An AI model provides the knowledge from which ChatGPT, Gemini and other chatbots generate results. The architecture of a model is a mathematical matrix (rows and columns of numbers) that is loosely based on the human brain (see
neural network). AI models and chatbots are nothing like traditional data processing applications (see
AI programming).
Pattern Recognition
Models are pattern recognition systems that are "trained" by being fed data, and in some cases, most of the knowledge available over the Internet. They are also fine-tuned by AI engineers to answer questions and/or generate original content.
The Model Is Part One
A model is part one of a two-part system. Part two is the "inference" software, which uses the model to answer questions. Inference software may be able to work with numerous models, each trained on a different subject or in a different manner. See
inference and
AI training vs. inference.
Language models deal with text while image models handle drawings and photos. Multimodal models support some combination of text, images, audio and video. See
large language model,
AI diffusion and
AI secret sauce.