OpenAI Data Partnerships launches today to expand AI training data

OpenAI Data Partnerships, a new program from the AI company, has arrived today. It practically collaborates with organizations to create datasets for training AI models. 

The Microsoft-backed tech maker is looking for large datasets that represent human society and are not yet publicly available online. Not too long ago, they also announced the GPT-4 Turbo model that supports 128K tokens.

The company can work with any type of data, including text, images, audio, and video, but is particularly interested in data that shows human intent, such as long-form writing or conversations.

“Data Partnerships are intended to enable more organizations to help steer the future of AI and benefit from models that are more useful to them, by including content they care about,” says the company in the announcement.

OpenAI offers two partnership options: open-source archives and private datasets. It’s pretty much self-explanatory: the company is looking to build a public dataset for training language models, which later could be used for free by anyone, and develop its own private datasets to train its proprietary AI models and fine-tune them. 

Organizations with private data that want OpenAI’s AI models to better understand their domain can partner with OpenAI under this program. OpenAI will treat all data with the level of sensitivity and access control that the organization prefers — and if your organization fits what they’re looking for, you can submit your interest here.