New research has revealed that ChatGPT, powered by OPenAI´s GPT-4 large language model (LLM), scored in the top 1% in tests assessing creativity, surpassing the performance of most college students. The study, according to SciTechDaily, suggests that AI's capabilities are increasingly aligning with human creative abilities.
The findings from the University of Montana and partners show that artificial intelligence can compete with the creative abilities of the top 1% of human participants in a standard test for creativity. Dr. Erik Guzik, an assistant clinical professor at UM's College of Business, led the research. The team used the Torrance Tests of Creative Thinking (TTCT) to evaluate creativity, a tool that has been in use for decades.
Methodology and Results
Eight responses generated by ChatGPT were submitted for evaluation. In comparison, answers from a control group of 24 UM students from Guzik's classes were also submitted. These scores were then compared with the results of 2,700 college students nationally who took the TTCT in 2016. Scholastic Testing Service, a provider of standardized tests, scored all submissions while being unaware of any AI involvement. ChatGPT excelled in fluency and originality, placing it in the top percentile. However, it ranked slightly lower, at the 97th percentile, for flexibility.
Dr. Guzik commented on the results, stating, “For ChatGPT and GPT-4, we showed for the first time that it performs in the top 1% for originality.” He also mentioned that some UM students scored in the top 1%, but ChatGPT outperformed the majority of college students nationally.
Presentation and Interpretation
The research was presented at the Southern Oregon University Creativity Conference. Guzik emphasized their cautious approach, saying, “We were very careful at the conference to not interpret the data very much. We just presented the results. But we shared strong evidence that AI seems to be developing creative ability on par with or even exceeding human ability.”
When asked about the implications of ChatGPT performing well on the TTCT, Guzik shared the AI's response: “ChatGPT told us we may not fully understand human creativity, which I believe is correct,” he said. “It also suggested we may need more sophisticated assessment tools that can differentiate between human and AI-generated ideas.”
TTCT Test Insights
The TTCT test uses real-life creative tasks as prompts. Guzik gave an example on how this works, saying “Let's say it's a basketball. Think of as many uses of a basketball as you can.” He expressed surprise at ChatGPT's ability to generate original ideas, which is typically a hallmark of human imagination.
Guzik's interest in creativity dates back to his school days in Palmer, Massachusetts, where he was introduced to the Future Problem Solving process by Ellis Paul Torrance, the creator of the TTCT. This experience sparked his passion for brainstorming and human imagination. Guzik and his team decided to test ChatGPT's creativity after observing its unexpected and novel responses.
Reflecting on the rapid advancements in AI, Guzik anticipates it becoming a crucial tool in the business world, driving innovation. He believes that creativity is about doing things differently and sees potential in AI's role in fostering creative thinking in business and innovation processes. Guzik concluded by emphasizing the importance of integrating AI education in the UM College of Business curriculum, acknowledging the inevitable role of AI in the future.