AI Garage Glossary
Understand AI terms with real-world use cases and simple explanations.
Showing 12 of 12 terms
A machine learning model trained on large volumes of text to understand and generate natural language.
ELI5
Like a very advanced autocomplete that can hold a conversation.
Why it matters: It powers most modern AI assistants used in daily business workflows.
The practice of structuring instructions and context to get reliable, high-quality output from AI systems.
ELI5
It is how you ask the AI the right way so it gives better answers.
Why it matters: Good prompts reduce rework, errors, and hallucinations.
A chunk of text processed by an AI model. Pricing, limits, and context windows are often token-based.
ELI5
Think of tokens like tiny word pieces the AI counts and reads.
Why it matters: Token usage directly affects speed, cost, and model performance.
Numeric vector representations of text, images, or other data that capture semantic meaning for retrieval and similarity search.
ELI5
A way to turn text into numbers so AI can compare meaning, not just exact words.
Why it matters: Embeddings are the backbone of high-quality knowledge retrieval.
Additional training on domain-specific examples to adapt a base model for a particular task or style.
ELI5
Teaching an already-smart AI your specific way of working.
Why it matters: Fine-tuning can improve consistency when prompts alone are not enough.
When an AI model generates confident but incorrect or fabricated output.
ELI5
The AI sounds sure, but it made things up.
Why it matters: Unchecked hallucinations can cause trust, compliance, and business issues.
A model architecture inspired by neurons, built from layers that learn patterns from data.
ELI5
A layered pattern-finder that gets better by learning from examples.
Why it matters: Most modern AI systems are built on neural network architectures.
An Application Programming Interface lets software systems communicate with each other through structured requests.
ELI5
A software messenger that lets one app ask another app to do something.
Why it matters: APIs are how AI moves from demos into real product workflows.
A pattern where an LLM first retrieves relevant documents, then uses them to generate more grounded answers.
ELI5
The AI checks notes before answering so it is less likely to guess.
Why it matters: RAG improves factuality and makes answers easier to trace to sources.
AI systems that understand or generate across multiple data types like text, image, audio, and video.
ELI5
AI that can work with more than just words, like pictures and sound too.
Why it matters: Multimodal workflows unlock richer product experiences and automation.
The amount of information (in tokens) a model can consider in one request.
ELI5
How much the AI can keep in mind at one time.
Why it matters: Exceeding context limits causes truncation and quality loss.
The process of running a trained model to generate predictions or outputs.
ELI5
When the trained AI actually does the task for you.
Why it matters: Inference performance drives user experience and infrastructure cost.
