Info
This page is a living glossary and gets revised whenever terms, tools, and technologies change.
Introduction – AI, ML, DL, and GenAI
As artificial intelligence continues to spread throughout every industry, it’s easy to get caught up in the lingo - AI, ML, DL, GenAI. What do they all mean, and how are they different? This piece demystifies the key terms and concepts so you can have a clear understanding of the world of AI. Whether you’re new to the industry or simply need a refresher on the basics, this glossary-driven primer will bring you up to date on the terminology of AI and machine learning.
Looking for courses and certifications instead of definitions? Check the AI Learning Guide.
Table of Contents
AI Glossary
Artificial Intelligence (AI)
The broader field of creating intelligent agents, aiming to mimic human intelligence.
Machine Learning (ML)
A subset of AI focused on teaching computers to learn from data without explicit programming.
Deep Learning (DL)
A subset of ML using artificial neural networks to analyze complex patterns in data.
Generative AI (Gen AI)
A subfield focused on creating new content, such as text, images, or code, based on learned patterns.
Agentic AI
Agentic AI refers to AI systems that can plan, take multi-step actions, and use tools to complete a task with limited step-by-step human direction, rather than just responding to a single prompt. An agent might break a goal into sub-tasks, call external tools or APIs, check its own results, and adjust its approach along the way. This is different from a standard chatbot interaction, where the model responds once and stops. Agentic AI is behind tools like autonomous coding assistants, research agents, and multi-step workflow automation.
MCP – Model Context Protocol
Model Context Protocol (MCP) is an open standard that lets AI models connect to external tools, data sources, and services in a consistent way, rather than requiring a custom integration for every tool. Think of it as a common connector: once a service exposes an MCP server, any MCP-compatible AI assistant can use it, whether that’s reading files, querying a database, or calling a company’s internal APIs. MCP has become a key building block for agentic AI systems that need to interact with the real world beyond just generating text.
Subagents / Multi-Agent Systems
A subagent is a smaller, focused AI agent that a main “orchestrator” agent can delegate a specific part of a task to, then use the result to continue the larger job. Multi-agent systems break complex work into parallel or sequential pieces handled by different agents, each often given a narrower role, context, or toolset. This pattern is increasingly used in agentic coding and research tools to keep individual tasks focused and easier to verify.
LAB (Large-scale Alignment for chatBots)
LAB is a research approach that focuses on making chatbot behavior harmonize with human values, goals, and expectations, at scale. While more powerful chatbots are possible, it’s no longer enough to just train them on large data sets. They need to be able to respond usefully, safely, and in line with user intent. LAB consists of model refinements through methodologies such as Reinforcement Learning from Human Feedback (RLHF), instruction tuning, and preference modeling, to ensure responses are well-suited, accurate, and in line with what humans indeed desire.
At a large scale, this alignment becomes more complex, you’re not just adjusting responses for one use case, but ensuring consistent and ethical behavior across many domains, languages, and cultures. LAB plays a key role in building trustworthy AI assistants, especially those used in customer support, education, healthcare, and other high-impact areas.
LLM – Large Language Models
LLMs are extremely sophisticated AI models that have been trained on massive amounts of text data to understand and generate human-like language. These models learn to identify patterns, grammar, facts, reasoning, and context from books, websites, articles, and more. After being trained, they can perform a wide range of language tasks such as answering questions, writing code, summarizing text, translating between languages, and holding conversations, all without direct programming for each task.
“Large” in LLM refers to both the model size (millions or even billions of parameters) and the size of the data they are trained on. Because of this scale, LLMs can generalize well across domains and adapt to a wide range of different user needs, making them a key part of modern AI applications like chatbots, virtual assistants, and content generation software.
Gen AI – Generative Artificial Intelligence
Generative AI, or Gen AI, is one of the subcategories of artificial intelligence that focuses on creating new content, including text, pictures, sound, video, or even programming code, instead of just analyzing or sorting existing data. Powered by large models trained on enormous datasets, Gen AI creates new output based on learned patterns. For example, it can generate a story, design a logo, write a song, or generate photorealistic images from a simple text prompt.
What sets Gen AI apart is the ability to mimic creativity and provide output that appears to be human-made. It’s being used across every industry, marketing, design, software development, education, and more. While it holds great promise, it also raises real questions around originality, bias, and usage, especially as its output becomes harder to distinguish from human-made work.
AI Inference
AI Inference is the process of using a machine learning model that has been trained on data to make predictions or decisions on new, unseen input. After a model has been trained on large datasets, inference is the deployment stage, where the model applies what it learned to real-world inputs. For example, an AI model trained to classify objects in images performs inference when it looks at a new image and classifies the objects in it. Inference can happen in various settings, such as cloud servers, edge devices, or even smartphones, depending on the use case and performance requirements. It’s a critical step in AI implementation, since the usefulness of a model is validated by how well and efficiently it processes new data and produces actionable results.
