promptowy.com / AI Glossary
AI Glossary
AI concepts explained without the jargon. Every entry includes a straightforward definition, practical examples, and answers to real-world questions.
77 terms
Fundamentals
Multimodal AIMultimodal AI refers to artificial intelligence capable of processing and combining different formats of data, such as text, images, audio, and video simultaneously.AGI (Artificial General Intelligence)Artificial General Intelligence (AGI) is a hypothetical AI system that can understand, learn, and perform any intellectual task at or above human capability.AI Parameters (Model Parameters)AI parameters are the internal numeric values that a machine learning model fine-tunes during training to recognize patterns and generate reliable answers.Artificial Intelligence (AI)Artificial Intelligence (AI) refers to computer systems that perform tasks requiring human-like problem solving, pattern recognition, learning, or language generation.Transformer ModelA transformer model is a neural network architecture that processes entire sequences at once, using self-attention to connect related words across long texts.NLP (Natural Language Processing)Natural Language Processing (NLP) is a branch of artificial intelligence that enables computers to understand, interpret, and generate human language.LLM (Large Language Model)A large language model (LLM) is an AI system trained on vast amounts of text to recognize patterns, process language, and generate relevant context-based text.Generative AI (GenAI)Generative AI is a branch of artificial intelligence that creates new content, such as text, images, audio, and code, from simple user prompts.Turing TestThe Turing Test is an evaluation method where a machine is judged intelligent if a human evaluator cannot tell its responses apart from those of a human.Neural NetworkA neural network is a computational model made of interconnected layers that learns to recognise patterns and make decisions by adjusting internal parameters.Machine Learning (ML)Machine learning is a branch of artificial intelligence where computer systems detect patterns in data to make decisions or predictions without explicit programming.AI ChatbotAn AI chatbot is software designed to simulate human conversation through text or voice, powered by natural language processing and modern machine learning models.
Prompting
Iterative PromptingIterative prompting is the process of refining AI instructions across multiple conversational steps based on previous outputs to reach an ideal result.System PromptA system prompt is a hidden, top-level instruction that sets an AI model's role, behavioral boundaries, and tone across an entire conversation.Negative PromptA negative prompt is an instruction telling an AI image generator which specific visual elements, styles, or errors to exclude from the final output.Chain-of-Thought (CoT)Chain-of-thought is a prompting technique where an AI breaks down complex problems into step-by-step reasoning before delivering a final answer.Prompt ChainingPrompt chaining is the technique of breaking a complex task into a sequence of smaller prompts, using the output of one step as the input for the next.Zero-Shot PromptingZero-shot prompting is an AI technique where you ask a model to complete a task by giving instructions only, without providing any prior examples.Few-Shot PromptingFew-shot prompting is an AI technique where you provide a small set of input and output examples in your prompt to demonstrate the exact pattern you want.Prompt (AI Prompt)A prompt is the instruction, question, or text input you give to an artificial intelligence system to guide what it generates in return.Prompt EngineeringPrompt engineering is the practice of structuring text instructions so that generative artificial intelligence systems produce precise and useful outputs.
Models and training
LoRA (Low-Rank Adaptation)LoRA (Low-Rank Adaptation) is an efficient fine-tuning technique that trains small adapter layers while keeping the underlying foundation model frozen.Model Quantization (Weight Compression in AI)Model quantization is a compression technique that reduces the numerical precision of an AI model's weights to save memory and speed up local inference.Mixture of Experts (MoE)Mixture of Experts is an AI architecture that divides a model into specialized sub-networks, activating only a few for each prompt to run fast and efficiently.Fine-Tuning (Model Customization)Fine-tuning is the process of taking a pre-trained AI model and training it further on custom data to adapt its behavior, style, or output format.Open-Weights Model (Open-Weight AI)An open-weights model is an AI system whose pre-calculated mathematical parameters are publicly shared, letting anyone download, run, and adapt it locally.OverfittingOverfitting happens when an AI model memorizes its training data instead of learning general rules, making it stumble on fresh information.Model Distillation (Knowledge Distillation)Model distillation is a machine learning technique where a compact student model is trained to reproduce the reasoning and performance of a much larger teacher model.Model TrainingModel training is the computational process where an AI system analyzes large datasets to learn patterns, adjust its internal parameters, and make accurate predictions.Synthetic Data (AI-Generated Training Data)Synthetic data is artificially generated information created by computer programs or AI models to train machine learning systems without relying on real-world records.RLHF (Reinforcement Learning from Human Feedback)RLHF is a training method that uses human preferences to teach language models how to follow instructions, stay polite, and act as helpful conversational assistants.AI Alignment (Alignment Problem)AI alignment is the practice of steering artificial intelligence systems so their actions and objectives match human values, intended goals, and safety standards.Training DataTraining data is the collection of text, images, code, or audio used to teach a machine learning model how to recognize patterns and make decisions.Model Weights (Neural Network Weights)Model weights are the numerical values inside an artificial intelligence system that determine how it processes input data and generates decisions or text.Reasoning ModelsReasoning models are artificial intelligence systems designed to deliberate, test internal logic, and verify steps before generating a final answer.
