Top News

Researchers found that these AIs tend to repeat patterns from dominant cultures and languages, producing less varied and often less creative outputs; in group settings, reliance on LLMs can reduce the number and creativity of ideas compared with unaided collaboration.
They also noticed group brainstorming takes a hit when people rely on LLMs. Everyone ends up thinking more similarly, which could limit creativity and problem-solving.

The team suggests training AIs with more real-world diversity so they reflect a wider range of voices and ideas.
As Sourati puts it, more diverse AIs could actually help us tackle bigger challenges together.