Glossary
The terms adopted by the encyclopedia, with their Arabic equivalent and a short definition. Each term is used with the same meaning in every entry.
Glossary
المسرد- Agentic AIالذكاء الاصطناعي الوكيلي
- Systems composed of multiple coordinated AI agents that break down, delegate and collaborate on tasks to reach complex goals over long periods with limited human supervision; distinct from a single AI agent.
See also: AI agent، Large language model
Source: What agentic AI is and does
- AI agentوكيل الذكاء الاصطناعي
- Software that can reason about user inputs in order to plan and execute actions on behalf of the user, often using external tools.
Elsewhere: Arabic Wikipedia: وكيل ذكاء اصطناعي؛ SDAIA: وكيل
See also: Agentic AI
Source: Machine Learning Glossary
- AI alignmentالمواءمة
- The process of encoding human values and goals into AI models to make them as helpful, safe and reliable as possible.
See also: Reinforcement learning from human feedback، AI guardrails، Interpretability
Source: What Is AI Alignment? | IBM
- AI guardrailsحواجز الحماية
- The safeguards that keep AI systems operating safely, responsibly and within defined boundaries.
See also: AI alignment، Prompt injection
Source: What Are AI Guardrails? | IBM
- AI literacyالإلمام بالذكاء الاصطناعي
- Under the EU AI Act, the skills, knowledge and understanding that allow providers, deployers and affected persons to make an informed deployment of AI systems and to gain awareness of their opportunities, risks and possible harms.
Elsewhere: Arabic Wikipedia: الإلمام بالذكاء الاصطناعي
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- AI red-teamingاختبارات الفريق الأحمر
- A structured testing effort to find flaws and vulnerabilities in an AI system, often in a controlled environment and in collaboration with its developers, adopting adversarial methods.
Elsewhere: Arabic Wikipedia: الفريق الأحمر
See also: Benchmark، Prompt injection، AI guardrails
- AI regulatory sandboxالبيئة التجريبية التشريعية
- Under the EU AI Act, a controlled framework set up by a competent authority that lets providers develop, train, validate and test innovative AI systems for a limited time under a sandbox plan and regulatory supervision.
See also: Provider، High-risk AI system
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- AI systemنظام الذكاء الاصطناعي
- A machine-based system that infers from its inputs how to generate outputs such as predictions, content, recommendations or decisions that can influence physical or virtual environments; such systems vary in autonomy and adaptiveness after deployment.
See also: Artificial intelligence، Provider، High-risk AI system
Source: Updates to the OECD's definition of an AI system explained
- Application programming interface APIواجهة برمجة التطبيقات
- A contract that enables software to access data or invoke functionality made available by other software; many models are offered to developers this way without releasing their weights.
Elsewhere: Arabic Wikipedia: واجهة برمجة التطبيقات
See also: Open-weight model
Source: What Is an API (Application Programming Interface)? | IBM
- Artificial general intelligence AGIالذكاء الاصطناعي العام
- A hypothetical stage of AI development in which a system could match or exceed human cognitive abilities on any task; no such system exists and there is no agreed definition.
Elsewhere: Arabic Wikipedia: ذكاء عام اصطناعي
See also: Artificial intelligence، AI alignment
Source: What is Artificial General Intelligence (AGI)? | IBM
- Artificial intelligence AIالذكاء الاصطناعي
- A non-human program or model that can solve sophisticated tasks, such as translating text or identifying diseases from images.
Elsewhere: Arabic Wikipedia: ذكاء اصطناعي؛ Dubai Future Foundation: الذكاء الاصطناعي
See also: Machine learning، AI system
- Attention mechanismآلية الانتباه
- A mechanism used in a neural network that indicates the importance of a particular word or part of a word relative to the rest of the input.
Elsewhere: Arabic Wikipedia: انتباه
See also: Transformer
Source: Machine Learning Glossary
- Automatic speech recognition ASRالتعرف التلقائي على الكلام
- A capability that enables a program to process human speech into a written format; also called speech-to-text.
Elsewhere: Arabic Wikipedia: تعرف الكلام
See also: Text to speech، Natural language processing
- Benchmarkاختبار الأداء
- A standardized framework for assessing model performance, comprising sample data, tasks that test specific skills, performance metrics and a scoring mechanism.
See also: Large language model، AI red-teaming
Source: What Are LLM Benchmarks? | IBM
- Bias (ethics/fairness)التحيز
- Stereotyping, prejudice or favoritism towards some things, people or groups over others; it can enter a model's data, design or outputs.
