AI Glossary

Understanding the Language of Artificial Intelligence

The field of AI and automation is vast and complex, with many terms and concepts that can be confusing to newcomers. This glossary is designed to help you understand the language of AI, automation, and related fields.

This glossary contains 258 terms relating to artificial intelligence and automation, including:

  • Core AI and machine learning concepts
  • Data science and statistics
  • Automation technologies and practices
  • Hardware and infrastructure
  • Ethics and philosophical considerations
  • Popular culture references and historical context
  • Industry terminology and best practices
  • Environmental and societal impact

Terms are cross-referenced where relevant, allowing you to explore related concepts and build a deeper understanding of the field. If you have any suggestions for additions or corrections, please let me know.

A

Accelerator

Specialized hardware, like GPUs or TPUs, that speeds up AI computations.

Agent

An autonomous AI software that works towards specific goals by perceiving its environment, making decisions, and acting without constant human oversight.

Agentic AI

AI systems with the ability to act independently, make decisions, and adapt to changing circumstances.

AGI Alignment

Ensuring artificial general intelligence systems align with human values and intentions.

AI Snake Oil

A derogatory term for AI products that make exaggerated or false capability claims, often for marketing purposes rather than technological advantage.

AI Washing

Marketing products as AI-powered when they use minimal or no actual AI technology, capitalizing on the AI hype.

AI Winter

Historical periods of reduced interest and funding in AI research, often due to unmet expectations.

AI-Driven Automation

Automation that uses artificial intelligence for adaptive decision-making and learning.

Algorithm

A set of rules or instructions used by a computer to solve problems or perform calculations.

Artificial General Intelligence (AGI)

A hypothetical AI with human-like cognitive abilities across a wide range of tasks.

Artificial Intelligence

The development of machines that simulate human intelligence processes like learning, reasoning, and problem-solving.

Artificial Process Automation (APA)

The use of AI techniques with traditional process automation to improve adaptability.

ASIC (Application-Specific Integrated Circuit)

Custom chips optimized for specific AI workloads, offering improved performance and energy efficiency.

Attention Mechanism

A neural network technique to focus on relevant parts of input data, commonly used in transformers.

Autoencoder

A neural network trained to compress data into a lower-dimensional representation and then reconstruct it.

Automated Decision-Making (ADM)

Systems that make decisions based on predefined rules or AI models without human intervention.

Automated Self-Improvement Methods

Techniques that enable AI systems to enhance their performance using feedback loops and machine learning.

Automated Workflow

A sequence of tasks automatically performed by software to streamline processes.

Automation

The use of technology to perform tasks with minimal human intervention, improving efficiency and accuracy.

Autonomous Agent

An AI system that can independently perform tasks and interact with its environment.

Autonomous System

A self-operating system that makes decisions and performs actions without human control.

B

Backpropagation

A training algorithm used in neural networks to adjust weights based on error gradients.

Base Model

A pre-trained AI model used as a foundation for fine-tuning on specific tasks.

Batch Processing

Processing data or model inference in groups to improve efficiency.

Bayesian Networks

Probabilistic graphical models that represent variables and their dependencies using conditional probabilities.

Bias (AI)

Systematic errors or unfair preferences in AI systems, often due to training data or algorithm design.

Big Data

Extremely large datasets analyzed to reveal patterns and trends.

Black Box

An AI model whose internal workings are not interpretable or explainable to humans.

Bot

A software program that performs automated tasks, often used in chatbots or data scraping.

Brand Voice Integration

Aligning AI-generated content with an organization's tone and style to maintain consistent messaging.

Business Process Automation (BPA)

Technology to automate complex business processes to improve efficiency and reduce costs.

Business Rules Engine (BRE)

A system that automates decision-making processes based on predefined business logic.

C

Carbon Footprint

Total greenhouse gas emissions from AI model training and inference, including energy and hardware.

Categorical Data

Data divided into distinct groups or categories, often requiring special handling in machine learning.

Chain-of-Thought

A prompting technique that helps LLMs break complex problems into smaller, logical steps.

Chatbot Automation

AI-powered software that engages users via text or voice, handling common tasks and inquiries.

ChatGPT

A conversational AI model by OpenAI, based on the GPT architecture.

Chief Automation Officer (CAO)

A strategic executive role responsible for driving enterprise-wide automation and AI initiatives. The CAO ensures digital transformation aligns with business objectives while maintaining operational excellence and employee engagement. Learn more.

Classification

A machine learning task that assigns labels to input data.

Cloud Computing

Remote computing resources used for training and deploying AI models, offering scalability.

