Glossary

HanyAI AI Glossary Navigating the terminology of artificial intelligence can be complex. The HanyAI Glossary is designed to demystify key concepts, acronyms, and technical terms you'll encounter in our directory and the broader AI field. This resource supports our mission of clarity and education, helping users from all backgrounds make sense of technologies like Machine Learning (ML), Natural Language Processing (NLP), Neural Networks, Computer Vision, and more. Understand the tools you discover with greater depth and confidence.rminology explained. Learn about common terms used in the AI tool landscape. HanyAI AI Glossary: The Definitive Guide to Artificial Intelligence Terminology Navigating the world of AI tools requires understanding a new language. Use this comprehensive glossary to decode technical specs, compare tool features, and make informed decisions. Updated regularly as the AI landscape evolves. A Agent / AI Agent An autonomous system that can perceive its environment, make decisions, and perform tasks to achieve specific goals without constant human intervention. Examples: AutoGPT, customer support chatbots. Why it matters: Represents a shift from "chatbots" to "doers." AI Ethics A set of moral principles and techniques aimed at ensuring AI systems are fair, transparent, accountable, and respect human rights. Why it matters: Essential for evaluating a tool's long-term viability and brand safety. Alignment The challenge of ensuring that an AI system's goals and behaviors are consistent with human values and intended objectives. Why it matters: A core challenge in advanced AI safety (crucial for tools like Claude). API (Application Programming Interface) A set of protocols that allows one software application to interact with another. In AI, an API allows developers to integrate a model (like GPT-4) into their own apps. Why it matters: Indicates if a tool is "plug-and-play" vs. a standalone product. Augmented Generation See RAG. B Bias (in AI) Systematic errors in an AI model that create unfair outcomes, often reflecting prejudices present in the training data. Why it matters: A critical filter for tools used in hiring, lending, or legal contexts. Black Box A system where the internal logic or decision-making process is invisible or incomprehensible to humans. Why it matters: Relevant for "Explainable AI" (XAI) features. BPE (Byte-Pair Encoding) A tokenization algorithm used by models like GPT to break down words into subword units. Why it matters: Affects how token counts (and thus cost) are calculated. C Chatbot A computer program designed to simulate human conversation, typically powered by NLP or LLMs. Why it matters: The most common entry point for consumer AI. Computer Vision (CV) A field of AI enabling computers to derive meaningful information from digital images, videos, and other visual inputs. Why it matters: Core category for tools like Midjourney, Stable Diffusion, and facial recognition. Context Window The maximum amount of text (in tokens) a model can consider at one time when generating a response. Why it matters: Crucial for analyzing long documents (e.g., "Claude offers 100K token context"). Convolutional Neural Network (CNN) A deep learning architecture primarily used for analyzing visual imagery. Why it matters: Legacy architecture, largely superseded by Transformers in modern tools. D Deep Learning (DL) A subset of machine learning based on artificial neural networks with multiple layers ("deep"). Why it matters: Powers almost all modern cutting-edge AI tools. Diffusion Model A generative model that creates data (usually images) by gradually adding noise to training data and then learning to reverse the process. Why it matters: The "D" in DALL-E and Midjourney; key for image generation. Discriminative Model A model that classifies or predicts labels based on input data, distinguishing between different types of data points. Opposite of Generative AI. Why it matters: Used for spam detection, fraud detection. Fine-Tuning The process of taking a pre-trained model and training it further on a specialized, smaller dataset to adapt it for specific tasks. Why it matters: Key enterprise feature; allows customizing generic tools for niche needs. Freemium A business model where basic features are free, but advanced features require a subscription. Why it matters: Highly filterable on HanyAI (e.g., "ChatGPT: Free plan available"). G Generative AI (GenAI) A type of AI that can create new content (text, images, audio, code) rather than simply analyzing existing data. Why it matters: The primary category for most tools on HanyAI. GPT (Generative Pre-trained Transformer) A family of LLMs developed by OpenAI, known for their ability to generate human-like text. Why it matters: Industry benchmark; often used as a generic term for text AI. GPU (Graphics Processing Unit) A specialized processor originally designed for rendering graphics, now essential for training and running AI models due to its parallel processing capabilities. Why it matters: Determines hardware requirements and speed for local AI tools. Guardrails Programmatic constraints placed on an AI model to prevent it from generating harmful, illegal, or off-brand content. Why it matters: Critical for enterprise deployment to mitigate risk. H Hallucination When an AI model generates confident-sounding but completely false or nonsensical information. Why it matters: Primary risk factor in using LLMs for factual research. Hugging Face A platform and community for hosting, sharing, and collaborating on machine learning models and datasets. Why it matters: The "GitHub for AI"; sign of open-source compatibility. Hyperparameters Configuration variables external to the model that govern the training process (e.g., learning rate, batch size). Why it matters: Advanced setting for custom fine-tuning. I Inference The process of a trained AI model making a prediction or generating output based on new input data. Why it matters: Most API costs are billed "per inference" or "per token." Instruction Tuning A fine-tuning method that trains a model to better follow human instructions/directions. Why it matters: What makes ChatGPT "helpful" vs. a raw base model. LLM (Large Language Model) A deep learning model trained on massive amounts of text data to understand, generate, and manipulate human language. Why it matters: The engine behind tools like ChatGPT, Claude, and Gemini. J Jailbreak A prompt or technique designed to circumvent the safety restrictions and guardrails of an AI model. Why it matters: Indicates security robustness of a tool. K Knowledge Cutoff Date The date of the last data point included in an AI model's training set. The model has no knowledge of events after this date unless connected to live search. Why