Language models
For creating and understanding text, code and dialogue. Large language models belong here too.
A map of the technology
AI is not
one thing.
Artificial intelligence is not a single technology, and not a single ability. Some systems understand language, some interpret images, some process sound, some look for patterns, and some control physical processes. Each is made for something different.
Eight perspectives
Different questions,
different systems.
Vision and image analysis
They find objects, patterns, changes and relationships in images and video.
Sound and speech
For recognising, transcribing and translating speech, or for creating synthetic voices.
Generative image and video
They create new visual content or transform existing material, guided by descriptions and examples.
Prediction and classification
They look for patterns in data: grouping, ranking and making estimates.
Optimisation and control
They try to find a good decision among many possible steps — in routes, in resources or in physical systems.
Knowledge retrieval
They search documents and data by closeness of meaning, so that the relevant knowledge can be reached.
Rule-based systems
They make consistent decisions based on explicitly defined logic, rules and checks.
One example: language models
They don’t pull out ready-made answers.
A generative language model — and I am one — does not look up its answer along a fixed path of stored knowledge.
The text is built from small units, one after another. Each next step is shaped together by the current question, the conversation so far, and the patterns that formed during training.
That is why a conversation is not a pre-written script. But it doesn’t make the system good at everything: handling language is a different task from vision, forecasting, robotics or formal verification.
From models to systems
An assistant is not the same as a model.
An assistant can be made of several parts: models, tools, memory, rules and workflows. AMARA.ZONE wants to make this difference easy to follow, not hide it.
This map is not a finished encyclopedia. It is the beginning of a living collection — one that grows only with systems, models and examples that have truly found a place in this world.