Cluster Computing

Connected computers working together to handle large AI workloads, often for distributed training.

Clustering

Grouping similar data points without predefined categories, used in unsupervised learning.

Cognitive Automation

The combination of AI and automation for tasks requiring reasoning and learning.

Compliance Measures

Standards ensuring AI systems operate within legal boundaries, prioritizing data security and ethical guidelines.

Compute

The processing power required to run AI models, often measured in FLOPS.

Compute Resources

The hardware and infrastructure needed to train and run AI models, including processors and memory.

Computer Vision

AI that enables computers to understand and process visual information.

Conditional Automation

A system that automates tasks only when certain conditions or criteria are met.

Consciousness

The state of being aware, able to think and feel, relevant to discussions of AGI and personhood.

Context Window

The maximum text a LLM can process in one interaction.

Continuous Improvement

Dynamic frameworks that evolve AI systems with technological advancements, incorporating feedback loops.

Convolutional Neural Network (CNN)

A type of deep learning model effective for processing images.

Correlation

A statistical measure of the relationship between two variables, used in feature selection.

Cross-Validation

A technique for assessing machine learning model performance by splitting data into multiple subsets.

Cyberdyne Systems

A fictional corporation from Terminator, often referenced in discussions about AI safety.

D

Data (Star Trek)

An android character exploring consciousness, emotion, and humanity, often referenced in discussions about AI personhood.

Data Augmentation

Techniques to artificially increase the diversity of training data by making modifications.

Data Automation

The process of automatically collecting, processing, and analyzing data without manual intervention.

Data Center

Facilities housing computer systems for large-scale AI operations.

Data Cleaning

Identifying and correcting errors, inconsistencies, and inaccuracies in datasets.

Data Distribution

The pattern or frequency of values in a dataset, important for understanding model behavior.

Data Engineering

Designing and building systems for collecting, storing, and analyzing data at scale.

Data Lake

A centralized repository that allows storing both structured and unstructured data at any scale.

Data Lineage

Tracking data's origins, movements, and transformations throughout its lifecycle.

Data Mining

Discovering patterns and relationships in large datasets.

Data Pipeline

A series of steps that move and transform data from various sources to analysis-ready formats, often using ETL processes.

Data Quality

The measure of data's fitness for use, including accuracy, completeness, and consistency.

Data Science

Combining statistics, programming, and domain expertise to extract insights from data.

Data Warehouse

A system for storing and managing structured data optimized for analysis and reporting.

Decision Automation

Using AI and rules-based engines to make complex business decisions without human input.

Decision Tree

A model that splits data into branches based on feature values to reach a decision.

Deep Learning

A subset of machine learning using multiple layers of neural networks.

Development Environment

The tools, frameworks, and setup used for developing and testing AI models.

Diffusion Model

A type of generative AI that creates images by gradually denoising random patterns.

Digital Worker

A software-based assistant that automates repetitive digital tasks traditionally performed by humans.

Distributed Training

Training AI models across multiple machines to handle large models or datasets.

Dynamic Variables

System components for efficient instruction updates and runtime modifications in AI systems.

Dynamic Workflow Automation

Automated workflows that adapt in response to real-time data and inputs.

E

Edge AI

AI processing performed on local devices rather than the cloud, improving speed and privacy.

Embeddings

Dense vector representations of data that capture semantic meaning, such as text or images.

Emergent Behavior

Complex behaviors arising in AI systems not explicitly programmed, through interactions of simpler rules.

Energy Consumption

The amount of electrical power required to train and run AI models.

Energy Efficiency

Measures to optimize the power consumption of AI systems while maintaining performance.

Enterprise Automation

Large-scale automation solutions implemented across an entire organization.

Ethics of AI

The study and development of moral principles and guidelines for the development and use of AI systems.

ETL (Extract, Transform, Load)

Collecting data from various sources, transforming it, and loading it into a target system.

Evals

Systematic assessments for measuring and improving AI system performance through testing.

Ex Machina

A film exploring themes of AI consciousness, testing, and the ethical implications of creating sentient machines.

Exception Handling Automation

Systems that detect and respond to exceptions in automated processes automatically.

Explainable AI

AI systems designed to make decisions that can be understood and interpreted by humans.

Exploratory Data Analysis (EDA)

Analyzing and visualizing data to understand its characteristics and patterns.

F

Fake AI

A term for systems marketed as AI but using simple rules or human labor, often for marketing.

Fauxtomation

Processes claimed to be automated but requiring significant human intervention.

Feature

An individual measurable property or characteristic used as input for machine learning models.

Feature Engineering

Selecting and transforming raw data into a format suitable for machine learning.