it matters: Essential for tasks requiring current information. L LangChain A framework designed to simplify the development of applications powered by LLMs, enabling chaining of calls and connection to external data sources. Why it matters: Dominant framework for AI app developers. Latency The time delay between sending a request to an AI model and receiving the response. Why it matters: Critical for real-time applications like voice assistants. LoRA (Low-Rank Adaptation) A parameter-efficient fine-tuning technique that adapts large models without retraining all weights. Why it matters: Makes customizing enterprise models faster and cheaper. M Machine Learning (ML) A subset of AI where systems learn from data to improve performance on specific tasks without being explicitly programmed for every rule. Why it matters: Foundational technology; all modern AI tools use ML. Multimodal AI An AI system capable of processing and integrating multiple types of data (e.g., text + images + audio) simultaneously. *Why it matters: Cutting edge; seen in GPT-4V and Gemini.* N Natural Language Processing (NLP) A branch of AI focused on enabling computers to understand, interpret, and manipulate human language. Why it matters: The category for text-based tools. Neural Network A computing system loosely inspired by the biological brain, composed of interconnected nodes (neurons) organized in layers. Why it matters: The architecture underlying deep learning. O One-shot / Few-shot Prompting A technique where you provide one or a few examples of the desired output format directly within the prompt to guide the model. Why it matters: A key skill for effective prompt engineering. Open Source / Open Weights Software or AI models released with licenses allowing free use, modification, and distribution (e.g., Llama 3, Stable Diffusion). Why it matters: Signifies lower cost and high customizability. Overfitting When a machine learning model learns the training data too well, including its noise, and performs poorly on new, unseen data. Why it matters: Sign of a low-quality or poorly trained model. P Parameter A numerical value inside a model that is learned during training and determines how the model interprets input data to generate output. Often used as a measure of model size (e.g., 7B, 70B). Why it matters: More parameters generally = more capability, but also higher cost. Prompt The input text provided to an AI model to instruct it on what to generate. Why it matters: The primary user interface for generative AI. Prompt Chaining Breaking a complex task into a sequence of simpler prompts, where the output of one prompt becomes the input for the next. Why it matters: Improves output quality and reliability. Prompt Engineering The practice of designing, refining, and optimizing prompts to achieve specific, high-quality results from generative AI models. Why it matters: A new, in-demand technical skill. R RAG (Retrieval-Augmented Generation) A technique that combines a retrieval system (searching a database/knowledge base) with an LLM. The model retrieves relevant information first, then generates an answer based on it. Why it matters: Reduces hallucinations and keeps answers current without retraining. Reinforcement Learning from Human Feedback (RLHF) A training technique where a model is fine-tuned using human preferences as a reward signal to align its outputs with human expectations. Why it matters: The secret sauce behind ChatGPT's conversational ability. Responsible AI An approach to developing and deploying AI systems that aligns with ethical values, fairness, and transparency. Why it matters: Governance requirement for regulated industries. S SaaS (Software as a Service) A software licensing and delivery model where applications are hosted in the cloud and accessed via web browser. Why it matters: The dominant distribution model for AI tools. Stable Diffusion A popular open-source deep learning model specifically designed for generating images from text descriptions. Why it matters: Democratized AI image generation; key alternative to Midjourney. Sentiment Analysis The use of NLP to identify and extract subjective information from text to determine the writer's attitude (positive, negative, neutral). Why it matters: Common business tool for social listening. T Temperature A hyperparameter that controls the randomness of a model's output. Low temperature = more deterministic/factual; High temperature = more creative/random. Why it matters: Advanced setting in most LLM APIs. Tensor A multi-dimensional array of numbers; the fundamental data structure used in deep learning frameworks like TensorFlow and PyTorch. Why it matters: The "language" neural networks speak. Token The basic unit of text that an LLM processes. A token can be a word, part of a word, or a punctuation mark (e.g., "ChatGPT" = 3 tokens). Why it matters: The unit of billing for virtually all commercial LLM APIs. Transformer A deep learning architecture based on a "self-attention" mechanism, introduced in the 2017 paper "Attention Is All You Need." It is the "T" in GPT and BERT. Why it matters: The single most important architecture in modern AI. U Unsupervised Learning A type of machine learning where the model finds patterns in unlabeled data without explicit instructions on what to predict. Why it matters: How base LLMs are initially trained (predicting the next token). Use Case A specific scenario or task for which an AI tool is applied (e.g., "Customer Support Automation," "Code Documentation"). Why it matters: Core filtering criteria on HanyAI. V Vector Database A database designed to store, index, and query high-dimensional vectors (embeddings). Essential for implementing RAG and similarity search. Why it matters: The backbone of enterprise "chat with your data" tools. Versioning The practice of managing different releases of a model (e.g., GPT-3.5 vs GPT-4). Why it matters: Critical for stability and regression testing in production. W Weights The learned parameters within a neural network that transform input data as it passes through the layers. Why it matters: The "file" you download when using an open-source model. Workflow Automation Using AI to automate a sequence of tasks across different applications without manual triggers. Why it matters: Key productivity benefit of advanced agents. X XAI (Explainable AI) A set of methods and techniques that allows human users to understand and trust the results and output created by machine learning models. Why it matters: Regulatory compliance in finance and healthcare. Z Zero-shot Learning The ability of a model to perform a task it was never explicitly trained on, based solely on its pre-existing knowledge and the prompt instructions. Why it matters: Defines the flexibility of a foundation model.