Can synthetic logic promote creativity? Sure – though at a price: NPR

Can an AI chatbot make a person extra talented?

Proponents of artificial logic say it can serve as an inspiration, but critics are skeptical – they say it does little more than remix existing works.

Now, the latest analysis means parts of each argument are reasonable. AI can help an individual become more creative, but it also risks reducing creativity in the family.

‘Unused’ concepts

Questions over the merits of AI in art have been swirling since Giant Language Models (often known as LLMs) burst onto the scene about two years ago. Companies like OpenAI have introduced their products as tools that artists can use to enhance their production. While some writers this week say they have adopted AI as a tool for their streamlining process, many alternative artists and creators have expressed skepticism. Some people have even filed a lawsuit alleging that the device attributes copyrighted paintings to copyrighted works.

Oliver Hauser, an economist at Exeter College in the United Kingdom who researches artificial reasoning, wanted to grapple with and answer the obvious question of whether AI can only produce creativity.

“It has a kind of incredible ability to present content at the click of a button,” he says. AI, on the other hand, can always create stories that are similar in nature.

“It may be that it’s not as creative as you might think, and it doesn’t help you be more creative,” he says.

To grapple with this complex question of creativity and gain some insight, Hauser teamed up with Anil Doshi at the College School London Faculty of Control. He recruited about 300 people, whom Doshi says were not established as skilled writers. “We asked them to write a short eight-sentence story,” he says.

About one-third of the writers needed to get a grip on concepts on their own, while others generated starter concepts through the chatbot ChatGPT 4.0. Those who were given support were divided into two subgroups: one who were given a single AI-generated idea, and the other who were given one of 5 options to choose from.

The important thing, says Doshi, is that both the human-only and AI-assisted groups had to write the stories themselves.

“Our intention was to focus on whether AI could help human creativity,” says Doshi. “This was not a horse race between AI versus humans.”

The results have been evaluated through a group of 600 evaluators. They have been requested to classify each story based on its “novelty” and “usefulness”. Novelty was a symbol of the originality of the story, pace was used as a measure of whether the story was in top form as it unfolded.

The results, published these days within the magazine progress of science, It was found that stories written with AI support were considered more useful and useful. Writers who had access to one AI performed better, but those who had access to five ideas saw the biggest boost — they identified stories as bookable about 8% more than humans did on their own. Wrote, and 9% more helpful.

Moreover, Doshi says, the worst writers benefited the most.

“People who were naturally the least creative experienced the greatest improvements in their creativity,” he says.

So actually AI seems to be oversimplifying the masses. However, there is a twist to the plot: when Hauser and Doshi looked at all the stories, they got a different impression.

“Collectively speaking, there was little diversity of innovation in the group with AI,” Hauser says.

social catch 22 status

In other words, the chatbot made each individual more creative, but it made the gang that had AI support less creative.

Hauser describes the disturbing end result as a “classic social dilemma” – a situation where the public benefits individually, but the team suffers.

“We’re concerned that, on a large scale, if too many people are using it… it will reduce diversity and creativity in the population overall,” he says.

Science myth maker and journalist Annalee Newitz questions the findings. It’s difficult to determine whether someone is extra-talented: “I think part of creativity is that it can’t really be measured in percentages like that,” says Newitz.

Nonetheless, when Newitz attempted to reproduce one of the key AI story concepts using paper forms, he clearly saw how using AI would generate similar stories.

For example, when asked to generate story ideas for an “adventure on the high seas”, they found that the AI ​​would consistently incorporate the clichéd idea of ​​finding charity into the story. And it looked like it would capture the phrase “The real treasure was…” – a familiar web meme. Newitz says, because the AI ​​is proficient at a variety of texts, it seems it might be attracted to frequently used clichés at first.

Newitz also says that the social Catch 22 situation that the study warns about has already hit the sci-fi population. The deadline for the sci-fi novel Clarksworld was required to be submitted online because “they were filled with AI-written stories.”

Ultimately, Newitz says he wouldn’t blame anyone who wanted to struggle with the use of AI to write a story. However, in the long run, he believes those tools outweigh the purpose of writing.

Inventive writing, says Newitz, is “humans communicating with other humans.” “Even if something is badly written – even if it’s not very creative – if it’s written by a human being, it’s serving its purpose.”