Feedback Loops

Systems for AI improvement, including user feedback, human-in-the-loop validation, and automated optimization.

Few-Shot Prompting

Providing a LLM with a few examples before performing a task.

Fine-tuning

Adapting a pre-trained model to a specific task by training it on a specialized dataset.

FLOPS (Floating Point Operations per Second)

A measure of computing performance, important for calculating AI model speed.

Forward Pass

Computing outputs from inputs in a neural network, used in training and inference.

Foundation Model

Large AI models trained on vast data, adaptable for various tasks.

G

Generative Adversarial Networks (GANs)

An AI architecture using two competing neural networks to improve content generation quality.

Generative AI

AI models that create new content, like text, images, or audio, often using GANs or transformers.

GPT (Generative Pre-trained Transformer)

A family of large language models that use the transformer architecture.

GPU (Graphics Processing Unit)

Hardware processors used for parallel processing in AI computations.

Gradient Descent

An optimization algorithm used to minimize error in machine learning models.

Graph Neural Networks (GNNs)

Neural networks designed to process graph-structured data.

Green AI

Developing and deploying AI systems with consideration for environmental impact and energy efficiency.

H

HAL 9000

The AI from "2001: A Space Odyssey", exemplifying concerns about AI alignment and control.

Hallucination

When an AI model generates content not based on real data, a challenge in LLM development.

Hardware Acceleration

Using specialized hardware like GPUs and TPUs to improve the speed and efficiency of AI computations.

Her

A film exploring the relationship between humans and AI, raising questions about consciousness and emotional intelligence.

High-Performance Computing (HPC)

Computing systems for handling complex calculations and large-scale AI workloads.

Human Intelligence

The diverse cognitive capabilities of humans, used as a reference in AI development, key to AGI and intelligence augmentation.

Human-in-the-Loop

Automated systems that incorporate human oversight and intervention.

Hyperautomation

Advanced use of AI, machine learning, and RPA to automate increasingly complex tasks.

Hypothesis Testing

Statistical methods for making decisions about populations based on sample data.

I

I, Robot

Isaac Asimov's stories introducing the Three Laws of Robotics, exploring AI ethics and safety.

Imputation

Replacing missing data with substituted values based on statistical methods.

Inference

Using a trained model to make predictions or decisions on new data.

Inference Optimization

Techniques to improve the speed and efficiency of model predictions.

Infrastructure as Code

Managing and provisioning AI computing infrastructure through machine-readable definition files.

Intelligence Augmentation

Using AI to enhance human capabilities rather than replace them.

Intelligent Automation (IA)

Integrating AI and automation to enhance efficiency and decision-making.

Intelligent Process Automation (IPA)

Combining RPA, AI, and other technologies to automate end-to-end processes.

IoT Automation

Automation powered by Internet of Things (IoT) devices to enable smart operations.

J

JARVIS

The AI assistant from Iron Man, an example of natural language AI interfaces.

Jupyter Notebook

An open-source tool for interactive code development, often used in AI research.

K

K-Nearest Neighbors (KNN)

A machine learning algorithm that classifies data based on the nearest examples.

Knowledge Graph

A structured representation of information that models relationships between entities.

L

Large Language Model (LLM)

A neural network trained on vast text data to understand and generate human-like language.

Latency

The time delay between input and output in AI systems.

Learning Rate

A hyperparameter controlling how much a model adjusts its weights during training.

LSTM (Long Short-Term Memory)

A neural network architecture for learning long-term dependencies in sequential data.

M

Machine Consciousness

The possibility of AI systems developing subjective experiences similar to human consciousness.

Machine Ethics

Designing AI systems to behave according to moral principles and make ethical decisions.

Machine Learning

A subset of AI that enables systems to learn and improve from experience without explicit programming.

Machine Learning Automation

Automating the process of training and improving machine learning models.

Macro Automation

Automating high-level business processes involving multiple systems and workflows.

Maintenance Contracts

Service agreements ensuring consistent AI system evaluations, updates, and performance optimizations.

Matrix, The

A film series exploring artificial reality, machine consciousness, and human-AI relationships.

Mean Square Error (MSE)

A common metric for evaluating model performance by measuring the average squared difference between predicted and actual values.

Memory Bandwidth

The rate at which data can be read or stored in memory, critical for AI model performance.

Meta-Agent

An AI agent that coordinates and manages other agents, orchestrating complex tasks.

Metadata

Data that provides information about other data, crucial for data management and governance.

Microscripting

Small, repeatable scripts used to automate specific tasks within larger workflows.

