Spring til indhold
Kristian Primdal

AI Ordbog

263 AI-begreber forklaret på dansk. Fra grundlæggende koncepter til de nyeste trends inden for kunstig intelligens, maskinlæring og agentisk AI.

Grundlæggende AI-koncepter

Artificial General Intelligence (AGI) Artificial Super Intelligence (ASI) Autoencoder Computer Vision Deep learning Diffusion model Foundation model GAN (Generative Adversarial Network) Generativ AI Kunstig intelligens (AI) Large Language Model (LLM) Large Multimodal Model (LMM) Maskinlæring (ML) Multimodal AI Narrow AI (ANI) Natural Language Processing (NLP) Natural Language Understanding (NLU) Neurale netværk Recursive Self-Improvement (RSI) Small Language Model (SLM) Transformer

Modeller & Udbydere

API (Application Programming Interface) Claude (Anthropic) Claude Opus Claude Sonnet Closed source modeller DALL-E DeepSeek Gemini (Google) Gemma (Google) GLM (Zhipu AI) Google DeepMind GPT (Generative Pre-trained Transformer) GPT-4 / GPT-4o GPT-5 Grok (xAI) Hermes (Nous Research) Imagen Kimi (Moonshot AI) Llama (Meta) MAI (Microsoft) Midjourney MiniMax Mistral Nano Banana Nemotron (NVIDIA) Open source modeller Perplexity Perplexity Computer Qwen (Alibaba) Sora Stable Diffusion Whisper

Teknik & Infrastruktur

Attention mechanism Backpropagation Batch processing Batch size Cloud AI Cosine similarity CUDA Dimensionality reduction Edge AI Embedding Endpoint Epoch GPU (Graphics Processing Unit) Gradient descent Hyperparameter Inference Latency Learning rate Loss function Model serving Ollama On-device AI Overfitting Self-attention Semantic search (AI-søgning) Softmax Throughput TPU (Tensor Processing Unit) Underfitting Vector database

Træning & Læringsmetoder

Active learning Continual learning Curriculum learning Distillation (model distillation) Federated learning Few-shot learning Fine-tuning In-context learning LoRA (Low-Rank Adaptation) Meta-learning One-shot learning PEFT (Parameter-Efficient Fine-Tuning) Pre-training QLoRA Quantization (kvantisering) Reinforcement Learning (RL) RLAIF (Reinforcement Learning from AI Feedback) RLHF (Reinforcement Learning from Human Feedback) Self-supervised learning Supervised learning Transfer learning Unsupervised learning Zero-shot learning

Prompting & Interaktion

Chain-of-thought (CoT) Completion Context window Grounding Guardrails Hallucination Jailbreaking Prompt Prompt engineering Prompt injection Retrieval-Augmented Generation (RAG) Streaming (response streaming) System message System prompt Temperature Token Tokenizer Top-k sampling Top-p (nucleus sampling) Tree-of-thought

AI-produktivitet & Anvendelse

AI automation AI workflow AI-assistent Chatbot Classification Copilot (koncept) Image-to-text Low-code AI Machine translation Named Entity Recognition (NER) No-code AI OCR (Optical Character Recognition) Personalization med AI Robotic Process Automation (RPA) Sentiment analysis Speech-to-text (STT) Summarization Text-to-code Text-to-image Text-to-speech (TTS) Text-to-video

Agentisk AI & Kodning

A2A (Agent-to-Agent Protocol) Agentic AI Agentic coding Agentic Engineering AI-agent Bash Autonomi (AI autonomi) CLI (Command-Line Interface) Claude Code Claude Cowork Coding agent Code generation Code review med AI Codex (OpenAI) Context Engineering Cursor GitHub Copilot Harness Engineering Human-in-the-loop (HITL) IDE (Integrated Development Environment) Loop Engineering Manus (AI-agent) MCP (Model Context Protocol) MCP Server Multi-agent system NemoClaw Open WebUI OpenClaw Orchestration (agent orchestration) Pair programming med AI Planning (agent planning) Plugin Marketplace PRD (Product Requirements Document) Progressive Disclosure Ralph Loop ReAct (Reasoning + Acting) Scaffold / Scaffolding Skill Creator Skills Spec-Driven Development Tool use (function calling) Vibe coding Windsurf

Etik, Sikkerhed & Regulering

AI alignment AI governance AI safety Ansvarlig AI (Responsible AI) Bias (AI bias) Black box Constitutional AI Copyright og AI Data privacy (AI) Deepfake EU AI Act Explainability (XAI) Fairness Interpretability Model card Red teaming Shadow AI Syntetisk data

Forretning & Strategi

AI literacy AI readiness AI-first AI-modenhed (AI maturity) AI-strategi AI-transformation Augmented intelligence Churn prediction Conversational AI Digitaliseringstilskud Enterprise AI Forward Deployed Engineer Human-AI collaboration Predictive analytics Prescriptive analytics Recommendation engine ROI af AI Tokenomics Use case

Nye / Trending Termer (2025-2026)

Agentic workflow AI Research Automation AI Slop AI-native Benchmark (AI benchmark) Browser use Catastrophic forgetting Chinchilla scaling Compound AI system Computer use Context caching ELO rating (for AI-modeller) Emergent abilities Extended thinking GenBI (Generativ Business Intelligence) HumanEval Inference-time compute JSON mode Leaderboard Long context MCP marketplace Mixture of Experts (MoE) MMLU Model collapse Multimodal reasoning Proof of human Reasoning model Scaling laws Sparse model Structured output SWE-bench Synthetic data generation Thinking tokens Tokenmaxxing Vision model World model

Ordbogen udvides løbende. Mangler du et begreb? Skriv til mig på LinkedIn.