Zum Inhalt springen

Fragen und Glossar derzeit auf Englisch; deutsche Inhalte folgen.

11 Begriffe

Context window
Kap. 1Kap. 4
The amount of preceding text, measured in tokens, that a language model can consider when generating a response.
Embedding
Kap. 1Kap. 4
A numerical representation of a token that encodes semantic, syntactic, and contextual relationships as a vector in a high-dimensional space.
Generative AI
Kap. 1
A branch of artificial intelligence that uses large, pre-trained models to generate human-like output such as text, images, or code.
Hallucination
Kap. 3
LLM output that appears factually incorrect or irrelevant to a given task.
Multimodal model
Kap. 1
A model capable of processing multiple data types, such as text, images, sound, and video.
Retrieval-Augmented Generation (RAG)
Kap. 4
An approach that enhances language-model response generation by retrieving and incorporating relevant information from additional data sources.
Shadow AI
Kap. 5
Use of AI tools outside an organization's approved integration and deployment approach, creating risks involving information security, data privacy, compliance, regulation, and intellectual property.
Structured prompt
Kap. 2
A prompt for software testing built from role, context, instruction, input data, constraints, and output-format components.
Symbolic AI
Kap. 1
A rule-based form of artificial intelligence that represents knowledge using symbols and logical rules.
Tokenization
Kap. 1
The process in language models of breaking text into smaller units called tokens, which may be characters, sub-words, or words.
Vector database
Kap. 4
A data store that can hold embeddings and support efficient retrieval of related content through semantic similarity.