Artificial intelligence (AI) — software that performs tasks which normally require human intelligence: understanding language, recognising images, making predictions.
Machine learning (ML) — the main way modern AI is built: instead of programming rules by hand, the system learns patterns from many examples.
Neural network — a mathematical structure loosely inspired by the brain: layers of simple units that together can learn very complex patterns. Deep learning means neural networks with many layers.
Model — the result of training: a file of learned numbers that can be run to answer questions, generate speech, detect objects and so on.
Dataset — the collection of examples a model learns from: recorded speech, texts, labelled photographs. For low-resource languages like Karakalpak, building the dataset is usually the hardest part.
Training — the process of showing a dataset to a model so it adjusts itself. Requires large computing power, usually GPUs (graphics processors).
Inference — using a trained model to get answers. Much cheaper than training; this is what happens every time you use an AI product.
LLM (large language model) — a model trained on enormous amounts of text so it can read and write language. ChatGPT, Gemini and Claude are LLM-based assistants.
TTS (text-to-speech) — technology that turns written text into spoken audio. karakalpakvoice.uz provides TTS for Karakalpak.
STT / ASR (speech-to-text / automatic speech recognition) — the reverse: turning spoken audio into written text.
Voice cloning — reproducing a specific person’s voice from a short recording, so one consistent voice can speak any text.
Computer vision — AI that understands images and video: detecting objects, reading documents, describing scenes.
NLP (natural language processing) — the branch of AI that works with human language: translation, summarisation, question answering.
Chatbot — a program you can talk to in ordinary language. Modern chatbots are built on LLMs.
Hallucination — when an AI confidently states something false. A reminder that AI output always needs human checking in serious matters.