Elsewhere: Arabic Wikipedia: تحيز خوارزمي
See also: Training data، Synthetic data
Source: Machine Learning Glossary
- Chain-of-thought prompting CoTسلسلة الأفكار
- A prompt engineering technique that encourages a large language model to explain its reasoning step by step.
See also: Prompt engineering، Inference
Source: Machine Learning Glossary
- Computeالقدرة الحاسوبية
- The computational resources used by a model or system, such as processing power, memory and storage.
See also: Graphics processing unit، Tensor Processing Unit، Inference
Source: Machine Learning Glossary
- Computer vision CVالرؤية الحاسوبية
- A subfield of AI that equips machines to process, analyze and interpret visual inputs such as images and videos, using machine learning to derive meaningful information.
Elsewhere: Arabic Wikipedia: رؤية حاسوبية
See also: Optical character recognition، Multimodal AI
Source: What Is Computer Vision? | IBM
- Context windowنافذة السياق
- The number of tokens a model can process in a given prompt; a larger window lets the model use more information.
Source: Machine Learning Glossary
- Data poisoningتسميم البيانات
- A type of cyberattack where threat actors manipulate or corrupt the training data used to develop AI and machine learning models.
See also: Training data، Prompt injection
Source: What Is Data Poisoning? | IBM
- Deep fakeالتزييف العميق
- Under the EU AI Act, AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful.
Elsewhere: Arabic Wikipedia: تزييف عميق
See also: Watermarking، Generative AI
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- Deep learning DLالتعلم العميق
- A subset of machine learning driven by multilayered neural networks whose design is inspired by the structure of the human brain.
Elsewhere: Arabic Wikipedia: تعلم عميق؛ Dubai Future Foundation: التعلم العميق
See also: Machine learning، Neural network
Source: What Is Deep Learning? | IBM
- Deployerالمشغِّل
- Under the EU AI Act, a natural or legal person, public authority or other body using an AI system under its authority, except in the course of a personal non-professional activity.
See also: Provider، High-risk AI system
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- Diffusion modelنموذج الانتشار
- A generative model used primarily for image generation, trained to gradually add noise to samples and then reverse the process to produce high-quality images.
Elsewhere: Arabic Wikipedia: نموذج الانتشار
See also: Generative AI، Deep fake
Source: What are Diffusion Models? | IBM
- Distillation (knowledge distillation)التقطير
- Reducing a larger teacher model into a smaller student model that emulates its predictions, gaining speed and lower memory and energy use at some cost in quality.
See also: Quantization، Parameters
- Edge AIالذكاء الاصطناعي على الحافة
- Deploying AI algorithms and models directly on local edge devices such as sensors or IoT devices, enabling real-time processing without constant reliance on the cloud.
See also: Federated learning، Quantization
Source: What Is Edge AI? | IBM
- Embedding (vector embedding)التضمين
- A numerical representation of data points (such as words or images) as arrays of numbers that machine learning models can process, with similar items placed close together.
See also: Token، Retrieval-augmented generation
Source: What is Vector Embedding? | IBM
- Explainable AI XAIالذكاء الاصطناعي القابل للشرح
- A set of processes and methods that allows human users to comprehend and trust the results and output of machine learning algorithms; it goes beyond interpretability to show how the AI arrived at a result.
Elsewhere: Arabic Wikipedia: ذكاء اصطناعي قابل للتفسير
See also: Interpretability
- Federated learning FLالتعلم الفيدرالي
- A decentralized approach to training machine learning models in which each node trains a global model on its local data and a central server aggregates the updates, without moving the raw data.
Elsewhere: Arabic Wikipedia: تعلم متحد
See also: Edge AI، Synthetic data
- Fine-tuningالضبط الدقيق
- A second, task-specific training pass performed on a pre-trained model to refine its parameters for a specific use case.
Elsewhere: Arabic Wikipedia: صقل
See also: Pre-training، Low-Rank Adaptation، Reinforcement learning from human feedback
- Foundation modelالنموذج التأسيسي
- A very large pre-trained model trained on an enormous and diverse training set, able to respond to a wide range of requests and to be customized through fine-tuning.
Elsewhere: Arabic Wikipedia: نموذج الأساس
See also: General-purpose AI model، Pre-training، Fine-tuning
- General-purpose AI model GPAIنموذج الذكاء الاصطناعي ذو الأغراض العامة
- Under the EU AI Act, a model—often trained on large amounts of data using self-supervision—that displays significant generality, competently performs a wide range of distinct tasks and can be integrated into many downstream systems; models used only for research or prototyping before market placement are excluded.
