AI Ethics:  Issues that AI stakeholders such as engineers and government officials must consider to ensure that the technology is developed and used responsibly. This means adopting and implementing systems that support a safe, secure, unbiased, and environmentally friendly approach to artificial intelligence.

Algorithm: A set of mathematical rules or instructions a computer follows to solve a problem or complete a task.

Agentic: Autonomous systems designed to achieve complex, long-term goals by proactively planning, taking multi-step actions, and adapting to new information with minimal human oversight.

Alignment: The process of ensuring an AI’s goals and behaviors match human values and safety standards.

Artificial General Intelligence (AGI): A theoretical version of AI that can understand, learn, and apply its intelligence to any intellectual task a human can do. We aren’t there yet.

Artificial Intelligence (AI): The broad field of creating machines capable of performing tasks that typically require human intelligence, such as reasoning, learning, and problem-solving.

Asset Performance Management (APM): A strategy combining artificial intelligence, IoT, and data analytics to monitor, predict, and optimize the health, reliability, and lifespan of physical assets. It uses machine learning to identify risks, reduce unplanned downtime, and enable proactive maintenance by analyzing real-time data to improve operational efficiency.

Autonomous/Autonomy: The capacity of AI systems to operate, make decisions, and achieve specific goals in complex environments with minimal or no direct human intervention.

Bias: When an AI application produces prejudiced results because the data it was trained on contained human biases or unfair stereotypes.

Black Box: A term used to describe AI systems (especially deep learning) where the internal decision-making process is so complex that even the creators can’t fully explain how it reached a specific result.

Computational Knowledge Graph (CKG): A structured, graph-based data model that represents real-world entities (people, places, concepts) as nodes and their semantic relationships as directed edges, designed specifically to be queried, computed, and understood by artificial intelligence systems.

Computer Vision: An interdisciplinary field of science and technology that focuses on how computers can gain understanding from images and videos. Computer vision allows AI engineers to automate activities that the human visual system typically performs.

Condition-Based Monitoring: A predictive maintenance technique that continuously monitors the condition of equipment or assets using sensor-derived data that relates information about real-time conditions.

Data Mining: The process of closely examining data to identify patterns and glean insights. Data mining is a central aspect of data analytics; the insights you find during the mining process will inform your business recommendations.

Deep Learning: A more advanced type of machine learning inspired by the structure of the human brain. It uses multiple layers of “neurons” to process complex data like photos or speech.

Emergent Behavior: (also called emergence) When an AI system shows unpredictable or unintended capabilities that only occur when individual parts interact as a wider whole.

Fine Tuning: The process of taking a pre-trained AI model and training it further on a smaller, highly specific dataset to adjust its behavior or improve its performance in a particular field.

Generative AI: AI specifically designed to create new content, such as text, images, music, or video, rather than just analyzing existing data.

Guardrails: Mechanisms and frameworks designed to ensure that AI systems operate within ethical, legal, and technical boundaries. They prevent AI from causing harm, making biased decisions, or being misused.

Hallucination: Instances where an AI model generates false, misleading, or entirely fabricated information that it presents as fact.

Inference: The process of the AI actually using what it has learned. When a chatbot is asked a question and it answers, it is “inferring” the best response based on its training.

Large Language Model (LLM): A type of AI trained on vast amounts of text to understand, generate, and predict human language (e.g., Gemini, ChatGPT, Claude).

Knowledge Fabric: An advanced, interconnected layer of organizational memory that combines, structures, and continuously updates data from various sources—such as documents, conversations, workflows, and databases—into a single, semantically connected framework.

Machine Learning (ML): A subset of AI where computers use statistics to find patterns in data and “learn” how to perform a task without being explicitly programmed for every step.

Multimodal Model: An AI system capable of understanding and processing multiple types of information (like text, audio, images, and video) simultaneously.

Natural Language Processing (NLP): The branch of AI that focuses on the interaction between computers and human language, allowing machines to “read” and “speak.” 

Neural Network: The digital architecture used in deep learning. It consists of interconnected nodes (neurons) that pass information to one another to reach a conclusion.

Normal Behavior Modeling (NBM): A machine learning approach used for process, system, and equipment management that involves training a model to understand and quantify the “normal” operating state of a system based on historical data. Once the baseline for normal operation is established, the model continuously monitors live data to detect anomalies, deviations, or impending failures without needing prior exposure to specific failure modes.

Parameters: The internal variables AI adjusts during training, i.e., “knobs” the model turns to get the right answer; more parameters often correlate with higher complexity.

Pattern Recognition: The use of computer algorithms to analyze, detect, and label regularities in data. This informs how the data are classified into different categories.

Predictive Maintenance: A proactive strategy that uses sensors, machine learning, and data analytics to monitor asset/equipment conditions in real time and forecast potential failures before they occur.

Prescriptive Maintenance: An advanced, data-driven strategy that uses artificial intelligence, machine learning, and IoT sensors to not only predict equipment failures but also recommend, and sometimes execute, specific corrective actions. It determines the root cause of potential issues, advising on the exact, optimal maintenance to prevent failures, reduce downtime, and increase asset lifespan.

Prompt: The instruction or question you provide to an AI to elicit a specific response.

Reinforcement Learning: A type of machine learning that learns by interacting with its environment and receiving positive reinforcement for correct predictions and negative reinforcement for incorrect predictions. Common algorithms are temporal difference, deep adversarial networks, and Q-learning.

Risk or Anomaly Score: Numerical value derived by aggregating all feature output values from an NBM model through statistical analysis. The risk score determines whether action is required on the part of maintenance staff.

 

Structured Data: Data that is defined and searchable. Structured data is formatted data; for example, data organized into rows and columns. Structured data is typically easier to analyze than unstructured data because of its tidy formatting. This includes data like phone numbers, dates, or product SKUs.

Supervised Learning: A type of AI machine learning that trains algorithms using labeled input-output pairs to learn mappings, allowing the model to predict outcomes for new, unseen data. By analyzing categorized data (labeled data), the model learns to identify patterns, classify data, or forecast values.

Tag: The specific name assigned to a unique data element in an input data set to a neural network (e.g., Temp Pump 37A).

 

Token: The basic unit of data processed by AI models. In text, tokens can be whole words or just parts of words.

Training Data: The massive dataset (text, images, or code) used to “teach” an AI model. The quality of the AI’s output depends heavily on the quality of this data.

Transformer Model: A transformer model is a type of neural network architecture that excels at processing sequential data, most prominently associated with large language models (LLMs). Transformer models have also achieved excellent performance in other fields of artificial intelligence, such as computer vision, speech recognition, and time series forecasting.

Unstructured Data: Data that is not organized in any apparent way. In order to analyze unstructured data, it is typically necessary to implement some type of structured organization.

Unsupervised Learning: A machine learning technique that analyzes unlabeled, unstructured datasets to discover hidden patterns, clusters, and relationships without human guidance. Unlike supervised learning, the algorithm acts independently to categorize data based on similarities or reduce dimensionality, making it ideal for exploratory data analysis.