anomaly detection
Anomaly detection identifies observations or behavior that differ meaningfully from an expected pattern and may warrant investigation.
AI Terminology
Search clear AI terminology definitions, acronyms, related concepts and the reviewed videos that explain them.
Showing 461–480 of 1136 terms
Clear filtersanonymity
Anonymity allows a person to act or communicate without their real-world identity being known or readily linked to the activity.
API rate limiting
API rate limiting restricts how many requests a client can make within a defined period or resource budget.
application programming interface compatibility layer
An application programming interface compatibility layer translates between different interface formats so one client can communicate with another service.
application-specific integrated circuit
ASICAn application-specific integrated circuit is a chip designed for a defined class of tasks rather than broad general-purpose computing.
architecture diagram
An architecture diagram visually maps a software system's components, boundaries and important connections.
art
AI art is visual or multimedia artwork created with meaningful involvement from generative models or other AI systems.
assignment shortcut detection
Assignment shortcut detection identifies when a learner may be asking AI to bypass the intended learning process and can redirect the interaction toward guided study.
audio encoder
An audio encoder converts sound into a compact representation that an AI system can store, compare, or use while generating new audio.
audio-native language model
An audio-native language model processes learned sound representations directly, allowing it to reason about speech content and acoustic cues without relying entirely on a text transcript.
auditable reasoning trail
An auditable reasoning trail preserves the claims, evidence, decisions, revisions, and reviewer actions needed to reconstruct how a research result was produced.
authentic assessment
Authentic assessment evaluates learning through realistic tasks that require students to apply knowledge, make decisions and explain their process.
automated cyber reconnaissance
Automated cyber reconnaissance uses software to continuously identify and analyze exposed systems, services and other security-relevant information.
automated form filling
Automated form filling maps available data to form fields and enters it programmatically or through a browser agent, usually with validation before submission.
automated research evidence collection
Automated research evidence collection captures experiment outputs and metadata in a structured form that can be inspected and reproduced.
automated theorem proving
ATPAutomated theorem proving uses algorithms to search for a formal proof of a mathematical or logical statement with limited human guidance.
autonomous cybersecurity agent
An autonomous cybersecurity agent uses models and security tools to pursue a cyber objective across multiple steps with limited continuous human direction.
autoregressive video generation
Autoregressive video generation creates later AI video content using the sequence or context produced before it.
backpressure
Backpressure is a mechanism that limits incoming work when an AI service or other system cannot safely process requests at their current rate.
base model
A base model is the starting version of an AI model before it is adapted, fine-tuned or connected to additional tools for a particular use.