The Department of Energy’s Office of the Chief Information Officer has released the second version of DOE’s Generative Artificial Intelligence Reference Guide.
DOE said the document offers an overview of the key benefits, risks, best practices and considerations associated with the responsible development, use and implementation of generative AI tools.
The reference guide provides a background on generative AI and cites opportunities to apply the technology, including key applications for text, speech, code and image or video.
For text, the use cases cited in the document are summarizing documents, inferring reference, translating language and writing emails.
The guide’s operationalization section provides foundational knowledge of three key concepts surrounding generative AI technology: organizational roles, public versus protected data and service models.
The document outlines generative AI-related best practices that are classified into people, organization and technology categories.
Under the tech category, best practices include providing a safe space for experimentation, exercising data minimization best practices when collecting data for a system and considering using synthetic data to mitigate risks surrounding privacy, data integrity and insufficient data.