<- Scope Go to ToC References ->
Table 1 references the Terms defined by MPAI-WMG V1.0. The definition of all MPAI-defined Terms is accessible online.
Table 1 – Terms defined by MPAI-GAM V1.0
| Term | Definition |
| Adaptive learning | The characteristic of an AI system that can change its behaviour as a result of the processing of input Data. |
| AI System | A machine able to infer from input Data how to generate outputs relevant to its function. |
| Data augmentation | A technique that increasing the training dataset with new training examples obtained by altering some features of the original training dataset. |
| Federated learning | A machine learning technique that trains algorithms collaboratively keeping the Data in edge devices. |
| Large Language Model | An machine learning technique to train an AI System on extremely large datasets to make it able to understand and generate natural language. |
| Loss function | A function used in training that produces a quantitative assessment of an AI system producing an output. |
| Machine Learning | Techniques that make a system capable of learning how to perform a task from data without explicitly programming it. |
| Model | A component of an AI System that produces outputs by making inferences from inputs. |
| Natural language processing | The processing, analysis, and generation of human language by machine. |
| Neural Network | A set layers of simple processing elements connected by weighted links with adjustable weights. |
| Prompt | Inputs to a generative AI system describing the task the system is requested to perform, such as respond to question. |
| Reinforcement learning | A process that enables a machine to optimise its behaviour in an environment by maximising the advantage earned of its actions. |
| Small Language Model | A Language Model characterised by smaller values of the model’s neural network size, the number of parameters, and the volume of data it is trained on. |
| Training data | Data used for training an AI system, e.g., by determining the weights of a neural network through fitting its learnable parameters. |
