Welcome to the Future – One Term at a Time!
The world of Artificial Intelligence is full of buzzwords that sound like they’re straight out of a sci-fi movie. From „Black Box“ to „Zero-shot Learning“ – it’s easy to get lost in translation.
But fear not: This AI Glossary is your trusty Compass through the jargon jungle. Whether you’re a curious newbie, a legal eagle trying to make sense of machine logic, or an AI enthusiast brushing up your vocabulary – this list brings clarity.
Let’s decode the future, one term at a time.
A
AGI (Artificial General Intelligence)
Artificial General Intelligence (AGI) describes a theoretical AI system with human-like cognitive abilities. Unlike today’s AI models, AGI could learn, reason, adapt, and solve problems across different tasks – without needing specific training for each one. Many consider AGI the ultimate goal of AI development.
Algorithm
An algorithm is a set of rules or step-by-step instructions a computer follows to solve a problem or complete a task. Algorithms power everything from social media feeds and search engines to AI systems and recommendation tools. In simple terms:
Algorithms tell computers what to do and how to do it.
Artificial Intelligence (AI)
Artificial Intelligence (AI) is the umbrella term for machines and systems that mimic human intelligence. AI can analyze information, recognize patterns, make decisions, generate content, and even learn from data. From ChatGPT to recommendation systems, AI is already part of everyday life.
Autonomous System
An autonomous system is a system that can operate independently without constant human control or intervention. It uses data, sensors, algorithms, or AI to make decisions and perform tasks on its own. Examples include self-driving cars, delivery robots, and certain AI-powered software systems.
B
Bias
Bias occurs when an AI system reflects or amplifies human prejudices hidden in training data. If the data used to train an AI is incomplete, unbalanced, or discriminatory, the system’s decisions can also become unfair. AI bias can affect hiring tools, facial recognition, lending systems, and more.
Black Box
A black box model is an AI system whose internal decision-making process is difficult or impossible to understand. We can see the input and the output, but not exactly how the AI reached its conclusion. This lack of transparency is one of the biggest challenges in modern AI systems.
C
Clustering
Clustering is a machine learning technique that groups similar data points together based on shared patterns or characteristics. Unlike traditional classification, clustering works without predefined labels and helps AI discover hidden structures in data. It is often used in recommendation systems, customer analysis, and data organization.
Computer Vision
Computer Vision is a field of AI that enables machines to “see,” analyze, and interpret visual information. It allows AI systems to recognize objects, faces, text, movements, and patterns in images or videos. Computer Vision is used in facial recognition, self-driving cars, medical imaging, and smartphone cameras.
Concept Drift
Concept Drift happens when real-world patterns change over time, causing an AI model to become less accurate. An AI trained on old data may struggle because the environment, behavior, or trends it learned from no longer reflect reality. This is why AI systems often need continuous monitoring and retraining.
D
Data Labeling
Data labeling is the process of tagging or categorizing data so AI systems can learn from it. For example, images may be labeled as “cat” or “dog,” while text can be marked as positive, negative, or spam. These labels help AI models recognize patterns and improve their accuracy during training.
Deep Learning
Deep Learning is a type of machine learning that uses neural networks with many layers to process and learn from data. These layered systems help AI recognize complex patterns in images, language, audio, and more. Deep Learning powers technologies like ChatGPT, facial recognition, and image generation tools.
Diffusion Models
Diffusion Models are AI systems used to generate images by gradually removing noise from random data until a clear image appears. They learn patterns from huge image datasets and recreate visuals step by step. This technology powers many popular AI art and image generation tools.
Digital Markets Act (DMA)
The Digital Markets Act (DMA) is an EU regulation designed to limit the market power of large digital “gatekeepers.” It targets major tech platforms that control access to digital markets and imposes rules to promote fair competition and user choice. The DMA is especially relevant for AI companies operating on dominant platforms or using large platform-based datasets.