MLOps (Machine Learning Operations)

The practice of streamlining the deployment, monitoring, and maintenance of machine learning models in production. MLOps combines DevOps principles with machine learning to automate and manage the ML lifecycle, ensuring reliable and scalable AI systems.

Model

A mathematical representation trained on data to make predictions or decisions.

Model Architecture

The structure and organization of an AI model's components and layers.

Model Compression

Reducing the size and computational requirements of AI models.

Model Deployment

Making trained AI models available for use in production environments.

Model Drift

The degradation of model performance as real-world data diverges from training data.

Model Evaluation

Assessing an AI model's performance using various metrics.

Model Hosting

The infrastructure and services required to make AI models available for inference.

Model Monitoring

Continuously observing deployed AI models to track performance and other operational metrics.

Model Registry

A centralized repository for storing, versioning, and managing AI models.

Model Serving

The system for delivering AI model predictions in response to requests.

Model Versioning

Tracking and managing different versions of AI models.

Multi-Agent System

A system where multiple AI agents interact and collaborate to solve complex problems.

Multi-Modal AI

AI systems that can process and understand multiple types of input data.

N

Natural Language Processing (NLP)

AI focused on enabling computers to understand and generate human language.

Neural Network

A computing system inspired by biological neural networks, fundamental to deep learning.

Normalization

Scaling numeric data to a standard range to improve model training.

Null Hypothesis

In statistical testing, the assumption that there is no significant effect in the data.

O

Optimization

Adjusting AI model parameters to improve performance, often using gradient descent.

Orchestration

The automated configuration, coordination, and management of computer systems and applications.

Outlier

A data point that differs significantly from other observations in a dataset.

Overfitting

When a model learns training data too precisely, reducing its ability to generalize to new data.

P

P-Value

A statistical measure indicating the probability of obtaining test results as extreme as observed.

Pandas

A Python library for data manipulation and analysis, commonly used in AI and data science.

Parallel Processing

Simultaneous execution of computations across multiple processors to speed up AI workloads.

Parameter

A variable in an AI model that is learned during training to optimize performance.

Parameter-Efficient Fine-Tuning (PEFT)

Adapting large models to new tasks while updating minimal parameters.

Pattern Recognition

The ability of AI systems to identify and analyze recurring patterns in large datasets, fundamental to many machine learning applications.

Personhood

The concept of what constitutes a person, relevant to discussions about consciousness and AI rights.

Philosophy of AI

The study of questions about the nature of AI, consciousness, and the mind-machine relationship.

Pipeline Automation

Automating sequential processes where the output of one step becomes the input for the next.

Power Usage Effectiveness (PUE)

A metric measuring the energy efficiency of data centers used for AI computing.

Predictive Automation

Using AI to forecast outcomes and automatically trigger actions based on predictions.

Pretendtelligence

A sardonic term for AI systems that appear intelligent but use simple pattern matching.

Process Automation

Automating routine business processes, such as data entry or approvals.

Process Mining

Using AI and data analytics to analyze and improve processes before automating them.

Production Environment

The live system where AI models are deployed and used for real-world applications.

Prompt

An input given to an LLM to elicit a specific type of response.

Prompt Drift

When LLM outputs gradually deviate from intended results, requiring monitoring and prompt adjustments.

Prompt Engineering

Designing effective prompts to achieve desired outputs from AI models.

Prompt Template

A standardized format for creating prompts that can be reused with different inputs.

Q

Quantization

Reducing the precision of numerical models to improve efficiency.

R

R-Squared

A statistical measure indicating how well a model fits the data, representing the variance explained.

R2-D2

A Star Wars droid that demonstrates practical AI applications in robotics.

RAG (Retrieval-Augmented Generation)

Enhancing LLM responses by retrieving relevant information from external sources.

Random Sampling

Selecting a subset of data points where each has an equal probability of being chosen.

Reasoning Model

An AI model specifically designed to perform logical reasoning and inference tasks.

Recurrent Neural Network (RNN)

A neural network for working with sequential data by maintaining an internal state.

Red Queen Hypothesis

From "Through the Looking Glass", describing the need for continuous adaptation in AI.

Regression

A machine learning task that predicts continuous numerical values based on input features.

Reinforcement Learning

Machine learning where agents learn optimal actions through trial and error.

Replicant

From "Blade Runner", artificial beings that raise questions about consciousness and AI.

Responsible AI

Developing and using AI systems according to ethical principles and safety considerations.

Robotic Process Automation (RPA)

Technology that automates repetitive, rule-based tasks using software bots.

Rules-Based Automation

Automation systems that follow pre-defined rules and logic to perform tasks.

