OpenAI is introducing Data Partnerships²³ to work with organizations to produce public and private datasets for training AI models. It wants to broaden the training data for AI to cover more subjects, industries, cultures and languages so models can better understand and serve all of humanity.

What’s going on here?

OpenAI is on the lookout for more data for its models via its data partnership program.

What does this mean?

OpenAI is already partnering with organizations and governments to contribute data representing specific industries or countries. It has two stated goals for this program: Create an open-source archive of diverse datasets and private datasets to train proprietary OpenAI models.

In turn, Open AI is promising world-class OCR (for PDFs, images etc.) and ASR (for audio and video files) to turn your data future-ready.

Why should I care?

AI works if it helps everyone. To make that a reality diverse datasets are a need. Data partnerships can get more representation from areas that don’t have an active presence on the open web. For countries, it’s crucial to have a footprint in the systems that are being used globally.

OpenAI says if you have data you wish to keep private but want AI to better understand, you can partner for private datasets. But do you really want to share what might be the only edge a couple of years in the future?