AI Glossary (at last)
Core Concepts
Artificial Intelligence (AI)
This is what happens when computers try to act smart. They understand language, recognize faces, and solve problems. Some even try to predict what you will eat for lunch.
Machine Learning (ML)
This is how computers learn from data without you telling them every little thing. Like a dog that teaches itself to fetch only stock market tips.
Generative AI
The creative cousin that doesn’t just analyze things. It makes stuff. Words, pictures, songs, code. You name it. It is basically the artist who also balances the books.
Large Language Models (LLMs)
These are the talkers. They have read so much text they think they are professors. Chatbots love them.
Reasoning Engines
These are the thinkers. They use logic to reach conclusions. Sometimes brilliant, sometimes like your GPS insisting that “the lake is a shortcut.”
Diffusion Models
They start with random noise and turn it into a realistic image or video. Sort of like a sculptor who begins with chaos and ends up with a masterpiece.
Artificial General Intelligence (AGI)
This is the dream (or nightmare) where machines become as smart as humans. We are not there yet, but everyone has an opinion about when we will be.
Artificial Superintelligence (ASI)
This is the version that is smarter than all of us put together. Still science fiction. Let’s hope it stays polite.
Foundational Model Builders
These are the people behind the curtain. OpenAI, Google DeepMind, Anthropic, Mistral, Meta. They make the digital brains everyone else borrows.
Hyperscalers
Amazon Web Services, Microsoft Azure, Google Cloud. They provide the computing power that keeps AI running. Think of them as the gym membership AI never cancels.
Agents and Agentic Systems
Agents
Independent digital assistants that handle tasks. They summarize, schedule, and occasionally surprise you with how well they fake enthusiasm.
Agentic Systems
A group of agents working together in harmony. Like a team of interns that never argues about lunch.
Model Context Protocol (MCP)
This helps AI tools talk to each other without turning the conversation into gibberish. Humans should take notes.
Ad Context Protocol (AdCP)
This one keeps digital advertising honest and efficient. It prevents your marketing AI from accidentally buying a herd of alpacas.
Agentic Commerce Protocol (ACP)
This teaches machines how to buy things safely and legally. Hopefully, without adding 200 toasters to the cart.
Connecting AI Systems
Application Programming Interface (API)
This is how one program talks to another. It keeps your AI and your CRM from ignoring each other’s calls.
Retrieval Augmented Generation (RAG)
This is how AI looks up facts before answering. Imagine giving your intern Google access before they start guessing.
Context Window
The amount of information an AI can remember at once. Bigger windows mean fewer awkward moments when it forgets your name.
Knowledge Graph
A giant web of facts. It connects people, places, and things so AI can sound smarter than it really is.
Vector Database
Stores information by meaning instead of exact words. It knows “film” and “movie” are basically cousins.
Data Fabric
A magic layer that connects all your company data so the AI does not wander off into the wrong spreadsheet.
Developing and Adapting AI
Fine Tuning
Teaching a trained model to focus on your specific task. It is like retraining your dog to fetch the newspaper instead of the neighbor’s shoe.
Low Code AI
Builds AI tools with little code. For people who know just enough to be dangerous.
No Code AI
Builds AI tools with no code at all. Perfect for those who think Python is still just a snake.
Synthetic Data
Fake data used to train AI when real data is too private or too scarce. Like a movie set made to look real, without the lawsuits.
Evaluation Dataset
The test every model must take before you let it near your customers.
Prompting and Context Design
Prompt Engineering
The art of asking AI the right question. Otherwise, it may write you a sonnet when you asked for a sales report.
Pre Prompt
Hidden instructions that guide AI behavior before you even say hello.
Meta Prompt
The grand director that tells all the other prompts who is in charge.
JSON Context Profile
The AI’s digital personality sheet. It defines tone, manners, and how formal it should sound when it talks about quarterly profits.
Prompt Orchestration
Coordinating multiple prompts like a maestro. When done right, it’s Beethoven. When done wrong, it’s a marching band in a thunderstorm.
Context Engineering
Giving AI the right background so it actually knows what you are asking. Think of it as briefing your assistant before the meeting.
Creative and Human Language Interfaces
Vibe Coding, Vibe Marketing, and Vibe Prompts
You tell the AI the mood you want and it delivers. “Friendly but not clingy” is now a valid design brief.
Answer Engine Optimization (AEO)
Making your website easy for AI assistants to understand and quote accurately. Basically SEO for robots with better manners.
Governance and Measurement
Guardrails
The rules that keep your AI from going rogue. They are the “do not touch” signs of technology.
AI Policy
The document that explains how your company plans to use AI responsibly. It is like the employee handbook but with fewer typos.
Model Card
A public report card showing what the model can and cannot do. Transparency for the digital age.
Audit Trail
The official record of who used the AI, when, and for what. Ideal for both compliance and finger pointing.
Evals
How you measure performance. Accuracy, reliability, safety, and cost. The AI equivalent of an annual review.
Information and Research Tools
Deep Research Tools
Apps like Perplexity, NotebookLM, and ChatGPT with browsing. They gather facts, summaries, and occasionally, confusion.
These Terms Will Continue to Evolve
AI changes faster than the plot of a soap opera. Agreeing on definitions before a strategy meeting saves everyone from arguing about what “context window” really means.
Thanks to Shelly Palmer for the above.