Technology and architecture
Token (AI Text Unit)A token is the basic unit of data an AI language model processes, typically representing a whole word, part of a word, or punctuation mark.Inference (AI)Inference is the stage where a trained artificial intelligence model processes live input to generate an answer, prediction, or decision.LatencyLatency in AI is the delay between sending a prompt and receiving a response from a model, most often measured as the time to first token.VRAM (Video RAM)VRAM is dedicated high-speed memory on a graphics processing unit used to store AI model weights, prompts, and temporary calculations.GPU (Graphics Processing Unit)A GPU is a specialized processor designed to handle thousands of calculations at once, making it the primary hardware engine behind modern artificial intelligence.Embeddings (Vector Embeddings)Vector embeddings are numerical representations of text, images, or audio that capture their meaning so computers can compare concepts mathematically.Grounding in AI (Model Grounding)Grounding in AI is the practice of tying a model's responses directly to verifiable external facts, documents, or tools rather than relying on guesswork.TokenizationTokenization is the process of breaking raw text into smaller computational units, called tokens, before an AI model processes or generates language.AI API (Application Programming Interface)An AI API is an interface that allows external software to send prompts or data directly to an artificial intelligence model and receive automated responses.RAG (Retrieval-Augmented Generation)RAG is an AI framework that fetches relevant information from external documents and feeds it to a language model before it generates a response.Seed (Random Seed in AI)A seed is an initial number used to control randomness in artificial intelligence, allowing users to recreate or fine-tune specific generation results.Vector Database (Vector DB)A vector database is a specialized storage engine designed to index and search data based on conceptual similarity rather than exact keyword matches.Streaming (in AI)Streaming is a method where an AI model returns its output token-by-token in real time, rather than waiting to generate the entire response before sending it.Context WindowA context window is the maximum amount of text, measured in tokens, that an AI model can read and process in a single interaction.Temperature (AI Parameter)Temperature is a setting that controls how predictable or creative an AI model is by adjusting the randomness of its word choices.
Agents and tools
Model Context Protocol (MCP)Model Context Protocol (MCP) is an open standard that connects AI assistants to external data sources, enterprise tools, and local computer environments.Vibe CodingA software development approach where you steer an AI assistant with natural language prompts, guiding code through conversation rather than typing it yourself.AI AgentAn AI agent is an autonomous software system that perceives its environment, plans multi-step tasks, and uses digital tools to achieve a specific goal.Function Calling (Tool Use)Function calling is a capability that allows an AI model to detect when external tools are needed and generate structured instructions to execute them.AI Benchmark (Model Evaluation Test)An AI benchmark is a standardized test used to evaluate, grade, and compare the performance, reasoning, or language abilities of artificial intelligence models.Computer Use (AI)Computer use is an AI capability that allows models to control a PC like a human by viewing screens, moving cursors, clicking buttons, and typing text.
Image, video, and audio
AI Upscaling (AI Image Enhancement)AI upscaling uses machine learning models to enlarge low-resolution images and videos by synthesizing sharp, realistic details rather than just stretching pixels.AI Lip SyncAI lip sync is a technology that automatically adjusts the mouth and facial movements of a person in a video to match a specific spoken audio track.TTS and STT (Text-to-Speech and Speech-to-Text)TTS turns written text into spoken audio, while STT converts spoken words into written text, forming the foundation of voice AI.Img2img (Image-to-Image)Img2img is an AI generation mode that uses an existing image alongside a text prompt to create a new picture while preserving the original composition.Text-to-Image (AI Image Generation)Text-to-image is an artificial intelligence technology that generates digital illustrations, artwork, or realistic photos directly from written natural language prompts.InpaintingInpainting is an AI technique that reconstructs, replaces, or removes selected parts of an image while preserving the surrounding style, lighting, and context.Voice CloningVoice cloning is an AI technology that creates a synthetic replica of a person's voice from recorded audio samples, enabling it to speak any written text.Text-to-Video (T2V)Text-to-video is an artificial intelligence technology that generates short, moving video clips directly from written descriptions or still reference images.Diffusion ModelA diffusion model is an AI architecture that generates media by taking pure random noise and gradually filtering it into a clear, detailed output based on a prompt.
Safety and quality
AI WatermarkAn AI watermark is an imperceptible digital signature embedded directly into synthetic text, images, or audio to prove that the content was generated by a machine.AI GuardrailsAI guardrails are technical rules, filters, and safety boundaries designed to keep artificial intelligence outputs safe, accurate, and aligned with intended tasks.Content Moderation AI (Automated Safety Filters)Content moderation AI is software that automatically scans text, images, and audio to detect and block hate speech, violence, self-harm, or illegal material.DeepfakeA deepfake is synthetic video, audio, or imagery created or altered by artificial intelligence to convincingly depict someone doing or saying things they never did.AI HallucinationAn AI hallucination occurs when an artificial intelligence generates incorrect or fabricated information while presenting it as an established, confident fact.Prompt InjectionPrompt injection is a security vulnerability where an attacker manipulates an AI system by embedding malicious instructions inside untrusted input text.AI SlopAI slop refers to low-quality, mass-produced synthetic content churned out by automated tools to game algorithms, harvest clicks, and fill digital spaces.Jailbreak (LLM)An AI jailbreak is a prompt engineered to bypass a language model's built-in safety filters and force it to produce restricted or prohibited outputs.
AI on the web
GEO (Generative Engine Optimization)Generative Engine Optimization (GEO) is the practice of tailoring online content to be cited and recommended in AI-generated answers and search summaries.Zero-Click SearchA zero-click search occurs when an AI engine or search platform answers a query directly on the results page, without requiring a click through to an external site.llms.txt (AI Web Standard)A proposed web standard: a plain Markdown file placed in a website's root directory to help large language models parse and understand key content.AI Web Crawlers (AI Scrapers)AI web crawlers are automated software bots deployed by artificial intelligence labs to scan, collect, and process online content across the web.
