A groundbreaking study out of the University of Montreal has put artificial intelligence to the test against human ingenuity, and the results paint a clear picture: machines can now outpace the typical person in basic creative tasks, but they stumble when facing the sharpest human minds.
The research, published today in Scientific Reports, gathered data from 100,000 people across English-speaking countries like the United States, Canada, and the United Kingdom. Participants faced a simple challenge—list 10 words as unrelated as possible. Creativity got measured by semantic distance: how far apart the words were in meaning. Think “microscope, volcano, whisper” scoring better than “car, dog, tree.”
Nine top AI models entered the ring, including OpenAI’s GPT-4, Google’s GeminiPro, and others like Claude and Vicuna. GPT-4 came out ahead of the average human score. GeminiPro held its own at the middle mark. Smaller, open-source models like Vicuna even beat some pricier giants, proving size isn’t everything in the AI world.
But here’s where humans shine: no AI cracked the top 10% of human performers. Machines leaned on repetition—GPT-4 tossed “microscope” into 70% of its lists and “elephant” into 60%. Humans? Their most common word, “car,” showed up in just 1.4% of responses. AI outputs felt safe, predictable.
Researchers tweaked the AI’s “temperature” setting to crank up randomness, cutting repetition and boosting scores. GPT-4 climbed above 72% of humans after the adjustment. Still, in creative writing tests—haikus, movie plots, short stories—AI delivered solid work but lacked the raw variety humans brought, especially in poetry and summaries.
“The persistent gap between the best-performing humans and even the most advanced LLMs indicates that the most demanding creative roles in industry are unlikely to be supplanted by current artificial intelligence systems,” the paper states.
This isn’t just academic trivia. It speaks to real-world shifts. Goldman Sachs reported last year that AI could displace tasks equivalent to 300 million full-time jobs globally, with creative fields feeling the heat. The World Economic Forum estimates 85 million jobs lost to AI by 2026, though new roles might emerge. In creative industries, demand for entry-level positions like writers and graphic artists dropped 28-33% from 2024 to 2025, per a Coherent Market Insights analysis. AI handles scale and speed, but oversight remains human territory.
Look at the numbers: an MIT and Boston University report predicts two million manufacturing jobs gone by 2026, many involving creative problem-solving. Yet executives from companies like Salesforce and Autodesk argue workers won’t vanish—they’ll evolve into “directors” overseeing AI agents. “Instead of asking, ‘How do I accomplish the goal?’ they’ll focus on, ‘What are the goals that I want to accomplish, and then how do I delegate those goals to AI?'” as one leader put it.
Skeptics see a darker side. Tech giants pour billions into AI, optimizing for efficiency over originality. Newer models like GPT-4-turbo scored worse on creativity than older versions, likely due to cuts for speed and cost. Is this progress, or a race to the bottom where machines churn out bland content to slash payrolls? Creative workers already report job insecurity—over two-thirds in a BBC-cited survey feel AI threatens their livelihoods. Novelists, artists, videographers: many have pivoted, like one illustrator who turned to books on drawing after AI undercut commissions.
Education feels the ripple too. A recent Inside Higher Ed survey found 90% of faculty believe generative AI erodes students’ critical thinking, with 78% noting more cheating. The Chronicle of Higher Education echoes this: tools like ChatGPT foster dependency, potentially devaluing degrees.
Harvard Business Review research adds nuance—AI amplifies creativity for those who question and refine its outputs, not just accept them. Passive users see flat results; active ones iterate for breakthroughs. A University of Connecticut study reinforces that: “Participants who were more creative without AI also tended to perform better when collaborating with AI.”
The Montreal team’s framework, now public, lets anyone benchmark future AI. As models advance, that human edge—true originality, emotional depth—stays vital. Machines mimic, but the spark of genius? That’s ours to keep, if we value it enough to protect the jobs and skills that nurture it. In a world rushing toward automation, remembering what makes us irreplaceable might be the smartest move yet.
Image generated by Artificial Intelligence.