AI Ethics
AI Ethics is the study and practice of ensuring artificial intelligence systems are used and created in ways that are fair, transparent, and responsible. It deals with making sure AI doesn’t cause harm, discriminate unfairly, or make decisions that can’t be understood or explained. Since AI models are trained on large datasets, which are often full of human bias or gaps in information, there’s a real risk that those imperfections get passed on to the systems built from them. AI ethics is the practice of managing how we design, test, and deploy AI, with human values and social impact in mind.
Deepfakes
Deepfakes are media, predominantly images and videos, altered by AI to depict people, activities, or events in realistic but false ways. Advances in generative AI, such as neural networks and deep learning, have made it far easier to create manipulated media that looks convincingly real. As the technology has become more accessible, even people with little technical knowledge can now produce realistic but misleading content. This is especially concerning for trust, security, and disinformation, since deepfakes can be used to shape opinion, spread false information, or damage reputations. The widespread availability of deepfake tools remains an ongoing challenge to the authenticity of digital media, with real risks for individuals, institutions, and society as a whole.
AI Knowledge vs Skill
Knowledge in AI:
- Knowledge represents the information and data that an AI system has access to, including facts, rules, and concepts about the world or a specific domain. It’s the foundational material that the AI uses to make decisions, predictions, and perform tasks.
- AI systems accumulate knowledge through data training (e.g., large datasets of text, images, or structured data) and can retrieve and apply this knowledge when required.
- Knowledge can be encoded as rules (in rule-based systems) or derived from patterns in data (in machine learning systems).
Examples:
- A language model is trained on vast amounts of text data, giving it knowledge about grammar, facts, and even context.
- A medical diagnostic AI might have knowledge about symptoms, diseases, and treatments stored in its database or learned from medical literature.
Skill in AI:
- Skill refers to the ability of an AI to apply knowledge effectively to perform tasks or solve problems. It involves leveraging knowledge in a way that demonstrates proficiency in a specific task or set of tasks.
- AI develops skills through training, experience, and fine-tuning. Skill in AI can be thought of as the execution part, where the system demonstrates its capability to solve real-world problems using the knowledge it possesses.
Examples:
- A self-driving car’s ability to navigate roads by understanding traffic rules and reacting to real-time conditions.
- A machine learning model’s skill to recognize objects in images, using knowledge about how certain objects appear.
Relationship Between Knowledge and Skill in AI:
- Knowledge is the foundation, while skill is the application of that knowledge.
- An AI system may have vast knowledge but limited skill if it cannot effectively use the information to solve tasks.
- Conversely, a highly skilled AI must be backed by a strong knowledge base to perform its tasks accurately and consistently.
AI Tools and Products
InstructLab
InstructLab is an open-source project, co-created by IBM and Red Hat, that makes it easier to align large language models with user intent, opening up possibilities for innovative AI applications.
Podman AI Lab
Podman AI Lab is Red Hat’s answer to simplifying the AI development process. This extension provides a local environment with essential open-source tools and curated recipes to guide you through building AI solutions.
Example Use Cases
- Running a Jupyter Notebook preloaded with TensorFlow or PyTorch
- Testing open-source LLMs locally in a container
- Prototyping AI-powered apps with minimal setup
- Learning or teaching AI/ML concepts using clean, reproducible environments
Podman AI Lab vs Docker
- Podman is rootless by default, so it’s more secure for personal or educational setups.
- No daemon needed, Podman doesn’t require a background service like Docker, making it lightweight and easier to manage.
- Fully open source, with better integration into Linux-native environments (especially for Fedora, RHEL, etc.)
- With Podman Desktop + AI Lab, the experience is user-friendly and great for quick local AI testing, without needing the cloud or expensive GPU clusters.
Red Hat OpenShift AI
Red Hat OpenShift AI is a platform for deploying and scaling AI applications across hybrid cloud environments. Built on open-source technologies, it offers a trusted foundation for teams to experiment, serve models, and deliver AI-driven apps.
Claude Code and MCP
For teams already working with Kubernetes, Ansible, or other automation tooling, Claude Code is worth knowing about, it’s an agentic coding tool that runs in the terminal and can use MCP servers to connect to your existing infrastructure and tools. See the AI Learning Guide for more on getting started with Claude and MCP.
Some Facts
Warning
This section is updated periodically as models move fast, treat any specific model name here as a snapshot in time, not a permanent reference.
- Llama 4 (Scout and Maverick) is Meta’s current generation of openly available large language models, and the successor to the Llama 2 and Llama 3 families referenced in older AI articles.