Foundation Model
A large AI model trained on broad data that serves as the base for many downstream applications.
AI-generated
A foundation model is a large model trained on diverse, general-purpose data that can be adapted to many different tasks. GPT-4, Claude, and LLaMA are foundation models. They are called "foundation" because they serve as the base layer on which specialized applications are built, either through fine-tuning or prompting.
The foundation model paradigm has reshaped the AI industry. Instead of building specialized models from scratch for each task, organizations start with a foundation model and customize it. This dramatically reduces the cost and expertise needed to deploy AI for specific use cases.
Stanford CRFM: "On the Opportunities and Risks of Foundation Models" - https://crfm.stanford.edu/report.html
Wikipedia: Foundation models - https://en.wikipedia.org/wiki/Foundation_model