AI Glossary

Plain-English definitions of the AI terms you keep hearing, with Miami companies as the examples.

Part of AI Explained, MAIN's guides to one AI concept at a time.

Jump to: AEO · Agentic AI · AI Agent · AI-as-a-Service · AI Compliance · AI Governance · AI Infrastructure · AI Safety · AI Underwriting · Clinical AI · Colocation · Computer Vision · Context Window · Data Center · Deep Learning · Embedded Finance · Embeddings · Explainable AI · Fine-Tuning · Foundation Model · Generative AI · Geospatial Intelligence · GPU · Hallucination · Hyperscaler · Inference · Liquid Cooling · LLM · Machine Learning · MLOps · Moratorium · Multimodal AI · Natural Language Processing · Neocloud · Neural Network · Prompt Engineering · RAG · Responsible AI · Sovereign AI · Stablecoin · Take-or-Pay Contract · Tokenization · Vector Database · Voice AI

AEO (Answer Engine Optimization)

Answer engine optimization, or AEO, makes a business easy for AI answer tools to find, understand and recommend. It covers how a company appears when people ask ChatGPT, Gemini or Perplexity a question.

Miami examples: LinguaScale, ReadableIQ, Dr. Eva Wolf.

Agentic AI

Agentic AI describes systems built from AI agents that plan and carry out multi-step work. The agents act with limited human input, and a person can step in when needed.

Miami examples: QAD Redzone, Strata.ai, Maestro.

Read more: What Is an AI Agent?

AI Agent

An AI agent is software that takes actions toward a goal, not just answers questions. It can use tools, follow a sequence of steps and hand work to a person when it should.

Miami examples: Clave, CRE Agents, Maestro.

Read more: What Is an AI Agent?

AI-as-a-Service (AIaaS)

AI-as-a-service means renting AI capabilities from a provider over the internet instead of building them in-house. The customer pays for access and the provider runs the models.

Miami examples: Lety.ai.

Read more: How Do AI Companies Make Money?

AI Compliance (AI compliance software)

AI compliance covers two jobs. One is using AI to do compliance work, such as collecting evidence. The other is keeping AI itself inside the rules.

Miami examples: Comp AI, SOVAIR.AI.

Read more: What Is AI Compliance Software?

AI Governance

AI governance is the set of rules, roles and records a company uses to control how AI is used. It answers who may use which tools, on what data, and who is accountable.

Miami examples: Arpia, PrivacyPal, Work-Smart AI.

AI Infrastructure

AI infrastructure is the power, buildings, computing and routing that AI runs on. Each layer has its own builders and its own bottlenecks.

Miami examples: Exowatt, TrustedRouter, BlockRun.

Read more: What Is AI Infrastructure?

AI Safety

AI safety is the work of keeping AI systems from causing harm. It covers misuse, unintended behavior and the limits placed on what a system can do.

Miami examples: Molt AI, SOVAIR.AI.

AI Underwriting

AI underwriting uses models to assess credit or risk and to decide on loans or coverage. It is often much faster than manual review.

Miami examples: QUASH.ai, Split Pay.

Clinical AI

Clinical AI is software that supports medical decisions, such as answering clinicians' questions or reading images. It is used in care settings and faces safety and regulatory review.

Miami examples: OpenEvidence, Ferrum Health.

Colocation

Colocation lets a tenant place its own equipment in someone else's data center. The tenant pays for space, power and cooling.

Miami examples: Equinix and Flexential.

Computer Vision

Computer vision lets software interpret images and video. It can spot a flaw on a factory line or count the walls and floors on a construction drawing.

Miami examples: Togal AI, Brightpoint AI.

Context Window

A context window is the amount of text a model can consider at one time, measured in tokens. A longer window lets it work with longer documents.

Miami examples: Subquadratic. Subquadratic describes its model as built for long-context tasks.

Data Center

A data center is a building that houses servers, power and cooling. AI data centers are built for denser, hotter GPU equipment than older sites.

Miami examples: Hut 8, Metrobloks MIA-A1, Iron Mountain MIA-1.

Deep Learning

Deep learning is a type of machine learning that uses neural networks with many layers. It powers image recognition, speech tools and large language models.

Embedded Finance

Embedded finance builds payments, lending or accounts into non-financial software. Customers use them without leaving the app.

Miami examples: Payabli, Realfinity. Payabli provides embedded payments infrastructure, and Realfinity describes itself as an embedded lending platform.

Embeddings

Embeddings turn text, images or other data into lists of numbers that capture meaning. Similar items get similar numbers, which makes search by meaning possible.

Explainable AI

Explainable AI refers to methods that show why a model reached a result, so people can check and trust it.

Fine-Tuning

Fine-tuning takes an existing AI model and trains it further on a narrower set of data. The aim is better results for one job or industry.

Foundation Model

A foundation model is a large AI model trained on broad data and adapted to many tasks. Large language models are one kind.

