OpenAI has introduced Data Partnerships, a program aimed at collaborating with external organizations to build diverse datasets for training AI models. The initiative seeks to involve more entities to enhance model utility.

The program aims to collect large-scale datasets reflecting human society, particularly focusing on diverse modalities like images, audio, and video, with an emphasis on human intention expression.

OpenAI plans to develop open-source and private datasets. The open-source dataset will be publicly accessible for AI model training, while private datasets cater to organizations wanting to maintain data privacy while improving AI models' domain understanding.

Initial collaborations include working with the Icelandic government and others to enhance AI's capabilities in specific domains.

Analyst QuickTake: OpenAI previously partnered with companies like Shutterstock to obtain access to its image, video, and music libraries and associated metadata, as well as Associated Press to obtain news stories to train its models.