unfair or discriminatory outcomes.
Bilingual Evaluation Understudy (BLEU) : Metric for evaluating the quality of machine-translated text.
C
Chain-of-Thought : Prompt engineering technique that encourages models to show their reasoning process.
Chunking : Process of breaking down large text into smaller, manageable pieces for processing.
Classification : Machine learning technique for predicting categorical outcomes.
Clustering : Unsupervised learning technique for grouping similar data points.
Computer Vision : AI field focused on enabling machines to interpret and understand visual information.
Continuous Pre-training : Process of further training a foundation model on additional data.
D
Data Curation : Process of organizing, cleaning, and maintaining data for machine learning.
Data Lineage : Documentation of data origins and transformations throughout its lifecycle.
Deep Learning : Subset of machine learning using neural networks with multiple layers.
Diffusion Models : Type of generative model that creates data by reversing a noise process.
E
Embeddings : Numerical representations of data that capture semantic relationships.
Exploratory Data Analysis (EDA) : Process of analyzing datasets to summarize their main characteristics.
F
F1 Score : Performance metric that combines precision and recall into a single score.
Fairness : Principle ensuring AI systems treat all individuals and groups equitably.
Feature Engineering : Process of selecting and transforming variables for machine learning models.
Few-shot Learning : Machine learning approach using only a few examples for training.
Fine-tuning : Process of adapting a pre-trained model for specific tasks or domains.
Fit : How well a machine learning model captures the underlying patterns in data.
Foundation Models : Large-scale AI models trained on broad data that can be adapted for various tasks.
G
Generative AI : Type of artificial intelligence that can create new content, including text, images, and code.
Guardrails for Amazon Bedrock : AWS feature that implements safeguards for responsible AI applications.
H
Hallucinations : When AI models generate false or nonsensical information presented as fact.
Human-centered Design : Design approach that prioritizes human needs and experiences in AI systems.
Hyperparameter Tuning : Process of optimizing model parameters to improve performance.
I
In-context Learning : Ability of models to learn from examples provided in the input prompt.
Inferencing : Process of using trained machine learning models to make predictions on new data.
Instruction Tuning : Fine-tuning technique that trains models to follow specific instructions.
J
Jailbreaking : Technique used to bypass AI model safety restrictions and guardrails.
L
Large Language Model (LLM) : Type of AI model trained on vast amounts of text data to understand and generate human language.
M
Machine Learning (ML) : Subset of AI that enables systems to learn and improve from data without explicit programming.
MLOps (Machine Learning Operations) : Practices for deploying and maintaining machine learning models in production.
Model : Mathematical representation that makes predictions or decisions based on input data.
Multi-modal Models : AI models that can process and generate multiple types of data (text, images, audio).
N
Natural Language Processing (NLP) : AI field focused on enabling machines to understand and process human language.
Neural Networks : Computing systems inspired by biological neural networks used in machine learning.
O
Overfitting : When a model learns training data too specifically and fails to generalize to new data.
P
PartyRock : Amazon Bedrock Playground for experimenting with generative AI applications.
Pre-training : Initial training phase of foundation models on large datasets.
Prompt Engineering : Technique of crafting inputs to guide AI model responses.
Prompt Hijacking : Attack method that manipulates AI model behavior through malicious prompts.
Prompt Injection : Security vulnerability where malicious prompts are used to manipulate AI systems.
Prompt Poisoning : Attack that corrupts training data to influence model behavior.
Prompt Templates : Pre-designed structures for creating effective prompts.
R
Real-time Inferencing : Type of machine learning inference that provides immediate predictions.
Recall-Oriented Understudy for Gisting Evaluation (ROUGE) : Metric for evaluating automatic summarization and machine translation.
Recommendation Systems : AI systems that suggest relevant items or content to users.
Regression : Machine learning technique for predicting continuous numerical values.
Reinforcement Learning : Machine learning approach where agents learn through interaction with an environment.
Reinforcement Learning from Human Feedback (RLHF) : Training technique that uses human preferences to improve model behavior.
Retrieval Augmented Generation (RAG) : Technique that combines information retrieval with text generation.
S
Single-shot Learning : Machine learning approach using only one example for training.
Speech Recognition : AI technology that converts spoken language into text.
Supervised Learning : Machine learning approach using labeled training data.
T
Temperature : Parameter that controls randomness in AI model outputs.
Tokens : Basic units of text that language models process.
Training : Process of teaching machine learning models using data.
Transfer Learning : Machine learning technique that applies knowledge from one task to another.
Transformer-based LLMs : Language models built using the transformer architecture.
U
Underfitting : When a model is too simple to capture underlying patterns in data.
Unsupervised Learning : Machine learning approach that finds patterns in data without labeled examples.
V
Vectors : Mathematical representations of data points in multi-dimensional space.
Z
Zero-shot Learning : Machine learning approach that makes predictions without specific training examples.