AI Fundamentals Exam Prep
AI Fundamentals Exam Glossary - 40 Terms
Search the terminology pack for Artificial Intelligence Fundamentals Certificate. Use these definitions with the study guide and practice questions.
A
- agent
- An entity in reinforcement learning that takes actions in an environment to achieve a goal.
- AI lifecycle
- The full sequence of stages in an AI system, from design and development to deployment and monitoring.
- audio data
- Sound-based digital data used in AI applications such as speech recognition or classification.
- audit scope
- The defined boundaries, processes, systems, and objectives included in an audit review.
B
- bias in content output
- Unfair, skewed, or discriminatory results produced by an AI system due to patterns in its training data or design.
C
- chatbot
- A software application that simulates conversation with users, often for customer service or support.
- classification
- A predictive task where a model assigns inputs to predefined categories or labels.
- cloud-based AI services
- AI tools and platforms delivered over the cloud that provide scalable computing resources and managed capabilities.
- clusters
- Groups of similar data points formed by a clustering algorithm.
- conversational AI
- AI systems designed to interact with users through natural language in chat or voice interfaces.
- customer segmentation
- The process of grouping customers into segments based on similarities in behavior or characteristics.
D
- data augmentation
- Techniques used to create modified versions of existing data to improve model training.
- decision tree
- A machine learning model that splits data into branches based on input attributes to make classifications or predictions.
- deployment
- The stage where an AI model or system is put into operational use.
E
- environment
- The external system or context in which a reinforcement learning agent operates and receives feedback.
G
- GPU
- Graphics Processing Unit; specialized hardware commonly used to accelerate AI and machine learning workloads.
I
- information extraction
- The process of identifying and pulling specific structured information from unstructured text or speech.
- input attributes
- The features or variables used by a model to analyze data and make decisions.
- intent classification
- The task of identifying the user’s goal or purpose from a message in a conversational AI system.
K
- K-means clustering
- An unsupervised learning algorithm that partitions data into K groups based on similarity.
L
- labeled data
- Data that includes both input values and the correct target output used for training supervised models.
M
- Manage function
- The function in the NIST AI Risk Management Framework focused on prioritizing and acting on identified AI risks.
- monitoring
- The ongoing observation of an AI system’s performance, behavior, and risks after deployment.
N
- neural network
- A machine learning model made of interconnected nodes that learns complex patterns from data.
- NIST AI Risk Management Framework
- A framework from NIST for identifying, assessing, prioritizing, and managing risks associated with AI systems.
- nodes
- The processing units in a neural network that receive inputs, apply transformations, and pass outputs forward.
P
- pitch shifting
- An audio augmentation technique that changes the perceived pitch without altering the label.
- prediction
- The process of using learned patterns from data to estimate outcomes for new inputs.
- predictive model
- A model trained on data to estimate future or unknown outcomes.
R
- reinforcement learning
- An AI approach where an agent learns by interacting with an environment and improving through rewards and penalties.
- representative data
- Data that accurately reflects the diversity and characteristics of the population relevant to the AI system.
- reward signal
- Feedback in reinforcement learning that indicates how beneficial an action was toward achieving a goal.
- risk treatment
- Actions taken to reduce, transfer, accept, or mitigate identified risks.
S
- scalability
- The ability of a system to increase or decrease resources efficiently based on workload demand.
- supervised learning
- A machine learning approach that learns from labeled examples to predict outputs for new data.
T
- time stretching
- An audio augmentation technique that changes the speed or duration of audio without changing its label.
- TPU
- Tensor Processing Unit; specialized hardware optimized for machine learning computations.
- training data
- The dataset used to teach an AI or machine learning model to recognize patterns.
U
- unsupervised learning
- A machine learning approach that discovers patterns or structure in unlabeled data.
V
- vulnerability
- A weakness in a system that can be exploited by threats or attackers.
About These Definitions
These definitions are loaded from the shared release pack. Use them with the study guide and practice questions to connect vocabulary to exam scenarios.