Plain-language definitions of AI and machine-learning terms, curated by AI Balkans.
Terms
AI Adoption - The process of bringing AI tools and practices into an organization - from first experiments to everyday use.
AI Agent - An AI system that does not just answer questions but takes actions - browsing, using tools, and completing multi-step tasks on its own.
AI Alignment - The challenge of making AI systems reliably pursue what humans actually intend - not just what we literally asked for.
AI Bias - When an AI system produces unfair results for certain groups because of imbalances or prejudices baked into its training data.
AI Ethics - The field concerned with making AI fair, transparent, accountable, and beneficial - and avoiding harm from its use.
AI Literacy - The basic understanding of what AI can and cannot do - enough to use it productively and judge its output critically.
AI Model - The trained program that comes out of the machine-learning process - the thing that actually answers, predicts, or generates.
AI Strategy - A deliberate plan for how an organization will use AI - which problems to solve, what to build versus buy, and how to manage the risks.
Algorithm - A precise set of steps a computer follows to solve a problem or complete a task.
Artificial General Intelligence (AGI) - A hypothetical AI that could match or exceed human ability across virtually any intellectual task - it does not exist yet.
Artificial Intelligence (AI) - Computer systems that perform tasks which normally need human intelligence, like understanding language, recognizing images, or making decisions.
Auditable AI - AI designed so its decisions can be traced, reviewed, and verified after the fact - with logs and documentation to prove it.
Autoencoder - A neural network that learns to compress data down to its essence and reconstruct it - useful for spotting anything unusual.
Automation - Using technology to perform tasks with little or no human involvement - AI extends it from routine work to judgment work.
Bayesian Network - A model that maps how events influence each other with probabilities - built for reasoning under uncertainty.
Chatbot - A program you converse with in natural language. Modern AI chatbots like ChatGPT can handle open-ended conversation.
Computer Vision - AI that lets machines understand images and video - recognizing objects, faces, text, and scenes.
Context Window - The amount of text an AI model can consider at one time - its working memory for a conversation or document.
Continual Learning - The ability of an AI system to keep learning new things over time without forgetting what it already knows.
Convolutional Neural Network (CNN) - The neural network design specialized for images - it scans pictures for patterns, from edges up to whole objects.
Data Privacy - Your right to control how your personal information is collected and used - including by AI systems trained on it.
Deep Learning - A powerful type of machine learning that uses many-layered neural networks to learn from very large amounts of data.
Deepfake - AI-generated fake audio, images, or video that convincingly imitates real people saying or doing things they never did.
Diffusion Model - The AI technique behind most image generators - it creates pictures by gradually turning random noise into a coherent image.
Digital Transformation - The broader shift of organizations to digital ways of working - AI is its newest and most powerful chapter.
Embedding - A way of representing words, sentences, or images as lists of numbers so that similar meanings end up close together.
EU AI Act - The European Union law regulating artificial intelligence - the first comprehensive AI regulation in the world.
Explainable AI (XAI) - Techniques for making AI decisions understandable to humans - answering why the model decided what it decided.
Fine-Tuning - Taking a pre-trained AI model and training it further on specific data so it becomes better at a particular job.
Foundation Model - A huge model trained once on broad data, then adapted for many different uses - the base layer of modern AI products.
GAN (Generative Adversarial Network) - Two neural networks locked in a contest - one generates fakes, the other judges them - until the fakes become convincing.
Generative AI - AI that creates new content - text, images, music, video, or code - instead of just analyzing existing data.
GPT - Generative Pre-trained Transformer - the family of language models developed by OpenAI that powers ChatGPT.
Guardrails - The safety rules and filters built into AI systems to prevent harmful, illegal, or off-limits behavior.
Hallucination - When an AI confidently states something false - inventing facts, sources, or details that do not exist.
Inference - The moment an already-trained AI model is actually used - answering your question or generating your image.
Knowledge Graph - A network of facts - entities and the relationships between them - that lets software connect the dots like a mind map.
Large Language Model (LLM) - An AI model trained on enormous amounts of text to understand and generate human language. ChatGPT, Claude, and Gemini are built on LLMs.
Machine Learning - A way of building software where the computer learns rules from examples instead of being programmed step by step.
Machine Translation - AI that translates text or speech between languages automatically, like Google Translate or DeepL.
Multimodal AI - AI that works with more than one kind of input or output - text, images, audio, and video together.
Narrow AI - AI built for one specific job - excellent inside its lane, useless outside it. All AI in use today is narrow AI.
Natural Language Processing (NLP) - The field of AI focused on understanding and generating human language - the foundation under chatbots, translation, and search.
Neural Network - A computing structure loosely inspired by the brain, made of layers of simple units that together learn complex patterns.
Neuro-Symbolic AI - A hybrid approach combining neural networks that learn from data with symbolic systems that reason by rules.
Parameters - The internal numbers a model learns during training. Model sizes are described by parameter count, like 7B or 70B (billions).
Predictive AI - AI that forecasts what is likely to happen next - demand, churn, risk, failures - from patterns in historical data.
Prompt - The instruction or question you give an AI system. Better prompts produce noticeably better answers.