Agents & protocols
AI Agent
An AI system built from an LLM, a set of tools it can call, and enough autonomy to plan and take a sequence of actions toward a goal - as distinct from a system that only answers a single prompt.
agentagentic AI
Intelligence · AI Encyclopedia
The AI Encyclopedia
19 concepts, architectures, techniques and protocols that shape how AI actually works - each one defined in plain English and precisely, sourced, and linked to the models, companies, research and regulations it connects to. A term here is never the end of the trail.
A
Core concepts
AI Hallucination
When a language model generates false or fabricated information - a made-up citation, a wrong fact, a nonexistent feature - stated with the same fluent confidence as a correct answer.
hallucinationreliability
C
Core concepts
Context Window
The maximum amount of text (measured in tokens) a model can 'see' at once - the prior conversation, a document, an instruction - when generating its next output.
context windowtokens
D
Architectures & models
Diffusion Model
A generative model that learns to create data - typically images - by learning to reverse a process that gradually adds noise to training data, starting from pure noise and denoising it step by step into a coherent output.
diffusionimage generation
E
Core concepts
Embedding
A numeric vector that represents the meaning of a piece of data - a word, sentence, image - such that similar meanings sit close together in vector space.
embeddingsvectors
F
UAE & regional
Falcon (LLM family, TII)
A family of open-weight large language models developed by Abu Dhabi's Technology Innovation Institute (TII), released from 2023 onward, including one of the largest openly available LLMs at the time of its release.
FalconTII
Techniques & methods
Fine-Tuning
Further training a pre-trained model on a smaller, labelled, task- or domain-specific dataset so it adapts its internal weights to that task - as opposed to prompting, which leaves the model unchanged.
fine-tuningtraining
Architectures & models
Foundation Model
A model trained on broad data at scale, usually via self-supervision, that can be adapted - by fine-tuning, prompting or other means - to a wide range of downstream tasks, rather than built for one task alone.
foundation modelpre-training
L
Core concepts
Large Language Model (LLM)
A deep-learning model, usually built on the transformer architecture, trained on huge amounts of text to predict the next token and, in doing so, learn grammar, facts, reasoning patterns and style.
LLMgenerative AI
M
Architectures & models
Mixture of Experts (MoE)
A model architecture that splits computation across many specialised sub-networks ('experts') and, for each input, routes it to only a few of them - giving a model the knowledge capacity of a huge network at the compute cost of a much smaller one.
MoEarchitecture
Agents & protocols
Model Context Protocol (MCP)
An open standard, created by Anthropic and released in November 2024, for connecting AI applications to external data sources and tools through one common protocol instead of a custom integration for each pairing.
MCPprotocol
Architectures & models
Multimodal AI
AI models that process and reason across more than one type of input or output - text, images, audio, video - within a single system, rather than bolting separate single-purpose models together.
multimodalvision
P
Techniques & methods
Prompt Engineering
The practice of designing the instructions and examples given to a language model to reliably get the output you want, without changing the model itself.
promptingzero-shot
R
Techniques & methods
Reinforcement Learning from Human Feedback (RLHF)
A post-training method that adjusts a pre-trained model's behaviour using human preference judgments, rather than more raw text - the technique that turned raw language models into helpful assistants.
RLHFalignment
Safety, governance & policy
Responsible AI & AI Governance
The organisational practice of overseeing AI systems for safety, fairness, accountability and regulatory alignment across their lifecycle - and the frameworks, like NIST's AI Risk Management Framework, organisations use to operationalise it.
governanceresponsible AI
Techniques & methods
Retrieval-Augmented Generation (RAG)
A technique that pairs a language model with a search step over an external knowledge source, so the model's answer is generated from retrieved, citable text rather than from memory alone.
RAGretrieval
T
Architectures & models
Transformer (Architecture)
The neural network architecture, introduced in 2017, that processes a whole sequence at once using self-attention instead of reading it word by word - and that underlies almost every modern LLM.
architectureattention
U
UAE & regional
UAE National Strategy for Artificial Intelligence 2031
The UAE Government's national framework, announced in October 2017, setting out its goal to become a global leader in artificial intelligence by 2031 through AI adoption across government, healthcare, education, transport and energy.
UAEnational strategy
V
Infrastructure & tooling
Vector Database
A database purpose-built to store embeddings (vectors) and answer 'find me the most similar items' queries fast, using approximate nearest-neighbour search - the storage layer under most RAG systems.
vector databaseembeddings
Today
Term of the day · 3 October 2026
Large Language Model (LLM)
A deep-learning model, usually built on the transformer architecture, trained on huge amounts of text to predict the next token and, in doing so, learn grammar, facts, reasoning patterns and style.
Beyond the encyclopedia
The Encyclopedia explains the concepts. For the systems built on top of them, see Nakoda’s wider coverage of the UAE and global AI landscape.