🔬 Research Summary by Simone Grassini, an Associate Professor in Psychology at the University of Bergen, Norway. Some of his research has been recently focused on understanding the interaction between humans and AI.
[Original paper by Mika Koivisto & Simone Grassini]
Overview: In the paper, we attempted to measure the creative behavior of AI chatbots using a psychological test designed to test creativity in humans. The study results show that, on average, AI chatbots achieve a better score than our human sample. However, the best performers among humans produced scores that were higher than any response produced by the AI models.
Introduction
Exploring the creative capability of artificial intelligence, a study recently published in Scientific Reports illuminates the captivating interplay between human ingenuity and computational creativity. Researchers Mika Koivisto and Simone Grassini utilized the Alternate Uses Task (AUT) as a medium to explore divergent thinking, a cognitive process wherein individuals formulate varied ideas or solutions in response to a singular task. Encompassing responses from 256 human participants and three advanced AI chatbots—ChatGPT3, ChatGPT4, and Copy.Ai—the study scrutinized creative alternatives for commonplace objects such as a pencil and a candle. Delving into semantic distance and creativity as measured through computational and human evaluators, respectively, the findings unveiled a nuanced narrative: while chatbots, on average, surpassed human scores in semantic distance and creativity, the peak of human creativity ultimately eclipsed the AI-generated responses in seven of the eight scoring categories. Thus, while AI chatbots demonstrably weave a tapestry of ideas that align with, or even surpass, average human creativity, the best level of human imaginative thought, at least presently, remains unparalleled. This investigative foray invites further exploration into the application and integration of AI within the realm of creative processes, foreshadowing an intriguing symbiosis of human and artificial creative potentials.
Key Insights
Background
Creativity has always been regarded as a unique human attribute, distinguishing us humans from other animal species. However, the rapid advancements in Artificial Intelligence (AI) technologies, especially generative AI chatbots, have begun touching upon the creative domain, once thought exclusive to humans. These AI chatbots have demonstrated their ability to produce high-quality artwork, which has sparked intriguing discussions surrounding the fundamental differences between human and machine creativity.
Methodology
Attempting to quantify the creativity of AI, a study employed a well-established measure of creative thinking known as the Alternate Uses Task (AUT). This task was utilized to draw a comparative analysis between the creative capacities of 256 human participants and three modern AI chatbots. The AUT required participants to propose uncommon and inventive uses for everyday objects, gauging their divergent thinking ability—a crucial aspect of creativity.
Findings
On a broader scale, AI chatbots seemed to outshine human participants in generating creative ideas. Their responses were generally more original and innovative compared to the average human participant. However, it was observed that the most creative ideas generated by human participants exceeded for quality rating those produced by AI chatbots. This finding underscores the fact that while AI can mimic or even surpass human creativity on a general scale, the best of human creativity remains out of reach by AI.
Implications
The findings of this study harbor significant implications for both the fields of AI and human creativity. On the one hand, it opens up opportunities where AI can be utilized as a potent tool to augment human creativity, especially in scenarios where divergent thinking is essential. On the other hand, it also accentuates the intricate nature of human creativity that remains elusive to AI technology. This balance between machine and human creative behavior opens the door for a collaborative future where both entities can complement each other in pursuing innovative solutions and creative endeavors.
Between the lines
From the insights provided in the article, it’s important to acknowledge that AI is advancing to a level of sophistication capable of mimicking human creativity. However, it’s crucial to acknowledge the potential substantial differences between these two forms of creativity, originating from the disparate operational mechanisms of the human brain and generative AI. While divergent thinking is often explored as an important factor of creativity, it is not the complete picture of the multifaceted nature of creative behavior.
The advancements highlighted in the article are noteworthy from a scientific perspective, marking a significant milestone in technological evolution. Yet, the article touches on themes important in light of ethical and societal concerns. The advent of the widespread use of AI to undertake tasks presently exclusive to humans raises profound questions. Nonetheless, it’s important to recognize that a machine excelling in a particular cognitive task (like the divergent thinking tasks as in the article) doesn’t necessarily translate to machines evolving into artists or competing with humans in jobs demanding a wide array of creative aptitudes.
In a future where humans will use AI agents to improve their performance regarding creative skills, it is the most likely scenario when talking about complex tasks, at least for the current stage of the technology.
Furthermore, the study has certain limitations, as acknowledged by the authors. The creative behavior exhibited by AI chatbots may merely reflect retrieving responses for the specified test from their training database sans a genuine task elaboration. This behavior, although indicative of memory capacity, doesn’t necessarily constitute genuine creative engagement. Through this lens, the discourse around AI’s creative capabilities invites a deeper exploration, fostering a balanced understanding of its potential and constraints.