Foundation Model

Large-scale AI models trained on massive, diverse datasets that serve as the foundation for multiple downstream applications rather than being purpose-built for specific tasks. Foundation models like the latest GPT, Claude and Gemini models, plus the big image generators, learn general representations of language, images, or other data that can be adapted to countless specific uses through prompting, fine-tuning, or additional training. These models represent a paradigm shift. Instead of training separate models for each task, you build on powerful general-purpose foundations. Foundation models require enormous computational resources to train (millions of dollars), but once created, they're accessible through APIs.

Why it matters

Foundation models democratise access to cutting-edge AI capabilities. Rather than needing ML expertise and resources to train models, entrepreneurs can leverage foundation models through simple API calls or interfaces, dramatically lowering barriers to AI-powered businesses.