.png)
AI Glossary
Search or filter by alphabet.
3D Reconstruction
Converting 2D images to 3D models.
Accountability
Assigning responsibility for outcomes.
Activation Function
Function controlling neuron output.
Adapter Layers
Lightweight fine-tuning technique.
Adversarial Attack
Manipulating model behavior.
AI
Computational methods that allow machines to perform tasks that normally require human intelligence.
AI Adoption Curve
Stages of organizational maturity.
AI Agent
Autonomous system performing tasks.
AI Assistant
Virtual assistant for employees or customers.
AI Automation
Using AI technologies to automate tasks and processes.
AI Automation System
A system that integrates AI to automate workflows and processes.
AI Center of Excellence (CoE)
Centralized AI leadership team.
AI Competency Model
Skill levels for employees.
AI Governance Framework
Rules for AI adoption and oversight.
AI Literacy
Understanding AI fundamentals.
AI Maturity Model
Framework assessing readiness and capability.
AI Roadmap
Phased plan for AI adoption.
AI Safety
Ensuring models do not cause harm.
AI Strategy
Roadmap aligning AI with business goals.
Algorithm
Step by step set of rules a system follows to solve a problem.
Anomaly Detection
Identifying outliers or unusual patterns.
API
A standard interface that allows two systems to communicate or exchange data.
Artificial Intelligence (AI)
Systems that mimic human intelligence.
Assistant Output
Model's response.
Audio Classification
Categorizing audio signals.
Audio Embeddings
Vector representations of sound.
Autonomous Planning
AI planning steps to reach goals.
Autoscaling
Dynamically adjusting compute resources.
Backpropagation
Training method that adjusts weights based on errors.
Batch Size
Number of examples processed per step.
Beam Search
Search method exploring multiple candidate outputs.
Bias
Systematic unfairness in AI outputs.
BYOK
Bring Your Own Key; a security model allowing users to manage their own encryption keys.
Chain-of-Thought
Model reasoning through steps.
Change Management
Ensuring adoption of AI tools.
Chatbot
Automated conversational tool.
CI/CD
Continuous integration and deployment pipelines.
Classification
Predicting discrete categories.
Cloud AI
Running AI on cloud infrastructure.
Clustering
Grouping similar data points.
Code Generation
Models writing software code.
Cognitive Computing
Systems simulating human thought.
Compute
The processing power required to perform calculations or run algorithms.
Computer Vision (CV)
AI interpreting images/videos.
Containerization
Packaging models into portable units (Docker).
Context Window
How much text a model can "remember."
Contextual AI
AI aware of environment or situation.
Conversational AI
Chatbots/voicebots powered by LLMs.
Convolutional Neural Network (CNN)
Architecture for image tasks.
Correlation
Relationship between variables.
Curriculum Learning
Training models from easy → hard tasks.
Data Augmentation
Expanding datasets artificially.
Data Poisoning
Corrupting training data intentionally.
Data Privacy
Handling data according to policy.
Dataset
Collection of structured data for training.
Decision Engine
System making rule-based choices.
Deep Learning (DL)
ML using multi-layer neural networks.
Differential Privacy
Adding noise to protect individuals.
Digital Transformation
Integrating digital tech into operations.
Digital Twin
Virtual replica of a real-world system.
Dimensionality Reduction
Reducing complexity (e.g., PCA).
Diffusion Models
Models generating images through noise removal.
Diffusion Process
Noise removal generation method.
Drift Detection
Identifying degradation due to data changes.
Dropout
Randomly disabling neurons during training.
Early Stopping
Ending training before overfitting occurs.
Edge AI
Running AI locally on devices.
Edge Compute
Local computation on device.
Embedding
Numerical representation of text meaning.
Entity Extraction
Pulling key details from messages.
Epoch
One full pass through training data.
Explainability (XAI)
Methods explaining model behavior.
Explainable Regression
Transparent regression models.
Face Recognition
Identifying individuals from images.
Fairness
Ensuring equitable treatment.
Feature
Individual measurable variable.
Feature Engineering
Creating additional useful features.
Feature Extraction
Identifying key elements of an image.
Federated Learning
Training across decentralized data sources.
Few-Shot Learning
Model learns with minimal examples.
Fine-Tuning
Retraining a model on domain-specific data.
GAN (Generative Adversarial Network)
Two-model architecture that generates new images.
GDPR Compliance
European data protection standards.
General AI (AGI)
Hypothetical AI with human-like reasoning across domains.
Generative AI
Models that create content.
Generative Consistency
Alignment between prompt and output.
Generative Image Model
AI that creates images.
GPU
Hardware used for ML acceleration.
Gradient Descent
Optimization algorithm.
Ground Truth
Verified correct data.
Hallucination
Confident but incorrect model output.
Hidden Prompt
System-level instructions invisible to user.
Human-in-the-Loop (HITL)
Human oversight during AI tasks.
Hyperparameters
Settings defined before training.
Image Captioning
Generating descriptions of images.
Image Generation
Creating images via prompts.
Image Recognition
Identifying what's in an image.
Image Segmentation
Partitioning image into regions.
Imbalanced Data
Uneven class distribution.
Inference
Model generating answers after training.
Inference Pipeline
Infrastructure serving predictions.
Innovation Lab
Internal space for experimentation.
Inpainting
Filling missing parts of an image.
Instruction Following
LLM's adherence to directives.
Instruction Tuning
Training models to follow human instructions.
Intent Detection
Understanding what the user wants.
