Just nine per cent of humans are more creative than AI

Just nine per cent of people are more creative than the best Artificial Intelligence chatbots new psychology research has revealed.  

The study shows that the majority of humans lag behind mechanical muses and struggled to match them in a scientific study on creative writing.

Dr Paul Hanel, from the University of Essex, and Jennifer Haase, at Humboldt University, Berlin, therefore found overall “no qualitative difference between AI and human-generated creativity”.

The pre-print paper – Artificial muses: Generative Artificial Intelligence Chatbots Have Risen to Human-Level Creativity – comes as the technology is poised to transform white-collar work.

Analysts at influential consulting firm McKinsey & Company have predicted the technology could displace an estimated 400 million workers worldwide, 15% of the total workforce, by 2030.

However, Dr Hanel believes the technology will act as “valuable assistants in reviewing thoughts” rather than “omnipotent tools”.

He said: “Artificial Intelligence chatbots are quite impressive.

“However, they still require humans to use them and to interpret their outputs, so, they can be great tools to assist us.

“Fears that unemployment rates are going to skyrocket seem a bit exaggerated though.

“In the past decades, many technological changes have caused worries that machines are going to replace humans, but currently there is a shortage of skilled workers in the UK and many other countries.”

As part of the study research assistant Jennifer Haase and Dr Hanel compared 100 humans with six Generative Artificial Intelligences – including the cutting-edge and recently released ChatGPT-4.

All of the chatbots studied varied in levels of creativity with ChatGPT-4 coming out on top.

The other platforms studied were Alpha.ai, Copy.ai, ChatGPT-3, Studio and YouChat.

Participants were asked to write down as many ideas as possible for the words “a ball, fork, pants, tire and toothbrush”.

Humans were given three minutes to write down as many ideas as possible with the prompt – “What can you do with ”.

Six people and a specifically trained AI then rated the responses in a blind test – with the scores from man and machine matching.

Due to the late release of ChatGPT-4 it was only judged by the AI as human testers would know it was artificial.

However, the scientists speculate that due to the correlation between ratings in previous tests that the results would be similar.

Dr Hanel and research associate Ms Haase hope to expand the study in the future as the technology develops.

Ms Haase added “Our study demonstrates that current chatbot technology in beta stages can efficiently generate multiple alternative ideas for everyday objects.

“However, this task requires low-level creativity, and humans still have an edge in dealing with more complex problems that demand higher levels of creativity.

“Therefore, humans cannot yet entirely rely on AI for creative problem-solving.”