AI Surpasses Average Humans in Creativity Tests, Study Reveals

A groundbreaking study has revealed that generative artificial intelligence (AI) can outperform the average human in specific creativity tests. Conducted by a research team led by Professor Karim Jerbi from the Université de Montréal, the study compared over 100,000 human participants with advanced AI systems like GPT-4, Claude, and Gemini. The findings indicate a significant shift in the landscape of creativity, with AI now capable of excelling in tasks designed to measure original thinking and idea generation.

The research stands out as the largest direct comparison between human creativity and that of large language models to date. While some AI systems have shown superior performance on defined creativity tasks, the most creative individuals still maintain a distinct edge, particularly in more nuanced creative endeavors such as poetry and storytelling.

AI Achieves New Levels of Creativity

The results point to a clear turning point in the capabilities of AI. According to the study, certain AI models have surpassed average human scores in tasks that assess divergent linguistic creativity. Professor Jerbi notes, “Our study shows that some AI systems based on large language models can now outperform average human creativity on well-defined tasks.” This finding, while surprising to many, underscores the ongoing limitations of AI in achieving the heights of human creativity.

To evaluate creativity across both humans and machines, the research team employed multiple assessment methods. The primary tool used was the Divergent Association Task (DAT), a psychological test designed to measure divergent thinking—the ability to generate a variety of solutions from a single prompt. Participants were asked to list ten nouns that are as unrelated in meaning as possible, allowing researchers to assess the breadth of their creative responses.

Performance in these tasks is closely linked to results from established creativity tests in writing, idea generation, and problem solving. The DAT’s accessibility, taking only two to four minutes to complete, adds to its appeal in measuring creative capacity among a broad audience.

AI’s Potential in Creative Writing

Researchers further investigated whether AI’s success in simple word association tasks could extend to more intricate creative writing challenges. By comparing the output from AI systems and human participants in tasks such as composing haiku, summarizing movie plots, and crafting short stories, the results revealed a consistent pattern. While AI sometimes surpassed the performance of average human writers, the most skilled human creators produced superior and more original work.

The study also raised important questions about the adaptability of AI creativity. Findings indicate that AI creativity can be influenced by changing technical settings, particularly the “temperature” parameter, which affects the predictability of generated responses. Lower temperature settings yield more conventional outputs, while higher settings allow for greater variation and exploration of ideas.

Additionally, the phrasing of prompts significantly impacts AI creativity. Instructions that encourage models to consider word origins and structures through etymology can lead to unexpected associations and improved creativity scores. This emphasizes the crucial role of human direction in shaping AI-generated content.

The implications of these findings challenge the narrative that AI might replace human creators in the creative sector. While AI can now achieve comparable levels of creativity on specific tasks, it still has limitations and relies heavily on human input. Professor Jerbi states, “Generative AI has above all become an extremely powerful tool in the service of human creativity.” Rather than signaling the end of creative careers, the research suggests a future where AI acts as a creative assistant, expanding ideas and opening new avenues for exploration.

The full study is published in the journal Scientific Reports under the title “Divergent creativity in humans and large language models.” The findings highlight the evolving relationship between humans and AI in the realm of creativity, suggesting that the potential for collaboration could enhance human imagination rather than diminish it.