C3 AI Inc., a developer of artificial intelligence models customized to different industries and use cases, today announced C3 Generative AI for Government Programs, a generative AI application aimed at helping federal, state and local governments deliver information about government programs via a chat interface.

The company cited overburdened and underfunded government call centers that can create long delays for citizens looking for basic information about such topics as healthcare, public safety and benefits programs. It said its application can handle many inquiries directly using search and chat in 130 languages with enterprise-grade security, access controls and full traceability.

The application is independent of underlying large language models and provides source references where appropriate. Its model-driven architecture is customized to handle structured and unstructured sources such as federal legislation, state databases, reference documents and website content.

The language model is tailored to government data and the types of questions government agencies typically receive. For example, a user can ask, “What are the enrollment steps for the Affordable Care Act?” and receive a detailed, step-by-step response with citations to source web pages. The system can also understand follow-up questions like “What is a special enrollment period?”, “What qualifies me for a special enrollment period?” and “Does changing jobs qualify me for a special enrollment period?”

C3 Generative AI for Government Programs runs on Google LLC’s Cloud Platform and uses Google Cloud’s Gemini family of language models. The recently announced Gemini 1.5 features a two-million token context window, multi-modal capabilities and strong reasoning capabilities.

A generative AI context window refers to the amount of text or data that the AI model can consider at one time when generating responses. C3 said it all but eliminates hallucination by separating language and data models. The software is available today.

Photo: Unsplash