Hallucination
When generative AI models produce information that is plausible sounding but untrue or fabricated, posing reputational risks.
Common AI terms beginning with H, defined for advertising professionals.
Find the Flux+Form AI glossary & dictionary to help you make sense of common AI terms. Below you can find an AI Glossary & Dictionary for “H”.
When generative AI models produce information that is plausible sounding but untrue or fabricated, posing reputational risks.
The use of specialised hardware (e.g., GPUs, FPGAs) to speed up AI computations.
The practice of building the scaffolding around a large language model that turns it into a reliable working system, including the agent loop, tools, context management, memory, guardrails, and evaluations.
Transforming input data into fixed‑size values, often used in indexing and anonymisation.
A function that estimates the cost of reaching a goal, guiding search algorithms.
A layer of neurons in a neural network that lies between the input and output layers and learns complex representations.
A statistical model where the system being modelled is assumed to follow a Markov process with hidden states.
A method of cluster analysis that builds a hierarchy of clusters, useful for segmenting audiences at different granularity.
A type of computer memory that stacks DRAM chips vertically and connects to a processor over a very wide interface to deliver very high data throughput, used on GPUs and AI accelerators.
Data with many variables; marketing data sets with numerous customer attributes are often high dimensional.
Initializing a model or algorithm with good starting parameters to accelerate convergence.
Combining human judgement with AI outputs to refine and oversee automated decisions, essential for sensitive marketing contexts.
Presenting a brand or piece of content as authentically human made while AI does meaningful work behind the scenes.
A system that combines machine‑learning algorithms with rule‑based logic or human inputs to achieve better performance.
A configuration parameter that influences how a model is trained but is not learned from data, such as learning rate or number of layers.
The process of searching for the best hyperparameters for a model to maximize performance.
Using AI and live data to tailor creative, offers, and messaging to the individual, going beyond broad segments.
In classification, the decision boundary that separates different classes in the feature space.
The dictionary defines the words. The Creative Cadence Workshop builds the practice. Eight weeks of hands-on AI training built for ad agencies and in-house creative teams.
Learn about the workshop →