Generative AI

Generative AI creates new content, such as text, images, code or audio, from patterns it learned in training data. Chatbots and image tools are common examples.

Geospatial Intelligence

Geospatial intelligence analyzes satellite, aerial and location data to understand what is happening on the ground. Defense and infrastructure teams use it.

Miami examples: Space-Eyes.

GPU (Graphics Processing Unit)

A GPU is a chip built to run many calculations at once. That makes it the standard hardware for training and running AI models.

Miami examples: Hut 8, Hydra Host, Optidata.

Hallucination

A hallucination is an AI answer that sounds confident but is wrong or invented. It happens because models predict likely text, not verified facts.

Hyperscaler

A hyperscaler is a very large cloud provider, such as Amazon, Microsoft or Google. Hyperscalers also rent capacity from neoclouds when they cannot build fast enough.

In South Florida, Palm Beach County commissioners rejected Project Tango, a hyperscale data center proposal in Loxahatchee, 5-1 on July 15. See MAIN's tracker.

Read more: What Is a Neocloud?

Inference

Inference is the stage where a trained AI model produces answers for users. It runs every time someone sends a request, while training happens up front.

Miami examples: Metrobloks MIA-A1, Optidata. MAIN's report describes MIA-A1 as designed for high-density AI inference.

Liquid Cooling

Liquid cooling removes heat from AI chips with fluid, not only air. It matters because dense GPU racks run hotter than older servers.

Miami examples: Metrobloks MIA-A1. MAIN's report says Metrobloks cools MIA-A1 by air, liquid or a hybrid of both.

LLM (Large Language Model)

A large language model is an AI system trained on vast amounts of text to read and write language. Chatbots such as ChatGPT and Claude run on LLMs.

Miami examples: Subquadratic, BlockRun, TrustedRouter. Subquadratic builds a model for long-context tasks. BlockRun and TrustedRouter route requests between models.

Machine Learning

Machine learning is a way of building software that learns patterns from data instead of following hand-written rules. Most modern AI is built this way.

MLOps (Machine Learning Operations)

MLOps is the practice of deploying, monitoring and updating machine learning models in production. It does for AI models what DevOps does for software.

Moratorium

A moratorium is a temporary pause on new approvals. Local governments have used them to halt new data center projects while they review the rules.

As of August 24, 19 Florida jurisdictions had paused or banned new projects, according to MAIN's tracker.

Multimodal AI

Multimodal AI works with more than one type of input, such as text, images and audio, in the same system.

Natural Language Processing (NLP)

Natural language processing is the field of AI that reads, interprets and writes human language, in text or speech. Chatbots and voice assistants depend on it.

Miami examples: ClaireAI, Xina AI.

Neocloud (also written neo cloud)

A neocloud is a cloud provider built around GPU computing for AI, not the broad menu big clouds sell. Customers often sign multi-year contracts for fixed capacity.

Miami examples: Optidata.

Read more: What Is a Neocloud?

Neural Network

A neural network is a computing system loosely modeled on the brain. Layers of connected nodes pass signals along and adjust as the system learns.

Prompt Engineering

Prompt engineering is writing instructions that get an AI model to produce useful output. It covers wording, examples and structure.

Miami examples: Lucci AI. Lucci AI pitches AI systems over prompt craft.

RAG (Retrieval-Augmented Generation)

Retrieval-augmented generation lets an AI model look up documents before it answers. The model then writes its reply from what it found, which helps keep answers tied to real sources.

Responsible AI

Responsible AI is the practice of building and using AI with attention to fairness, privacy, safety and accountability.

Sovereign AI

Sovereign AI keeps AI systems and data under one organization's or country's control, not in a shared public cloud. Air-gapped setups go further and run with no outside network connection.

Miami examples: SOVAIR.AI.

Stablecoin

A stablecoin is a digital token designed to hold a steady value, usually pegged to a currency like the dollar. Payments firms use them to move money across borders.

Miami examples: Jeeves, OpenFX, Global Ledger.

Take-or-Pay Contract

A take-or-pay contract commits a customer to pay for capacity whether or not it uses it. Neoclouds rely on them to support the debt that finances their GPUs.

Read more: What Is a Neocloud?

Tokenization

In language models, tokenization splits text into small pieces called tokens. Models read and bill by tokens, so token counts drive cost. In finance the word means something else: turning an asset into a digital token.

Miami examples: Sibos Miami poll on AI token costs.

Vector Database

A vector database stores embeddings and finds the closest matches quickly. AI tools use it to look up relevant documents, as in retrieval-augmented generation.

Voice AI

Voice AI lets software hold spoken conversations, such as answering phones or taking orders. It combines speech recognition, language understanding and speech generation.

Miami examples: Vocaly AI, Xina AI, Heytruffle.

See a term missing? Tell the MAIN team and we will add it.