Artificial intelligence
35 terms, explained one by one
Concepts for understanding and applying AI in products and processes. Includes language models, agents, information retrieval, evaluation and governance.
A
- Agent memoryAgent memory brings together mechanisms for storing and retrieving information between the steps or interactions of an AI system.
- Agent orchestrationAgent orchestration coordinates tasks, tools, states and communications within a system with one or more agents.
- AI adoptionAI adoption is the process of incorporating artificial intelligence capabilities into organisational work in a useful, controlled and sustainable way. It includes selecting tasks, adapting processes, preparing data and people, and evaluating outcomes.
- AI agentAn AI agent is a system that uses information and tools to decide on and carry out steps towards a goal, with a defined degree of autonomy.
- AI governanceAI governance establishes responsibilities, policies and controls for developing and using artificial intelligence systems.
- AI guardrailsAI guardrails are controls that limit or verify a system's inputs, responses and actions.
- AI hallucinationAn AI hallucination is generated output that presents false, invented or unsupported information as if it were valid.
- AI inferenceInference is the use of a trained model to produce an output from an input.
- AI model evaluationAI model evaluation checks whether a system meets quality and behaviour criteria on defined tasks.
- Algorithmic biasAlgorithmic bias is a systematic deviation in data, methods or results that can produce inappropriate effects or inequalities.
- Artificial intelligenceArtificial intelligence brings together methods and systems that can carry out tasks such as recognising patterns, generating content or supporting decisions based on information.
C
- ChatbotA chatbot is a system that holds a conversational interaction through text, voice or other means.
- Context engineeringContext engineering designs what information, instructions and resources an AI system receives at each moment.
- Context windowThe context window is the amount of information a model can take into account within an interaction, according to its processing limits.
E
F
G
H
L
M
- Machine learningMachine learning develops models that learn patterns from data in order to carry out tasks.
- MCPMCP, or Model Context Protocol, is a protocol for connecting AI applications to tools and information sources through common interfaces.
- Multi-agent systemA multi-agent system combines several agents that interact to tackle a task or a shared environment.
- Multimodal modelA multimodal model processes or generates more than one type of information, such as text, images, audio or video.
P
- PromptA prompt is the input that guides the response or action of a generative AI system.
- Prompt engineeringPrompt engineering is the practice of designing, testing and refining instructions to obtain suitable results from a generative model.
- Prompt injectionPrompt injection is an attempt to make an AI system follow instructions inserted into content that it should treat as data.
R
S
- Semantic searchSemantic search retrieves information based on relationships of meaning, as well as literal word matches.
- Small language modelA small language model, or SLM, is relatively small in scale compared with the other models it is being compared against.
- Synthetic dataSynthetic data is generated artificially to represent the characteristics or cases of a data set.