See also: General-purpose AI system، Foundation model، Systemic risk
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- General-purpose AI systemنظام الذكاء الاصطناعي ذو الأغراض العامة
- Under the EU AI Act, an AI system based on a general-purpose AI model that can serve a variety of purposes, both for direct use and for integration into other AI systems.
See also: General-purpose AI model، AI system
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- Generative AI GenAIالذكاء الاصطناعي التوليدي
- An emerging field with no formal definition whose models create content—text, images, audio, code and more—that is complex, coherent and original.
Elsewhere: Arabic Wikipedia: ذكاء اصطناعي توليدي
See also: Large language model، Diffusion model، Deep fake
- Graphics processing unit GPUوحدة معالجة الرسوميات
- An electronic circuit designed to speed computer graphics and image processing, whose parallel processing now underpins training and running deep learning models.
Elsewhere: Arabic Wikipedia: وحدة معالجة الرسوميات
See also: Tensor Processing Unit، Compute
Source: What is a GPU? | IBM
- Hallucinationالهلوسة
- AI outputs that sound plausible but are factually wrong, irrelevant or entirely fabricated; some prefer the term confabulation.
Elsewhere: Arabic Wikipedia: هلوسة
See also: Retrieval-augmented generation، Large language model
- High-risk AI systemنظام ذكاء اصطناعي عالي المخاطر
- Under the EU AI Act, a system that is either a safety component of (or itself) a product covered by Union harmonisation legislation in Annex I requiring third-party conformity assessment, or that falls within the use cases listed in Annex III.
See also: Provider، Deployer، AI system
Source: Article 6: Classification Rules for High-Risk AI Systems — EU Artificial Intelligence Act
- Inferenceالاستدلال
- The process of making predictions by applying a trained model to new, unlabeled examples—as opposed to training the model.
See also: Compute، Parameters
- Interpretabilityقابلية التفسير
- The ability to explain or present an ML model's reasoning in terms understandable to a human—the degree to which an observer can understand the cause of a decision.
See also: Explainable AI، AI alignment
- Knowledge graph KGالرسم البياني المعرفي
- A network of real-world entities and the relationships between them; also known as a semantic network.
Elsewhere: Arabic Wikipedia: رسم بياني معرفي
See also: Retrieval-augmented generation
Source: What Is a Knowledge Graph? | IBM
- Large language model LLMالنموذج اللغوي الكبير
- A language model with a very high number of parameters; informally, any Transformer-based language model such as Gemini or GPT.
Elsewhere: Arabic Wikipedia: نموذج لغوي كبير
See also: Transformer، Foundation model، Generative AI
- Low-Rank Adaptation LoRAالتكييف منخفض الرتبة
- A parameter-efficient tuning technique that freezes the model's pre-trained weights and inserts a small set of trainable weights, making fine-tuning much faster.
See also: Fine-tuning، Parameters
- Machine learning MLتعلم الآلة
- A program or system that trains a model from input data, so that the trained model can make useful predictions from new, never-before-seen data.
Elsewhere: Arabic Wikipedia: تعلم الآلة؛ Dubai Future Foundation: تعلم الآلة
See also: Deep learning، Supervised learning، Unsupervised learning، Reinforcement learning
- Machine learning operations MLOpsعمليات تعلم الآلة
- A set of practices designed to create an assembly line for building, deploying and running machine learning models.
See also: Machine learning، Model card
Source: What is MLOps? | IBM
- Mixture of experts MoEمزيج الخبراء
- A scheme that increases neural network efficiency by using only a subset of its parameters (experts), chosen by a gating network, to process each input.
Elsewhere: Arabic Wikipedia: خليط الخبراء
See also: Large language model، Parameters
- Model cardبطاقة النموذج
- A file that accompanies a model and provides key information about it; on Hugging Face, a Markdown file with metadata that supports discoverability and reproducibility.
See also: Open-weight model، Machine learning operations
Source: Model Cards
- Multimodal AIالذكاء الاصطناعي متعدد الوسائط
- Machine learning models capable of processing and integrating information from multiple modalities, such as text, images, audio and video.
See also: Computer vision، Natural language processing، Foundation model
Source: What is Multimodal AI? | IBM
- Natural language processing NLPمعالجة اللغات الطبيعية
- A subfield of computer science and AI that uses machine learning to enable computers to understand and communicate with human language.