Intelligent Automation
RPA + AI decision-making.
Knowledge Base
Collection of organizational data.
Label
Ground truth used in supervised learning.
Large Language Model (LLM)
AI trained on massive text datasets.
Latency
Time it takes for model to respond.
Latency Budget
Acceptable delay for predictions.
Latent Space
Compressed representation used in generation.
Learning Rate
Speed at which model updates.
Lip Reading AI
Decoding speech from movement visuals.
LoRA (Low-Rank Adaptation)
Efficient method for fine-tuning LLMs.
Loss Function
Measures prediction error.
Machine Learning (ML)
Algorithms that learn patterns from data.
MCP
Model Control Protocol; a framework for managing and controlling AI models.
Microservices
Modular deployment architecture.
MLOps
ML lifecycle management.
Model Audit
Evaluation for risks or compliance.
Model Convergence
When training stabilizes.
Model Monitoring
Tracking performance over time.
Model Registry
Central catalog of trained models.
Model Serving
Making a model available via API.
Model Weights
Parameters learned by the model.
Multi-Agent System
Multiple agents collaborating.
Multimodal Model
AI processing text, images, audio, and video simultaneously.
Music Generation AI
Creating music algorithmically.
Named Entity Recognition (NER)
Identifying entities in text.
Narrow AI
AI specialized in one task (e.g., image recognition).
Neural Network
Model inspired by the human brain.
Neural Radiance Fields (NeRFs)
3D scene reconstruction.
NLG (Natural Language Generation)
Creating human-like responses.
NLU (Natural Language Understanding)
Machine comprehension of language.
Object Detection
Identifying objects in images.
OCR (Optical Character Recognition)
Extracting text from images.
Optimizer
Algorithm that reduces loss.
Outpainting
Extending an image beyond its borders.
Overfitting
Model learns noise.
Path Planning
Determining optimal route in robotics.
Perceptron
Simplest neural network unit.
Personalization Engine
Tailoring experiences to users.
Pilot
Limited rollout prior to scaling.
Pose Estimation
Determining body/joint positions.
Privacy-Preserving ML
Methods protecting user data.
Process Mining
Analyzing workflows to optimize them.
Proof of Concept (POC)
Small experiment validating feasibility.
Prompt Engineering
Designing inputs for optimal outputs.
Prompt Template
Reusable structured prompt.
Prompt-to-Image
Generating images from text.
Quantum Machine Learning
ML running on quantum computers.
RAG
Retrieval-Augmented Generation; a method where models retrieve relevant information before generating answers.
RAG (Retrieval-Augmented Generation)
LLM retrieves relevant info before generating answers.
Reasoning Model
A model designed to simulate human-like reasoning processes.
Red Teaming
Stress testing AI for vulnerabilities.
Regression
Predicting continuous values.
Regularization
Techniques preventing overfitting.
Reinforcement Learning
ML with reward-based feedback.
Responsible AI
Ensuring ethical and safe use.
RLHF (Reinforcement Learning from Human Feedback)
Human feedback used to align model behavior.
ROI (Return on Investment)
Financial benefit of AI.
Robotics AI
AI controlling physical robots.
RPA (Robotic Process Automation)
Rule-based automation.
Scaling AI
Deploying across the organization.
Self-Supervised Learning
Model generates labels from the data itself.
Semi-Supervised Learning
Hybrid of labeled and unlabeled data.
Sentiment Analysis
Classifying emotional tone.
Sentiment Detection
Understanding customer emotions.
SLAM
Mapping and navigation in robotics.
SLURM Cluster
Distributed compute scheduling.
Speaker Diarization
Distinguishing who is speaking.
Speech Recognition (ASR)
Converting speech to text.
Strong AI
AI with consciousness and self-awareness (theoretical).
Style Transfer
Applying artistic style to images.
Super-Resolution
Enhancing image quality or scale.
Supervised Learning
ML using labeled data.
Synthetic Data
AI-generated training data.
System Prompt
Defines model behavior and role.
Task Automation
AI completing repetitive tasks.
TCO (Total Cost of Ownership)
True cost of AI deployment.
Temperature
Controls creativity vs. precision.
Test Data
Data used to measure final performance.
Text Classification
Categorizing text into classes.
Text Generation
Process of producing text from a model.
Text Generation
Producing human-like text.
Text-to-Speech (TTS)
Generating spoken audio.
Throughput
Predictions per second.
Token
Unit of text (word, part of word, punctuation).
Tokenizer
Converts text into tokens.
Top-K Sampling
Limits tokens to the top K likely options.
Top-P (Nucleus) Sampling
Chooses tokens from top cumulative probability P.
TPU
Google's ML-specific processors.
Training Data
Data used to teach the model.
Transfer Learning
Using knowledge from one model to improve another.
Transparency
Clarity around how AI works.
Underfitting
Model fails to learn underlying pattern.
Unsupervised Learning
ML finding patterns in unlabeled data.
Use Case Prioritization
Ranking AI opportunities by impact.
User Prompt
Instructions given by user.
Validation Data
Data used to tune hyperparameters.
Vibe Coding
A coding approach focused on creativity and intuition rather than strict rules.
Video Generation
AI-created videos.
Voice Analytics
Insights from call center conversations.
Voice Assistant
Speech-driven AI interface.
Voice Cloning
Replicating a voice using AI.
Weak AI
Practical AI used today, not conscious.
Workflow Automation
Streamlining processes with AI.
Zero-Shot Learning
Model performs tasks with no prior examples.