EN
Contact & Support
MenuMenu
Automation
Virtual commissioningvalidate the plant on a modelPLC programmingTIA, TwinCAT, CodeSysRobot programmingKUKA, ABB, FANUC, URFunctional safetysafe stop, guardingElectrical designdiagram and control cabinetOn-site commissioningstart-up at the real plantBuildings and processesbuilding and process automationTechnical project managementcoordinate teams and versionsMaterial flow simulationthroughput and cycle timeTrainingtrain staff on the modelOpen overviewVirtual commissioning FAQTerms in automation & virtual commissioning
Industrial IoTSoftwareAIIT/OTProductsCompany
Legal NoticePrivacy PolicyCookie PolicyTermsAccessibilitySitemap

Terms in Artificial intelligence

The entries explain technical terms from artificial intelligence and lead to the service in which they occur.

A

AIArtificial intelligence
Methods with which computers learn tasks such as recognising, predicting, summarising or deciding from data. Artificial intelligence
AI ActEU Artificial Intelligence Act
EU Regulation 2024/1689 that classifies AI systems by their risk and attaches obligations for providers and deployers. AI Act obligations
AI agent
Application in which a language model breaks a task down into steps and calls tools such as search, database or interfaces for them. Assistants and agents
AI assistant
Application that drafts texts, answers questions or searches for information on request and leaves the decision with the user. Assistants and agents
AI governance
Rules and responsibilities according to which a company selects, approves, operates and reviews AI. Rollout and evidence
AI literacy
Obligation under Article 4 of the AI Act: providers and deployers ensure that their staff use AI systems competently. It has applied since February 2025. AI Act obligations
Anomaly detection
Method that finds deviations from normal behaviour in measured data, for example unusual vibrations of a spindle. Data analysis and forecasting

C

Context window
Amount of text that a language model takes into account at the same time in one request, measured in tokens. Integrating tools

E

Embedding
Representation of a text or image as a numerical vector in which similar content lies close together. Knowledge management

F

Fine-tuning
Further training of a pre-trained model with own examples that adapts it to a task or technical language. Integrating tools

G

GDPRGeneral Data Protection Regulation
EU regulation that governs the processing of personal data, including when AI is used. AI Act obligations
GPUGraphics processing unit
Graphics processor that executes many arithmetic operations in parallel and speeds up training and inference of AI models. Operation on the premises

H

Hallucination
Answer of a language model that sounds plausible and is factually wrong or has no source. Rollout and evidence
High-risk AI system
AI system that the AI Act subjects to particular obligations on risk management, documentation and human oversight because of its area of use. AI Act obligations
Human in the loop
Way of working in which a person checks and approves the result of an AI before it takes effect. Rollout and evidence

I

Inference
Application of a trained model to new inputs, for example the answer of a language model to a question. Operation on the premises

L

LLMLarge language model
Language model trained on large amounts of text that understands, summarises and generates text. Integrating tools

M

Machine learning
Field of AI in which a model learns patterns from example data without every rule being programmed individually. Data analysis and forecasting
Machine vision
Analysis of camera images, for example to check components for dimensions, completeness or surface defects. AI in automation
MCPModel Context Protocol
Open protocol through which AI applications connect tools, data sources and interfaces in a uniform way. AI agents and workflows

O

OCROptical character recognition
Text recognition that converts printed or handwritten characters from images and scans into editable text. Integration into existing systems
On-premise
Software and models run on own hardware in the own building or own data centre. Operation on the premises
Open-weight model
Language model with published weights that can be operated on own hardware. Operation on the premises

P

Predictive maintenance
Maintenance that schedules service from condition data such as vibration, temperature or current draw before a failure occurs. Predictive maintenance
Prompt
Input to a language model made up of instruction, context and question. Integrating tools

R

RAGRetrieval augmented generation
A language model retrieves matching passages from own documents before answering and bases its answer on them. Knowledge management

T

Token
Smallest unit of text into which a language model splits input and output. Context length and costs are measured in tokens. Integrating tools
Training data
Data set from which a model learns. Quantity, quality and rights to these data determine what the model achieves. Data analysis and forecasting

U

Use case
Described task for which AI is to be used, with benefit, data situation and effort. Identifying use cases

V

Vector database
Database that stores texts or images as numerical vectors and finds entries with similar content. Knowledge management