Researchers are exploring the existential implications of creative AI, including whether developing technology will actually make humans more creatively capable — or limit our ideas.
The new study, published in Science Advances by researchers from University College London and the University of Exeter, tested hundreds of short stories created with the help of ChatGPT's creative AI against those created entirely by humans. are created. One group of writers had full access to their ideas, another group could ask ChatGPT for a story idea, and a third could work with a set of five prompts created by ChatGPT. The stories were then rated on “novelty, usefulness (ie, likelihood of publication), and emotional enjoyment.” Tech Crunch.
Prime Day deals you can shop now.
Products available for purchase through affiliate links here are selected by our merchandising team. If you purchase something through links on our site, Mashable may earn an affiliate commission.
“These findings point to an increase in individual creativity that is at risk of losing collective novelty,” the study reads. “This dynamic resembles a social dilemma: with creative AI, writers improve individually, but collectively produce a narrower range of novel content.”
Mashable Light Speed
A new report suggests that bad actors on TikTok are using AI to spread political misinformation.
Participants were “measured” for creativity before the writing session with a commonly used word-production task that creates a measure of creativity among respondents. People who scored lower on these proxy tests of creativity got better scores on their personal writing when they were given access to AI-generated ideas. But for those who already had high creativity scores, the AI ideas didn't benefit their story ratings.
Additionally, the pool of stories aided by AI-generated prompts was perceived as less diverse and showed less unique writing characteristics, demonstrating the limitations of ChatGPT's universality. Mashable's Cecily Mauran reports that the new study's literature findings raise concerns about the problem of AI's self-feeding training loops, or AI models that are trained only on AI output that degrades the AI models themselves. show
study author Oliver Hauser said in a comment Tech Crunch: “Our study represents an initial approach to the larger question of how large language models and creative AI will affect human activities, including creativity more generally… It will be important that AI actually is being rigorously evaluated – rather than being implemented on a large scale under the assumption that it will have positive results.”
Titles
Artificial intelligence