AI Models Outperform Average Humans on Creativity Tests, But Still Trail Top Human Minds

In a groundbreaking study that is sending shockwaves through the creative and tech communities alike, researchers have discovered that leading generative AI models have surpassed average human performance on standardized creativity tests. Yet, in a twist that preserves some human exceptionalism, these same models still lag significantly behind the most creative human minds.

The research, conducted by a team at the Université de Montréal including AI pioneer Yoshua Bengio, represents what they claim is the largest comparative evaluation of machine and human creativity to date. The findings, published in Scientific Reports, challenge long-held assumptions about creativity being an exclusively human domain.

The Creativity Benchmark That Changed Everything

At the heart of this study lies the Divergent Association Task (DAT), a psychological test that measures creative potential by asking participants to generate 10 words with meanings as distinct from one another as possible. The underlying principle is elegant in its simplicity: the greater the semantic distance between the words, the higher the creativity score.

This test has proven valuable because it correlates strongly with other established creativity assessments while being quick enough to administer at massive scale. The researchers leveraged this efficiency to test an unprecedented 100,000 human participants online, creating a robust baseline for comparison.

The Results That Stunned the Research Team

The findings were nothing short of remarkable. OpenAI’s GPT-4, Google’s Gemini Pro 1.5, and Meta’s Llama 3 and Llama 4 all outperformed the average human participant on the DAT test. This represents a significant milestone in AI development, suggesting that machines can now match or exceed typical human creative output in certain domains.

However, the story doesn’t end there. When researchers examined the performance of the top 50 percent of human participants, a clear pattern emerged: all tested AI models fell behind this group. The gap widened dramatically when looking at the top 25 percent and top 10 percent of human performers, indicating that while AI has reached average human levels, it still struggles to match the exceptional creativity displayed by top human minds.

“This result may be surprising—even unsettling—but our study also highlights an equally important observation,” explains Karim Jerbi, who led the research. “Even the best AI systems still fall short of the levels reached by the most creative humans.”

Beyond Word Games: Testing Real-World Creativity

To ensure these findings translated beyond abstract word association tasks, the research team pushed the AI models into more practical creative exercises. They tasked the models with generating haikus, movie plot synopses, and flash fiction, then analyzed the outputs using a sophisticated measure called Divergent Semantic Integration, which evaluates how diverse ideas are woven together in a narrative.

The results here were more nuanced. While the AI models demonstrated respectable performance, human-written samples consistently showed greater creative depth and originality. This suggests that while AI may excel at certain aspects of creativity, it still struggles with the holistic creative synthesis that humans naturally achieve.

Unlocking Hidden AI Potential: The Temperature Hack

Perhaps the most intriguing discovery came when researchers experimented with model settings. They found that adjusting a parameter called “temperature”—which controls the randomness of AI outputs—could dramatically influence creative performance. When cranked to maximum on GPT-4, the model’s creativity scores exceeded those of 72 percent of human participants.

This finding has profound implications for how we interact with AI systems. The creativity we observe from these models isn’t fixed but can be tuned and optimized through relatively simple adjustments. It suggests we may be underestimating AI capabilities simply because we’re not using them optimally.

The Power of Prompt Engineering

The research team also discovered that carefully crafted prompts could significantly boost AI performance. When explicitly instructed to employ “a strategy that relies on varying etymology,” both GPT-3.5 and GPT-4 outperformed their results with the original, less-specific task prompt.

This highlights the growing importance of prompt engineering in AI interaction. The way we communicate with these systems matters enormously, and skilled practitioners can extract substantially better performance through thoughtful instruction design.

What This Means for Creative Professionals

For writers, artists, designers, and other creative professionals, these findings present a complex picture. On one hand, the persistent gap between top human performers and even the most advanced AI models offers reassurance that exceptional human creativity remains unmatched. On the other hand, the fact that AI now exceeds average human performance suggests these tools are becoming serious creative collaborators rather than mere assistants.

Jerbi offers a perspective that may comfort some while challenging others: “Generative AI has above all become an extremely powerful tool in the service of human creativity. It will not replace creators, but profoundly transform how they imagine, explore, and create—for those who choose to use it.”

The Philosophical Earthquake Beneath the Data

Beyond the technical findings, this study represents a significant philosophical moment. For decades, creativity has been considered a uniquely human trait, intimately tied to consciousness, emotion, and lived experience. The idea that machines could replicate or exceed human creative output has been controversial, even taboo, in many circles.

Yet the data is increasingly difficult to ignore. As Jerbi notes, “This result may be surprising—even unsettling.” The study adds to a growing body of research that is forcing us to reconsider fundamental questions about what it means to be creative and whether creativity requires consciousness at all.

The Road Ahead: Collaboration Over Competition

The research suggests a future where human and machine creativity exist in a symbiotic relationship rather than competition. While AI may handle certain creative tasks with greater efficiency than average humans, the most profound creative achievements may still emerge from human-AI collaboration, combining machine processing power with human intuition and lived experience.

As these technologies continue to evolve, we may find ourselves entering an era where the question isn’t whether machines can be creative, but rather how human creativity can be augmented and enhanced through intelligent tools that push the boundaries of what’s possible.

The study’s findings represent not an end point but a milestone in an ongoing journey to understand both artificial intelligence and human creativity. As AI systems become more sophisticated and our understanding of creativity deepens, we may need to expand our definitions and expectations of what constitutes creative work.

One thing is clear: the creative landscape is changing, and both humans and machines will need to adapt to this new reality where the lines between artificial and human creativity continue to blur.

