Recent research has revealed that Generative Artificial Intelligence (GAI) chatbots, such as ChatGPT versions 3 and 4, Studio.ai, and YouChat, have already reached creativity levels comparable to humans.
The new findings, based on several recently published empirical studies, challenge the long-standing belief that creativity is an exclusive domain of human intelligence.
For years, the general assumption has been that while AI can excel in logical and structured tasks, creativity remains a uniquely human trait. However, a recent study led by Dr. Jennifer Haase from the Department of Computer Science at Humboldt University, Berlin, and Dr. Paul H. P. Hanel from the University of Essex is turning that notion on its head by demonstrating that the line between human and AI-generated creativity is blurring.
In a series of meticulously designed tests, Dr. Haase and Dr. Hanel compared ideas generated by humans with those produced by various GAI chatbots, including prominent names like alpa.ai, Copy.ai, ChatGPT (versions 3 and 4), Studio.ai, and YouChat.
The study employed the Alternative Uses Test (AUT), a standard measure frequently used in creativity tests. Using the AUT, 100 human participants and five general artificial intelligence chatbots were asked to generate original images of five everyday objects, such as pants, a ball, a tire, a fork, or a toothbrush.
The human and GAI-generated responses were then measured based on their originality and fluency of ideas.
What made the research particularly interesting was the assessments of creativity were conducted by humans (following the Consensual Assessment technique) and by an AI program designed explicitly for assessing AUT-trained large-language models. This dual approach was meant to ensure a more comprehensive evaluation of creativity, considering human intuition and AI’s analytical capabilities.
Remarkably, when examining the results, researchers found no significant qualitative difference in the creativity of general artificial intelligence chatbots compared to humans.
These findings are pivotal, suggesting that AI’s potential in creative domains is much broader and more profound than previously thought.
The study showed that although 9.4% of humans surpassed the most creative Generative Artificial Intelligence in the tests, GPT-4, the most recent version of ChatGPT, was notably the highest-performing AI model among those evaluated.
GPT-4’s ability to generate original and creative responses to the prompts underscores the rapid advancements in AI. It also strongly indicates that AI’s role in creative processes is not just supplementary but can be central.
Affirming these findings, another study recently published in the Journal of Creativity found that GPT-4 ranked in the top percentile for originality and fluency on the Torrance Tests of Creative Thinking.
Hellbound and Down 