Meta creates a new ‘inclusive AI training dataset’ so bots are fair

This could be an important step in combating inaccurate, racist and sexist replies from OpenAI’s ChatGPT or Google’s Bard.

Meta hopes that the launch of Casual Conversations V2 will help AI researchers make their tools and procedures universally inclusive. This was stated in a March 9 statement by the company.

The vast new dataset includes video clips of face-to-face interactions from a wide range of human participants, across diverse geographic, cultural and racial demographics. It is an upgrade to the 2021 AI audiovisual training dataset.

Source:
https://interestingengineering.com/culture/meta-creates-new-inclusive-ai-training-dataset-so-bots-can-be-fair-chatgpt-openai