AI vs. Human Creativity: New Study Compares LLMs & 100,000 People

A new study reveals AI systems, like GPT-4, can surpass average human creativity on specific tasks, but top humans still outperform AI, especially in complex creative activities. AI is best positioned as a powerful creative assistant, not a replacement.

When scientists took a look at the most innovative fifty percent of participants, their typical scores exceeded those of every AI design checked. The void grew also larger amongst the top 10 percent of one of the most innovative people.

Researchers examined several leading big language models, including ChatGPT, Claude, Gemini, and others, and compared their performance with arise from greater than 100,000 human individuals. The searchings for highlight a clear juncture. Some AI systems, including GPT-4, surpassed typical human scores on tasks created to determine different etymological creative thinking.

Measuring Creativity: The Divergent Association Task

To assess creativity fairly across people and makers, the study group made use of several techniques. The key device was the Divergent Association Task (DAT), an extensively made use of emotional test that gauges divergent imagination, or the capacity to create original and diverse ideas from a single prompt.

Professor Karim Jerbi led the research study, with Antoine Bellemare-Pépin (Université de Montréal) and François Lespinasse (Université Concordia) working as co-first writers. The study team likewise included Yoshua Bengio, owner of Mila and LoiZéro, and a leader of deep learning, the technology behind modern AI systems such as ChatGPT.

Efficiency on this task is strongly linked to outcomes on various other well-known imagination tests used in composing, concept generation, and creative trouble resolving. The job is language-based, it goes well past vocabulary. It involves broader cognitive processes associated with creative thinking across numerous domains. The DAT also has functional benefits, as it takes only two to 4 minutes to complete and can be accessed online by the general public.

Created by research co-author Jay Olson, the DAT asks participants, whether human or AI, to detail ten words that are as unrelated in meaning as feasible. An example of an extremely creative feedback includes words like “galaxy, fork, freedom, algae, harmonica, quantum, nostalgia, velour, cyclone, photosynthesis.”

The scientists after that explored whether AI success on this easy word association job might include much more sensible and complicated innovative activities. To test this, they contrasted AI systems and human individuals on innovative creating obstacles such as making up haiku (a brief three-line poetic kind), composing flick plot recaps, and producing short stories.

AI as a Creative Assistant: The Future Outlook

Instead of indicating the end of creative professions, the searchings for suggest a future where AI serves as a creative aide. By increasing concepts and opening up brand-new paths for expedition, AI might assist magnify human creativity rather than change it.

Copyright 1995-2026 ScienceDaily or by various other events, where suggested. All rights controlled by their corresponding proprietors. Web content on this website is for information just. It is not intended to supply other or medical professional suggestions. Sights revealed right here do not always reflect those of ScienceDaily, partners or contributors. Financial backing for ScienceDaily originates from ads and reference programs.

Dec. 9, 2023– A research study group has actually concentrated on generative AI as an information-gathering tool in the clinical field. The outcomes showed that of the generative AIs suggested primarily fictitious referrals, while the …

Unveiling the Core Study and Participants

The paper titled “Divergent imagination in human beings and big language designs” was published in Scientific Reports on January 21, 2026. The study brought together researchers from Université de Montréal, Université Concordia, University of Toronto Mississauga, Mila (Quebec AI Institute), and Google DeepMind.

Can generative artificial intelligence systems like ChatGPT really produce original ideas? A new research led by Teacher Karim Jerbi from the Division of Psychology at the Université de Montréal, with involvement from prominent AI scientist Yoshua Bengio, takes on that question at an extraordinary range. The research is the biggest direct comparison ever performed in between human creativity and the imagination of huge language designs.

The study supplies a balanced perspective on worries that artificial intelligence could change creative professionals. While AI systems can now match or exceed typical human creativity on particular tasks, they still have clear constraints and rely upon human direction.

Shaping AI Creativity: Temperature and Prompts

At lower temperature level settings, AI produces safer and more traditional results. At greater temperatures, actions come to be a lot more diverse, much less foreseeable, and extra exploratory, enabling the system to relocate beyond acquainted ideas.

These results highlight that AI imagination depends heavily on human assistance, making interaction and motivating a main component of the creative process.

“Although AI can currently get to human-level imagination on particular examinations, we require to relocate beyond this misleading sense of competition,” states Teacher Karim Jerbi. “Generative AI has most importantly become a very powerful device in the solution of human creative thinking: it will certainly not change creators, yet exceptionally transform exactly how they envision, explore, and produce– for those who choose to use it.”

Is AI creativity repaired, or can it be formed? The research reveals that creativity in AI can be changed by changing technical settings, particularly the model’s temperature.

University of Montreal. (2026, January 25). Researchers tested AI against 100,000 human beings on creativity. ScienceDaily. Gotten January 29, 2026 from www.sciencedaily.com/releases/2026/01/260125083356.htm

The researchers also discovered that creativity is strongly affected by just how instructions are composed. Triggers that urge models to think concerning word origins and structure making use of etymology lead to even more unforeseen associations and greater imagination ratings. These results stress that AI creative thinking depends greatly on human advice, making communication and triggering a main part of the imaginative process.

Research Insights: AI’s Strengths and Limitations

The research, released in Scientific Information (Nature Profile), indicates a significant shift. Generative AI systems have now reached a degree where they can outmatch the ordinary human on specific imagination steps. At the very same time, one of the most innovative people still reveal a clear and consistent benefit over even the greatest AI designs.

“We created a strenuous structure that permits us to compare human and AI imagination utilizing the very same tools, based on data from more than 100,000 individuals, in partnership with Jay Olson from the College of Toronto,” claims Teacher Karim Jerbi, that is also an associate teacher at Mila.

July 31, 2023– The artificial intelligence language version GPT-3 carried out as well as college students in solving particular reasoning problems like those that show up on standard tests. The scientists that performed …

A substantial new research study comparing more than 100,000 people with today’s most advanced AI systems provides a surprising result: generative AI can now defeat the average human on specific creative thinking tests. Models like GPT-4 showed solid performance on jobs developed to gauge original reasoning and idea generation, sometimes outperforming common human reactions. There’s a clear ceiling. The most creative human beings– specifically the top 10%– still leave AI well behind, specifically on richer imaginative job like verse and narration.

Generative AI systems have actually now reached a level where they can outperform the ordinary human on certain imagination measures. Some AI systems, including GPT-4, gone beyond average human scores on tasks made to determine different etymological creative thinking.

May 30, 2024– Cultivating creativity in colleges is important for a future driven by expert system (AI). But while instructors embrace creativity as a crucial 21st century ability, a lack of legitimate and reliable …

A substantial brand-new research contrasting more than 100,000 individuals with today’s most sophisticated AI systems delivers an unexpected outcome: generative AI can now defeat the typical human on specific imagination examinations. The study is the largest straight contrast ever before carried out between human creative thinking and the creativity of large language models.

“Our study reveals that some AI systems based on large language designs can now outshine ordinary human creative thinking on distinct jobs,” clarifies Professor Karim Jerbi. “This result might be unexpected– also upsetting– but our research study likewise highlights a similarly important observation: even the best AI systems still fall short of the levels gotten to by the most innovative human beings.”