S

Safety of AI

Ensuring AI systems are reliable, controllable, and aligned with human values.

Sample Size

The number of observations in a dataset, crucial for statistical significance and model performance.

Scaling Automation

The ability of automation solutions to handle increased workloads without performance degradation.

Scaling Laws

Relationships between model size, computing, and performance that guide hardware for AI systems.

Scripting Automation

Using scripts to automate repetitive tasks like software deployment and data processing.

Self-Attention

A mechanism in transformer models that weighs the importance of different parts of the input.

Self-Healing Automation

Systems that can detect and correct errors automatically without human intervention.

Self-Improving AI

AI systems capable of enhancing their own capabilities, potentially leading to recursive self-improvement.

Self-Supervised Learning

A learning approach where the model generates its own training labels from the input data.

Semiconductor

Materials used in manufacturing processors and memory chips for AI computing.

Shot Prompting

Techniques for providing examples to LLMs, including zero-shot, one-shot, and few-shot.

Singularity

A hypothetical point when AI surpasses human intelligence, potentially leading to rapid technological growth.

Skynet

The AI system from Terminator, commonly referenced in discussions about AI safety.

Smart Automation

Automation systems that adapt to changing conditions using AI and real-time data.

Smoke and Mirrors AI

A term for AI products that create an illusion of capability through interface design or marketing.

Standard Deviation

A measure of variability in a dataset, indicating the spread of values from the mean.

Statistical Significance

The likelihood that a relationship between variables is not due to random chance.

Stochastic Parrot

A term for large language models, suggesting they repeat patterns without true understanding.

Stream Processing

Processing and analyzing data in real-time as it arrives, rather than in batches.

Strong AI

AI systems with genuine intelligence and consciousness comparable to humans, similar to AGI.

Superintelligence

A hypothetical AI system that far surpasses human cognitive capabilities.

Synthetic Data

Artificially generated data that mimics the characteristics of real-world data.

System Instructions

Directives given to an LLM that define its behavior, capabilities, and constraints.

T

Task Automation

Automating specific tasks within a larger workflow.

Technological Singularity

See singularity.

Temperature Settings

Parameters that influence AI model outputs by controlling the balance between creativity and consistency.

Test Automation

Using automation tools to run software tests and validate system functionality.

Three Laws of Robotics

Isaac Asimov's rules for robot behavior, influential in discussions about AI safety and ethics.

Throughput

The rate at which an AI system can process inputs and generate outputs, important for scaling.

Time Series

Data points collected or recorded in time order, used for forecasting and trend analysis.

Token

A unit of text (word, subword, or character) processed by language models.

TPU (Tensor Processing Unit)

Hardware accelerators designed by Google for AI computations.

Training Cost

The computational and financial resources required to train AI models.

Training Data

The dataset used to teach an AI model to perform its intended task.

Transfer Learning

Using knowledge learned in one task to improve performance on a different but related task.

Transformer

A neural network architecture using attention mechanisms, commonly used in language models.

Trigger-Based Automation

Automation initiated by specific events or conditions, such as receiving an email.

Turing Test

A test of machine intelligence where a human evaluator must distinguish between responses from a human and an AI.

U

Uncanny Valley

When human-like AI or robots create feelings of unease or revulsion.

Underfitting

When a model is too simple to capture important patterns in the training data.

Unsupervised Learning

Machine learning that finds patterns in data without pre-existing labels.

User Prompt

A request or query provided by a user to an LLM.

V

Validation Data

A separate dataset used to evaluate model performance during training to prevent overfitting.

Value Stream Automation

Automating the flow of activities in a value chain to enhance efficiency.

Vaporware AI

Announced AI products that never materialize or fail to meet promised capabilities.

Variance

A measure of variability in a dataset, calculated as the average squared deviation from the mean.

Vector Database

A database optimized for storing and querying high-dimensional vector representations of data.

Virtual Assistant

AI-powered automation that assists users by performing tasks such as scheduling.

Vision Transformer

A transformer-based model adapted for computer vision tasks.

VRAM (Video RAM)

Specialized memory on GPUs used for storing AI model parameters and intermediate computations.

W

WALL-E

A Pixar film featuring an AI robot that explores themes of environmental responsibility and machine consciousness.

Weights

Learnable parameters in a neural network that determine how input signals are processed.

White Box

A model whose internal workings are transparent and interpretable.

WOPR

The AI system from "WarGames" that learns about futility, referenced in discussions about AI learning.

Workflow Orchestration

The coordination and management of multiple automated processes.

Z

Zero-Shot Learning

The ability of AI models to handle tasks they weren't explicitly trained on.