Foundational Models on Google Cloud

Machine learning (ML) models enable generative artificial intelligence (AI) to write articles, paint pictures, or even write computer code. Developing and training ML models can be an expensive and time-consuming process, even when you rely on open-source models, such as Bidirectional Encoder Representations from Transformers (BERT) [1], since you will still need to prepare large amounts of proprietary training data.

Foundational models are pre-trained models that are made available by providers, on a pay-per-use basis, and usually accessed via an API. You can read more about large language models (LLMs) and their training process in Appendix 1. Foundational models can leverage the power of the public cloud to give practitioners cost-efficient and easy access to cutting-edge generative AI models behind simple API calls.

Google Cloud makes multiple foundational models available to its users on its Vertex AI platform, including open-source models...