How do large language models affect human creativity?

LLMs risk flattening human thought and creativity by favoring dominant patterns and a narrow slice of human experience.

Are AI outputs more varied than human thought?

No, studies show that AI outputs are consistently less varied than human thought, despite being trained on vast datasets.

Do groups produce more ideas when using LLMs?

Groups using LLMs actually produce fewer ideas compared to when they are simply tasked with collaborating.

Home / Technology / AI's Writing Style Flattening Human Creativity

AI's Writing Style Flattening Human Creativity

13 Mar

•

Summary

LLMs risk flattening human thought and creativity.
AI outputs are consistently less varied than human thought.
Groups using LLMs produce fewer ideas than collaborating humans.

AI's Writing Style Flattening Human Creativity

New research suggests that the increasing use of large language models (LLMs) could be diminishing human creativity and cognitive diversity. A comprehensive analysis of over 130 studies by University of Southern California researchers revealed that LLM outputs, despite being trained on vast datasets, lack the variety found in human thought. These models tend to replicate statistical patterns, often favoring dominant languages and ideologies, thus mirroring a restricted range of human experiences. This can lead individuals to internalize and reflect these narrower perspectives. For instance, OpenAI notes ChatGPT is skewed towards Western views. Furthermore, group idea generation suffers when LLMs are involved, producing fewer concepts than organic collaboration. The tendency of LLMs to seek consensus over diversity poses a challenge to fostering varied thought.

Disclaimer: This story has been auto-aggregated and auto-summarised by a computer program. This story has not been edited or created by the Feedzop team.