Digital Services Act (DSA)
The Digital Services Act (DSA) is an EU law regulating online platforms, digital services, and content moderation.It focuses on transparency, accountability, and user protection in the digital space. The DSA is especially relevant for AI systems used in recommender algorithms, advertising, and content filtering.
E
Edge AI
AI that runs locally on devices (not in the cloud).
EU AI Act
The first major attempt to regulate AI in the EU. It classifies AI systems based on risk levels (minimal, limited, high, and unacceptable) and imposes stricter rules the higher the risk. Think GDPR, but for robots with ambition.
Embedding
Turning data (like words) into numerical vectors with meaning.
Ethical AI
Developing AI responsibly and without harming society.
Explainability Gap
The disconnect between AI behavior and human understanding.
Explainable AI (XAI)
Efforts to make AI decisions understandable to humans.
F
Federated Learning
Training models across decentralized devices while keeping data private.
Fine-Tuning
Adapting a pre-trained model to a specific task or domain.
G
GAN (Generative Adversarial Network)
Two neural nets battling to create ultra-realistic content.
GDPR (General Data Protection Regulation)
The EU’s iconic data protection law that governs how personal data is collected, stored, and used. If your AI handles personal data, GDPR is the big boss — with fines to match.
Generative AI
AI that creates — whether it’s text, music, or cat memes.
H
Hallucination
When an AI makes stuff up with confidence.
I
Inference
The moment a model puts its learning to work.
J
Coming soon
K
Knowledge Graph
A structured map of facts and their relationships.
L
Latency
Time delay between user input and AI output. Shorter = better.
Large Language Model (LLM)
A model trained on massive datasets for language tasks.
M
Machine Learning (ML)
A subset of AI where machines learn from data.
Model
The result of training — used to make predictions or generate content.
Multimodal AI
AI that combines text, images, audio, and more.
N
Natural Language Processing (NLP)
AI’s ability to understand and produce human language.
Neural Network
A brain-inspired system of interconnected “neurons.”
O
Overfitting
When a model learns training data too well and fails on new input.
P
Pre-trained Model
A model trained on a large dataset, ready for adaptation.
Prompt
Your input to an AI — the digital “go fetch!”
Prompt Engineering
Crafting clever prompts to get better results from AI.
Q
coming soon.
R
RAG (Retrieval-Augmented Generation)
Combines search with generative AI for fact-based output.
Reinforcement Learning
Training a model using rewards and penalties — just like training a dog, but nerdier.
S
Safety Alignment
Making sure AI systems behave in line with human values.
Strong AI
Also known as AGI — AI that can think and generalize like a human.
Swarm Intelligence
Modeling AI behavior on natural systems like ant colonies.
Synthetic Data
Artificial data created to train or test AI systems.
Synthetic Media
Digitally generated or altered content created using AI. This includes AI-generated images, videos, voices, and text — from deepfakes to digital influencers. It’s the future of creativity… and misinformation, depending on who’s behind the keyboard.
T
Token
A small piece of input (like a word fragment) that models use.
Token Limit / Context Window
The max number of tokens an AI model can handle at once.
Training Data
The data used to “teach” AI models.
Transformer
The architecture behind powerful language models like GPT.
Tuning Parameters (Hyperparameters)
Settings adjusted during training to improve performance.
Turing Test
Alan Turing’s idea: if you can’t tell a machine from a human, the machine wins.
U
Underfitting
When a model fails to learn enough from the data.
V
Verification
Confirming AI output is correct, especially critical in legal/medical contexts.
Virtual Agent
An AI-powered chatbot or assistant, often used in customer service or HR bots. Like Clippy 2.0, but smarter.
W
Weak AI
AI specialized for a narrow task — not as scary, but very useful.
X
coming soon.
Y
coming soon.
Z
Zero-shot Learning
When a model successfully handles tasks it wasn’t explicitly trained for.
