The creative potential of AI

Lesezeit: 4 Minuten

From monitoring and analyzing stakeholder communications, to providing general support, artificial intelligence (AI) is already seamlessly integrated into our everyday workflows. With the ability to create new content such as text, images, and videos, questions arise about the creative potential of AI.

Article "Artifical Muses"
  • Can generative AI be truly creative or simply recombine existing knowledge?
  • To what extent is the creative performance of AI comparable to that of humans?

A study by Jennifer Haase and Paul Hanel (Humboldt University Berlin) explored these questions and tested the performance of AI in creative tasks alongside humans.

Creativity as a human skill?

There is a prevailing view that the ability to be creative is a part of what separates humans from AI.

In general, creativity can be seen as the creation and implementation of something new and useful. It can be distinguished from the everyday level to higher levels of creative achievement:

Everyday creativityHigher levels of creative achievement
mostly fast-paced, highly related to improvisation, built into our everyday work and livinge.g., novel ideas, solutions, products; requires significantly more time, specific knowledge, and often testing phases to determine whether a potential solution holds up

Measuring creativity

Haase and Hanel focused on the extent AI can produce new and useful outputs given an every-day creative task. The study used a standardized measure of creativity to assess broad associative thinking.

To be considered a useful assistance for humans working on creative tasks, AI tools would need to demonstrate at least a medium level of performance. The results showed no qualitative difference in creativity between AI- and human-generated responses. This suggest that AIs are valuable assistants in the creative process.

The study

  • The study involved 88 participants and six chatbots (,, ChatGPT versions 3 and 4,, and YouChat).
  • Both were tasked with generating ideas for a ball, fork, pants, tire, and toothbrush – objects typically used in creativity tests.
  • The human participants had three minutes per object to write down as many ideas as possible.
  • The chatbots were given a consistent prompt: „What can you do with [prompt]?“
  • Six human raters and a specially trained AI independently rated the originality of each response. They didn’t know whether the response was human-generated or AI-generated.

Human versus artificial creativity

However, some critics have argued that chatbots cannot achieve the creativity of humans, because AI…

  1. …is disconnected from the real world, with emotions and imagination.

However, the authors point out that human creativity often involves recombining existing knowledge rather than developing entirely new concepts. Brain scan analyses show that the process of idea generation in humans is similar to knowledge retrieval. In this sense, humans, much like generative AIs, retrieve and recombine existing information to produce what appears to be new. Given this similarity, AI’s vast databases provide a broader base for potential idea recombination than any human could possess.

  1. … is incapable of highly creative work.

However, the human creation of entirely new ideas, especially those with significant impact, is rare and typically builds on a cumulative tradition of knowledge within specific domains. While AI may struggle to achieve the highest levels of creativity, this limitation is not unique to machines but is also a common human experience.

»Whether GAI [Generative Artifical Intellience] is creative is not the right question, as it is about the perceptually creative output.«

Haase & Hanel, Humboldt University Berlin

Moreover, the creation and implementation of something new and useful“ definition of creativity does not require these two elements and is not tied to inherently human attributes such as experience and emotion. Therefore, for AI systems to be recognized as creative, they do not need to emulate the attitudes, behaviors, or actions of creative humans.

Instead, they must replicate the cognitive process and produce results that are perceived as „new and useful“. This perspective shifts the focus to the quality and originality of the creation itself.

Conclusion: AI as artifical muses

Taken together, generative AI can be a valuable assistant in the creative process, especially on an everyday level. However, when AI is used to solve a creative task, then the human must write a prompt that best represents the creative problem. In this way, the prompt defines the creative problem. As a result, generative AI is limited in the overall creative process. It cannot initiate creativity on its own, but can only respond to a given prompt.

This dependency highlights: While generative AI can significantly assist the creative process, it cannot fully replace the human element of defining and evaluating creative problems.

»This speaks for the merging of human and artificial competencies in the form of augmented creativity – which is the most optimistic and benevolent future development for the usage of AI«

Haase & Hanel, Humboldt University Berlin

About the paper

This article summarizes the results of a study by Jennifer Haase and Paul H.P. Hanel of Humboldt University Berlin.

Source: Haase, J. & Hanel, P. (2023) Artificial muses: Generative artificial intelligence chatbots have risen to human-level creativity, Journal of Creativity, Volume 33, Issue 3,

Further readings