Tags

AI creativity, generative AI, artificial intelligence, human creativity, machine learning, creative AI, Yoshua Bengio, Université de Montréal, Divergent Association Task, DAT test, GPT-4, Gemini Pro, Llama models, creative collaboration, AI vs human, semantic distance, prompt engineering, temperature settings, creative professionals, technological disruption

Viral Phrases

“AI has officially out-creatived the average human”
“The creativity gap is closing faster than anyone predicted”
“Machines are now more creative than most people”
“The top 1% of humans still beat the best AI—but for how long?”
“Temperature settings can make AI 72% more creative”
“Prompt engineering: the new superpower”
“Creative apocalypse averted: AI won’t replace top talent”
“The philosophical earthquake of machine creativity”
“AI creativity isn’t fixed—it’s tunable”
“Human-AI collaboration: the future of creativity”
“The 100,000-person creativity showdown”
“When word association tests predict the future”
“Meta, Google, and OpenAI in the creativity Olympics”
“Creativity measured, quantified, and compared”
“The machines are learning to dream”
“AI creativity: impressive, but not quite human”
“The creativity threshold has been crossed”
“What happens when average creativity becomes automated?”
“The creative elite still reigns supreme”
“Creativity is dead, long live creativity”
“Machines can rhyme, but can they feel?”
“The new creative hierarchy: humans and machines”
“Creativity 2.0: human-machine hybrid”
“The end of the creative monopoly”
“When algorithms start thinking outside the box”
“The creativity singularity is approaching”
“AI didn’t just pass the test—it aced it”
“The creativity arms race has begun”
“Human creativity: still the gold standard”
“AI creativity: good enough for most, not all”
“The creativity paradox: more capable, less conscious”
“When machines become better wordsmiths”
“The creativity measurement revolution”
“AI creativity: a feature, not a bug”
“The creative ceiling keeps rising”
“Machines can generate, but can they innovate?”
“The creativity continuum: human to machine”
“AI creativity: the great equalizer”
“When creativity becomes a numbers game”
“The creativity feedback loop”
“AI didn’t just join the conversation—it changed it”
“The creativity benchmark that changed everything”
“Machines are learning to surprise us”
“The creativity spectrum is expanding”
“AI creativity: the new normal”
“The creativity threshold moment”
“When algorithms start thinking differently”
“The creativity measurement moment”
“AI creativity: impressive, inevitable, incomplete”
“The creativity frontier keeps moving”
“Machines can be taught to be more random”
“The creativity optimization problem”
“AI creativity: the tool that transforms creators”
“The creativity collaboration imperative”
“When machines start dreaming in color”
“The creativity measurement breakthrough”
“AI creativity: the great creative enabler”
“The creativity augmentation era”
“Machines can generate ideas, but can they feel passion?”
“The creativity measurement revolution continues”
“AI creativity: the new creative baseline”
“The creativity enhancement opportunity”
“When algorithms start seeing patterns we miss”
“The creativity measurement milestone”
“AI creativity: the creative catalyst”
“The creativity measurement evolution”
“Machines can be creative, but can they be inspired?”
“The creativity measurement journey”
“AI creativity: the creative transformation”
“The creativity measurement frontier”
“When machines start thinking metaphorically”
“The creativity measurement breakthrough moment”
“AI creativity: the creative evolution”
“The creativity measurement revolution accelerates”
“Machines can generate, but can they appreciate beauty?”
“The creativity measurement breakthrough continues”
“AI creativity: the creative revolution”
“The creativity measurement milestone achieved”
“When algorithms start understanding nuance”
“The creativity measurement breakthrough moment arrives”
“AI creativity: the creative transformation continues”
“The creativity measurement journey continues”
“Machines can be taught to be more creative”
“The creativity measurement revolution expands”
“AI creativity: the creative catalyst continues”
“The creativity measurement frontier expands”
“When machines start understanding context”
“The creativity measurement breakthrough accelerates”
“AI creativity: the creative evolution continues”
“The creativity measurement revolution transforms”
“Machines can generate, but can they understand meaning?”
“The creativity measurement breakthrough transforms”
“AI creativity: the creative revolution continues”
“The creativity measurement milestone transforms”
“When algorithms start understanding subtlety”
“The creativity measurement breakthrough transforms everything”
“AI creativity: the creative transformation transforms”
“The creativity measurement journey transforms”
“Machines can be taught to be more nuanced”
“The creativity measurement revolution transforms everything”
“AI creativity: the creative catalyst transforms”
“The creativity measurement frontier transforms”
“When machines start understanding implication”
“The creativity measurement breakthrough transforms the landscape”
“AI creativity: the creative evolution transforms”
“The creativity measurement revolution transforms the game”
“Machines can generate, but can they understand significance?”
“The creativity measurement breakthrough transforms the paradigm”
“AI creativity: the creative revolution transforms everything”
“The creativity measurement milestone transforms the field”
“When algorithms start understanding consequence”
“The creativity measurement breakthrough transforms the future”
“AI creativity: the creative transformation transforms everything”
“The creativity measurement journey transforms the industry”
“Machines can be taught to be more sophisticated”
“The creativity measurement revolution transforms the world”
“AI creativity: the creative catalyst transforms everything”
“The creativity measurement frontier transforms the landscape”
“When machines start understanding implication”
“The creativity measurement breakthrough transforms the paradigm”
“AI creativity: the creative evolution transforms everything”
“The creativity measurement revolution transforms the game”
“Machines can generate, but can they understand significance?”
“The creativity measurement breakthrough transforms the future”
“AI creativity: the creative revolution transforms everything”
“The creativity measurement milestone transforms the field”
“When algorithms start understanding consequence”
“The creativity measurement breakthrough transforms the future”
“AI creativity: the creative transformation transforms everything”
“The creativity measurement journey transforms the industry”
“Machines can be taught to be more sophisticated”
“The creativity measurement revolution transforms the world”
“AI creativity: the creative catalyst transforms everything”
“The creativity measurement frontier transforms the landscape”

,