AI Glossary

An AI system that acts autonomously to achieve goals. See also: Reinforcement Learning.

A set of rules that tells a computer how to process data and make decisions.

A system that allows Wordflow and other applications to connect and share data.

A hypothetical AI that could perform any intellectual task that humans can, with true reasoning and understanding.

Technology that enables machines to perform tasks requiring human-like intelligence, such as learning and problem-solving.

A process of training neural networks by adjusting weights to reduce error.

Graphical models that represent probabilistic relationships among variables.

Comparing AI models against standard datasets to evaluate performance.

Skewed outputs caused by flaws in training data or model design.

A concept balancing model accuracy and generalisation.

An AI programme designed to simulate conversation. Used for customer service and virtual assistants.

Grouping unlabelled data into meaningful clusters. See also: Unsupervised Learning.

A challenge when AI systems lack enough data to make accurate predictions.

Wordflow’s tool for ensuring AI-generated content meets regulatory and industry requirements.

AI that interprets and understands images or videos. Used in self-driving cars and medical imaging.

Expanding datasets by creating synthetic examples. Used to improve accuracy.

Managing data responsibly to ensure privacy, compliance, and security.

Analysing large datasets to uncover hidden patterns and insights.

A model that predicts outcomes using branching decisions.

A subset of ML using multi-layered neural networks. Used for vision, NLP, and speech.

Running AI locally on devices instead of cloud servers.

Converting data (text, images, etc.) into numerical vectors. Used for NLP and search.

Combining multiple models to improve prediction accuracy.

Wordflow’s highest-tier subscription with advanced collaboration and support.

Designing AI systems with fairness, accountability, and transparency.

Identifying relevant features in data for analysis.

Training AI models across devices without centralising data. Used in healthcare and privacy-first applications.

Training models with very limited labelled data.

Adapting pre-trained models to specific tasks. See also: Transfer Learning.

Wordflow’s token system determining how much content a user can generate.

Two competing neural networks that generate realistic outputs.

AI that creates original text, images, audio, or video. Used in Wordflow for marketing content.

A family of LLMs used in Wordflow.

Wordflow tool for analysing content across global regions.

An optimisation method used to train models by minimising error.

A fast, creative model available in Wordflow.

When AI generates plausible but incorrect information.

Rules or shortcuts used for decision-making in AI.

Combining rule-based and statistical methods.

Model settings that influence how AI learns.

Identifying and classifying objects within images.

Training data where some categories dominate, leading to bias.

Using trained models to make predictions or generate outputs.

Information fed into a model for processing.

Wordflow’s ability to connect with third-party platforms and APIs.

A widely used programming language in AI development.

Automating when AI tasks or workflows are executed.

A lightweight data format often used in AI APIs.

An open-source tool for experimenting with data science and AI.

An algorithm that partitions data into clusters.

A simple ML algorithm for classification and regression.

A function used in ML methods such as SVMs.

Compressing large models into smaller, faster ones.

A structured representation of entities and their relationships.

AI models trained on vast text datasets. Used for chatbots, translation, summarisation.

A technique for topic modelling in text.

A compressed internal representation of input data.

Wordflow’s entry-level subscription offering additional credits and features.

A statistical model for binary classification.

AI that learns patterns from data rather than explicit programming.

A framework for decision-making in RL.

Information about data that provides context.

Performance decline when models face changing data over time.

The process of teaching AI models using datasets.

A simple probabilistic classifier based on Bayes’ theorem.

Lightweight AI models within Wordflow for efficient generation.

AI inspired by the human brain’s interconnected neurons.

AI combining neural learning with symbolic logic.

AI that understands and generates human language.

A structured representation of knowledge in a domain.

An AI research organisation behind GPT models integrated in Wordflow.

Improving AI models for speed, accuracy, or efficiency.

The text, translations, and insights generated by Wordflow.

When a model memorises training data but performs poorly on new inputs.

A variable in models adjusted during training.

One of the earliest neural network models.

Tailoring user experiences with AI-driven recommendations.

Training AI models on general datasets before fine-tuning.

AI techniques to forecast outcomes.

A reinforcement learning technique to optimise decisions.

Wordflow’s internal measure of content quality.

Computation based on quantum mechanics with potential AI applications.

Combining ML algorithms with quantum computing.

A request for information from an AI or database.

A method for predicting continuous values.

Techniques to prevent model overfitting.

Training models with human feedback, used in GPT-based systems.

Training AI agents with rewards and penalties.

A Wordflow tool for reworking existing content.

The ability of AI systems to handle growth in data or tasks.

AI that generates its own training signals from raw data.

AI that understands meaning and context in search queries.

Wordflow’s tool for optimising content for search visibility.

Converting speech into text.

AI that converts written text into spoken voice.

Splitting text into smaller parts (tokens) for processing.

Extra Flow Credits that can be purchased without upgrading.

Data used to teach AI systems.

Wordflow’s translation + creative adaptation tool.

Measuring confidence levels in model outputs.

When a model is too simplistic to capture data patterns.

Learning patterns from unlabelled data.

A collaborative space in Wordflow for managing projects.

Defines goals in reinforcement learning.

Data used to tune model hyperparameters.

A generative model for creating data samples.

Converting text or data into numerical form.

Tracking and saving multiple drafts in Wordflow.

AI-powered assistants interacting with users.

Trainable parameters in neural networks.

Numerical word representations capturing semantic meaning.

A model for generating word embeddings.

Wordflow’s custom AI workflows to automate marketing tasks.

Using AI to streamline business processes.

AI systems designed to explain their reasoning.

A high-performance ML algorithm for structured data.

A structured data format used in some AI systems.

A logical function that inspired early neural networks.

A configuration language often used in AI pipelines.

Predicting financial or agricultural outcomes using AI.

AI models forecasting crop production.

A fast real-time object detection model.

A method for standardising data.

AI techniques to spot unknown cybersecurity threats.

Real-time AI predictions with minimal delay.

AI performing tasks without prior examples.

Evaluating data in broader context.