Accuracy in model evaluation is the share of predictions a model gets right, a simple metric that becomes misleading when classes are…
An activation function is the nonlinear rule a neural network neuron applies to its weighted input, letting stacked layers model complex patterns.
Agent memory is the set of mechanisms that let an AI agent keep, retrieve and update information across steps, sessions and tasks…
Agent orchestration is the control layer that sequences, routes and supervises the steps, tools and sub-agents of AI agents in a multi-step…
Agentic AI describes AI systems that pursue goals with autonomy, planning steps, calling tools and acting on results with limited human direction.
An AI agent is a system that pursues a goal by observing its environment, planning steps and acting through tools in a…
AI alignment is the work of making AI systems pursue the goals, values and constraints their designers and users actually intend, even…
An AI benchmark is a fixed set of tasks, data and scoring rules used to measure and compare the capabilities of AI…
AI bias is a systematic error that makes an AI system produce unfair or skewed outcomes for certain groups, inputs or situations…
An AI gateway is a proxy layer between applications and AI model providers that centralizes routing, access control, cost tracking, caching and…
AI guardrails are the technical and policy controls placed around an AI model that check its inputs, outputs and actions against safety…
AI red teaming is the structured adversarial testing of AI models and systems to uncover harmful, insecure or unintended behavior before attackers…
Artificial general intelligence (AGI) is a hypothetical AI system able to learn and perform most cognitive tasks at or above human level…
Artificial intelligence is the field of computer science that builds systems able to learn, reason and act on tasks that normally need…
An attention mechanism lets a neural network weight the most relevant parts of its input for each output, and it is the…