A new Australian study has smashed the myth that generative AI systems such as ChatGPT could soon replace society's most creative playwrights, authors, songwriters, artists and scriptwriters.
The existing large language models (LLMs) have a built-in mathematical ceiling on their creative capacity, meaning they will never rival the originality or ingenuity of the most creative individuals.
That's the finding from creativity expert David Cropley, a Professor of Engineering Innovation at the University of South Australia, whose study on the mathematical limits of generative AI has been published in the Journal of Creative Behaviour.
Prof Cropley computed the creative ability of LLMs using standard mathematical principles. The results showed that LLM creativity has a maximum of 0.25 on a scale from 0 (no creativity) to 1 (maximum creativity).
"While AI can mimic creative behaviour - quite convincingly at times - its actual creative capacity is capped at the level of an average human and can never reach professional or expert standards under current design principles," says Professor Cropley.
He says the findings challenge widespread assumptions about AI's creative powers and offer clarity amid a global debate that has often been clouded by hype and misunderstanding.
"Many people think that because ChatGPT can generate stories, poems or images, that it must be creative. But generating something is not the same as being creative. LLMs are trained on a vast amount of existing content. They respond to prompts based on what they have learned, producing outputs that are expected and unsurprising.
"Unfortunately, many people misunderstand the meaning of creativity. They think it means 'the act of bringing something into being' or 'generating' when in fact it means creating something that is new, original and effective.
"Creative performance, however, is not symmetrical. Typically, 60% of people are below average when it comes to creativity, so it's inevitable that a sizeable slice of society will think that LLMs like ChatGPT are creative, when they're not. Highly creative people will recognise the weaknesses in the generative AI systems."
Prof Cropley's study is the first formal calculation of AI creativity based on the internal mechanics of LLMs.
He says that while AI can be a useful support tool, it cannot replace a highly talented, creative person.
"A skilled writer, artist or designer can occasionally produce something truly original and effective. An LLM never will. It will always produce something average, and if industries rely too heavily on it, they will end up with formulaic, repetitive work.
"For AI to reach expert-level creativity, it would require new architecture capable of generating ideas not tied to past statistical patterns."
The findings are encouraging for anyone who values human innovation, he says.
"This research shows that the world still needs creative humans - perhaps more than ever."
The paper, 'The Cat Sat on the … ?' Why Generative AI has Limited Creativity' is published in the Journal of Creative Behaviour. DOI: 10.1002/jocb.70077
