Creative AIs just aren’t capable of human-like creativity… right?
Many would like to think so. After all, if machines can truly do everything human professionals can, do we even need people for the more creative aspects of any given workflow?
Researchers from Germany and the UK tested this assumption by comparing human ideas with those generated by six Generative Artificial Intelligence (GAI) chatbots, namely:
- ChatGPT (versions 3 and 4)
- and YouChat
Creativity as the last frontier
As the researchers point out in their paper, Artificial Intelligence has already outpaced human intelligence and capability on several fronts. AI is great at performing repetitive tasks without getting bored or making mistakes, which is useful for manufacturing, data entry, and quality control. It can quickly analyze large amounts of data, identify patterns, and make financial, healthcare, and marketing predictions.
AI can perform precise tasks such as surgery, where even a small mistake can have serious consequences. It has a great memory, too; AI can store and retrieve vast amounts of data quickly and accurately, which is useful for tasks such as language translation, speech recognition, and image recognition.
But can AI really be creative? Many believe that human creativity is the last frontier – the one place we continue to excel.
Recommended reading: What is ChatGPT? An Introduction to OpenAI’s Conversational AI
However, in their study ‘Artificial muses: Generative Artificial Intelligence Chatbots Have Risen to Human-Level Creativity,’ researchers Jennifer Haase and Paul H. P. Hanel declared, “We found no qualitative difference between AI and human-generated creativity, although there are differences in how ideas are generated.”
Interestingly, the study used both human and AI judges to assess the output of 100 humans and 5 GAI chatbots. As the report explains, the AUT requires the generation of multiple original uses for five everyday objects (pants, ball, tire, fork, toothbrush), which can also be called prompts. It demonstrates good predictive validity and is one of most often used creativity tests available.
Generative AIs vs. humans: key study findings
Haase and Hanel found that:
- None of the GAI chatbots emerged as more original than the other four across all five prompts.
- Chatbots are creative, as their output was judged as such by humans and AI and indistinguishable from human output.
- When chatbots are asked the same simple question as humans, they generate more ideas, which are, on average, as original as ideas generated by humans.
- The motivation to engage with a specific creative task and problem understanding must come from the human interacting with the tool.
- Although there was no qualitative difference between AI and human-generated creativity, 9.4% of humans were more creative than the most creative GAI, GPT-4.
The research paper is a good read, particularly the ‘Discussion’ section in which Haase and Hanel tackle two major arguments around AI creativity:
1. GAI is missing (so far) a connection to the real world, with emotions and imagination, and is therefore incapable of true creativity.
2. Humans aren’t exceptional at bringing forth world-changing ideas and largely act as knowledge bases, drawing on our learned experience and somewhat limited context for ideation.
This makes the finding that just 9.4% of humans are more creative than generative AIs all the more fascinating. Those few super-creatives could be an incredibly valuable asset to companies in the years ahead, as the bar for content quality, experiences, and engagement continues to rise.
Read the full paper here.