Elsewhere: Arabic Wikipedia: معالجة اللغات الطبيعية؛ Dubai Future Foundation: معالجة اللغات الطبيعية
See also: Large language model، Automatic speech recognition، Text to speech
- Neural network ANNالشبكة العصبية الاصطناعية
- A model containing at least one hidden layer of connected nodes (neurons); a network with more than one hidden layer is called a deep neural network.
Elsewhere: Arabic Wikipedia: شبكة عصبونية اصطناعية؛ Dubai Future Foundation: شبكة عصبونية
See also: Deep learning، Parameters
- Open-weight modelالنموذج مفتوح الأوزان
- A model whose developer publicly releases its weights—the numerical parameters set by training—so others can download, run and customize it; this does not necessarily include training data or code, and openness is a spectrum.
See also: Parameters، Foundation model، Application programming interface
- Optical character recognition OCRالتعرف الضوئي على الحروف
- Technology that uses automated data extraction to convert images of text into a machine-readable format; also called text recognition.
Elsewhere: Arabic Wikipedia: تعرف المحارف البصري
See also: Computer vision
- Overfittingفرط الملاءمة
- Creating a model that matches the training data so closely that it fails to make correct predictions on new data.
Elsewhere: Arabic Wikipedia: فرط الملاءمة
See also: Training data، Machine learning
- Parametersالمعاملات
- The weights and biases that a model learns during training; model size is usually expressed as a parameter count.
See also: Training data، Open-weight model، Quantization
- Pre-trainingالتدريب المسبق
- The initial training of a model on a large dataset, before it is adapted to specific tasks; a pre-trained model today typically means a trained LLM or other generative model.
See also: Fine-tuning، Foundation model
- Promptالأمر النصي
- Any text entered as input to a large language model to condition it to behave in a certain way—a question, instructions, examples or a partial input.
See also: Prompt engineering، Context window، Prompt injection
- Prompt engineeringهندسة الأوامر
- The art of creating prompts that elicit the desired responses from a large language model; also called prompt design.
Elsewhere: Arabic Wikipedia: هندسة التلقين
See also: Prompt، Chain-of-thought prompting
- Prompt injectionحقن التعليمات
- A cyberattack against large language models in which attackers disguise malicious instructions as legitimate inputs, manipulating the system into leaking data or spreading misinformation.
Elsewhere: Arabic Wikipedia: حقن التعليمات
See also: Prompt، AI guardrails، Data poisoning
- Providerالمزوِّد
- Under the EU AI Act, a natural or legal person, public authority or other body that develops an AI system or general-purpose AI model, or has one developed, and places it on the market or puts it into service under its own name or trademark, whether for payment or free of charge.
See also: Deployer، High-risk AI system، General-purpose AI model
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- Quantizationالتكميم
- Reducing the precision of data by converting it from a higher-precision to a lower-precision format; used in machine learning to make models smaller and faster.
See also: Distillation (knowledge distillation)، Edge AI، Parameters
Source: What is Quantization? | IBM
- Reinforcement learning RLالتعلم المعزز
- A type of machine learning in which autonomous agents learn to make decisions by interacting with their environment, guided by rewards.
Elsewhere: Arabic Wikipedia: تعلم معزز؛ Dubai Future Foundation: التعلم المعزز
See also: Reinforcement learning from human feedback، Supervised learning
- Reinforcement learning from human feedback RLHFالتعلم المعزز بالملاحظات البشرية
- Using feedback from human raters—for example, thumbs-up or thumbs-down ratings—to improve the quality of a model's responses.
Elsewhere: Arabic Wikipedia: التعلم المعزز من ردود الفعل البشرية
See also: Reinforcement learning، Fine-tuning، AI alignment
- Retrieval-augmented generation RAGالتوليد المعزز بالاسترجاع
- An architecture for optimizing an AI model's performance by connecting it with external knowledge bases, helping LLMs deliver more relevant, higher-quality responses.
Elsewhere: Arabic Wikipedia: توليد معزز بالاستحضار
See also: Large language model، Hallucination، Embedding (vector embedding)
- Supervised learningالتعلم الخاضع للإشراف
- A machine learning technique that uses labeled data sets to train models to find patterns and relationships, so they produce accurate results on new data.
Elsewhere: Arabic Wikipedia: تعلم مراقب؛ Dubai Future Foundation: التعلم المراقَب
See also: Unsupervised learning، Reinforcement learning
- Synthetic dataالبيانات الاصطناعية
- Artificial data designed to mimic real-world data, generated through statistical methods or AI techniques such as deep learning and generative AI.
