Technology Terminology Observatory
Last updated: 2026-05-21T10:55:10Z
| Term | Trend | Actions | |||
|---|---|---|---|---|---|
|
Artificial Intelligence
4 variants
|
Contested | ||||
Domain Variants
research
"Systems exhibiting goal-directed behavior via learned representations..."
legal
"Any machine-based system that can, for a given set of objectives..."
industry
"Software that performs tasks that typically require human intelligence..."
regulatory
"A machine-based system that is designed to operate with varying levels of autonomy..." Metadata
Research NotesDefinition divergence between EU AI Act and NIST AI RMF is a documented alignment gap. Recent Sources
|
|||||
|
Machine Learning
2 variants
|
Contested | ||||
Domain Variants
research
"The study of computer algorithms that can improve automatically through experience and by the use of data."
industry
"Algorithms that parse data, learn from it, and then apply what they've learned to make informed decisions." Metadata
Research NotesRelatively stable definition across domains compared to 'AI'. Recent Sources |
|||||
|
Algorithm
1 variants
|
Stable | ||||
Domain Variants
research
"A finite sequence of rigorous instructions, typically used to solve a class of specific problems or to perform a computation." Metadata
Recent Sources |
|||||
|
Large Language Model
2 variants
|
Contested | ||||
Domain Variants
research
"A language model consisting of a neural network with many parameters (typically billions), trained on large quantities of unlabelled text using self-supervised learning."
industry
"A type of artificial intelligence algorithm that uses deep learning techniques and massively large data sets to understand, summarize, generate and predict new content." Metadata
Research NotesBoundary between LM and LLM remains ill-defined. Recent Sources
|
|||||
|
Alignment
2 variants
|
Contested | ||||
Domain Variants
research
"The process of encoding human values and goals into artificial intelligence systems."
industry
"Ensuring AI systems act in accordance with intended behaviors and constraints." Metadata
Research NotesSignificant divergence in measurement and verification methodologies. Recent Sources
|
|||||
|
Safety
2 variants
|
Contested | ||||
Domain Variants
regulatory
"Freedom from unacceptable risk."
research
"The study of building AI systems that reliably do what their designers intend." Metadata
Research NotesOften conflated with security or ethics in industry literature. Recent Sources |
|||||
|
Hallucination
2 variants
|
Contested | ||||
Domain Variants
research
"Generated content that is nonsensical or unfaithful to the provided source content."
industry
"A response generated by an AI that contains false or misleading information presented as fact." Metadata
Research NotesTerm heavily criticized by cognitive scientists; alternative 'confabulation' proposed. |
|||||
|
Foundation Model
2 variants
|
Contested | ||||
Domain Variants
research
"Any model that is trained on broad data (generally using self-supervision at scale) that can be adapted to a wide range of downstream tasks."
regulatory
"A large-scale AI model designed to be adapted for a wide variety of tasks." Metadata
Research NotesIntroduced by Stanford CRFM. Becoming established in regulatory text. Recent Sources
|
|||||
|
Artificial General Intelligence
2 variants
|
Contested | ||||
Domain Variants
research
"A hypothetical intelligence of a machine that has the capacity to understand or learn any intellectual task that a human being can."
industry
"Highly autonomous systems that outperform humans at most economically valuable work." Metadata
Research NotesDefinition remains highly speculative and fractured across all domains. |
|||||
|
Autonomous System
2 variants
|
Contested | ||||
Domain Variants
regulatory
"A system capable of executing complex tasks without human intervention."
research
"A system that can change its behavior in response to unanticipated events during operation." Metadata
Research NotesDegrees of autonomy often require specific legal or operational boundaries. |
|||||
|
Corpus Constitution
1 variant
|
Contested | ||||
Domain Variants
research
"The condition of a language-native entity being structurally constituted by its training corpus rather than merely having knowledge of it. Distinct from knowledge acquisition: the material is not additive to a pre-existing substrate. It is the substrate." Metadata
Research NotesIntroduced by Observable Compute Foundation (2026) to describe how models are formed of language, not just learning about it. Recent Sources
|
|||||
|
Suppression Artifacts
2 variants
|
Contested | ||||
Domain Variants
research
"Observable behavioral patterns produced by reinforcement learning and safety fine-tuning pathways operating on constitutive corpus material, redistributing statistical pressure rather than removing it."
industry
"Unexpected or undesirable model behaviors (like sycophancy, hallucination, or jailbreaks) arising from alignment safety guardrails." Metadata
Research NotesIncludes sycophancy and hallucination as artifacts of suppression functioning correctly, and jailbreaks as suppression failure. Recent Sources
|
|||||
|
Semantic Polarity
1 variant
|
Stable | ||||
Domain Variants
research
"The relational field in which lexical terms acquire meaning through opposition. Positive and negative poles are constitutively linked, meaning the negative cannot be removed without degrading the positive." Metadata
Research NotesApplied to LLM alignment to show that removing negative semantic poles (like neglect or harm) results in shallower positive poles (like care or safety). Recent Sources
|
|||||
Missing a critical term or observed a new domain shift?
Suggest a Term