10 terms
Showing all terms starting with M
A branch of AI where algorithms improve automatically through experience and exposure to data, without being explicitly programmed.
A mathematical function trained on data that maps inputs to outputs. In AI, models range from simple regressions to billion-parameter LLMs.
AI models that can process and generate multiple types of data - such as text, images, audio, and video - within a single unified system.
A pre-training objective where random tokens are masked and the model learns to predict them, used in BERT-style encoder models.
A neural network with an external memory component it can read from and write to, enabling longer-term storage than context alone.
Training AI models to learn how to learn, enabling rapid adaptation to new tasks with minimal data, also called "learning to learn".
A structured document describing an AI model's intended use, training data, performance metrics, limitations, and ethical considerations.
A degradation phenomenon where models trained on AI-generated data progressively lose diversity and accuracy over successive generations.
An architecture that routes different inputs to specialised sub-networks (experts), enabling very large models to run efficiently at inference.
A framework where multiple AI agents collaborate, negotiate, or compete to complete complex tasks that exceed the ability of a single agent.