Elsewhere: Arabic Wikipedia: بيانات مصطنعة
See also: Training data، Generative AI
Source: What Is Synthetic Data? | IBM
- Systemic riskالمخاطر النظامية
- Under the EU AI Act, a risk specific to the high-impact capabilities of general-purpose AI models, with a significant impact on the Union market due to their reach or to actual or foreseeable negative effects on public health, safety, security, fundamental rights or society, which can be propagated at scale across the value chain.
See also: General-purpose AI model
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- Tensor Processing Unit TPUوحدة معالجة الموتّرات
- Google's custom-developed application-specific integrated circuits (ASICs) used to accelerate machine learning workloads.
Elsewhere: Arabic Wikipedia: وحدة معالجة تنسر
See also: Graphics processing unit، Compute
Source: Introduction to Cloud TPU
- Text to speech TTSتحويل النص إلى كلام
- Technology that converts digital text into natural-sounding audio; also called speech synthesis or computer-generated speech.
Elsewhere: Arabic Wikipedia: تصنيع كلامي
See also: Automatic speech recognition، Deep fake
Source: What is Text to Speech? | IBM
- Tokenالرمز
- The smallest unit of language that AI language models work with—a word, part of a word or punctuation mark. Inputs, outputs and context windows are measured in tokens.
See also: Context window، Embedding (vector embedding)، Large language model
- Training dataبيانات التدريب
- Data used for training an AI system through fitting its learnable parameters.
See also: Parameters، Synthetic data، Data poisoning
Source: Article 3: Definitions — EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
- Transformerالمحوِّل
- A neural network architecture that excels at processing sequential data; its central self-attention mechanism detects relationships between parts of the input. It underlies most large language models.
See also: Attention mechanism، Large language model، Token
- Unsupervised learningالتعلم غير الخاضع للإشراف
- Machine learning algorithms that analyze and cluster unlabeled data sets, discovering hidden patterns or groupings without human intervention.
Elsewhere: Arabic Wikipedia: تعلم غير مراقب؛ Dubai Future Foundation: التعلم غير المراقب / التعلم الاستنتاجي
See also: Supervised learning
- Watermarkingالعلامات المائية
- The act of embedding information, typically difficult to remove, into outputs created by AI—including text, images, audio and video—to verify their provenance or modifications.
Elsewhere: Arabic Wikipedia: علامة مائية رقمية
See also: Deep fake، Synthetic data
Abbreviations
الاختصارات| Abbreviation | English | Arabic |
|---|---|---|
| AGI | Artificial general intelligence | الذكاء الاصطناعي العام |
| AI | Artificial intelligence | الذكاء الاصطناعي |
| ANN | Neural network | الشبكة العصبية الاصطناعية |
| API | Application programming interface | واجهة برمجة التطبيقات |
| ASR | Automatic speech recognition | التعرف التلقائي على الكلام |
| CV | Computer vision | الرؤية الحاسوبية |
| CoT | Chain-of-thought prompting | سلسلة الأفكار |
| DL | Deep learning | التعلم العميق |
| FL | Federated learning | التعلم الفيدرالي |
| GPAI | General-purpose AI model | نموذج الذكاء الاصطناعي ذو الأغراض العامة |
| GPU | Graphics processing unit | وحدة معالجة الرسوميات |
| GenAI | Generative AI | الذكاء الاصطناعي التوليدي |
| KG | Knowledge graph | الرسم البياني المعرفي |
| LLM | Large language model | النموذج اللغوي الكبير |
| LoRA | Low-Rank Adaptation | التكييف منخفض الرتبة |
| ML | Machine learning | تعلم الآلة |
| MLOps | Machine learning operations | عمليات تعلم الآلة |
| MoE | Mixture of experts | مزيج الخبراء |
| NLP | Natural language processing | معالجة اللغات الطبيعية |
| OCR | Optical character recognition | التعرف الضوئي على الحروف |
| RAG | Retrieval-augmented generation | التوليد المعزز بالاسترجاع |
| RL | Reinforcement learning | التعلم المعزز |
| RLHF | Reinforcement learning from human feedback | التعلم المعزز بالملاحظات البشرية |
| TPU | Tensor Processing Unit | وحدة معالجة الموتّرات |
| TTS | Text to speech | تحويل النص إلى كلام |
| XAI | Explainable AI | الذكاء الاصطناعي القابل للشرح |