A new study highlighted by Neuroscience News reveals a surprising paradox: while AI systems can generate responses that appear highly creative, their outputs are far more repetitive and predictable than human thinking overall. Researchers tested humans against multiple large language models (LLMs) like GPT, Gemini, and Llama on creativity tasks (e.g., coming up with unusual uses for objects) and found that AI responses were significantly more similar to each other than human responses.
Interestingly, individual AI responses often scored as equally or even more creative than the average human answer when judged in isolation. However, when comparing many outputs together, a pattern emerged: AI systems repeatedly produced the same kinds of “creative” ideas. This phenomenon is called the “sameness paradox”—AI can seem creative at a glance but lacks diversity at scale.
The study also explored whether increasing randomness (known as “temperature”) could improve creativity. While higher randomness did produce more varied outputs, it quickly led to nonsensical or low-quality responses, showing that AI has a limited ability to balance originality with coherence. This suggests that current AI systems may have a built-in ceiling when it comes to true imagination.
Overall, the findings raise an important concern: relying heavily on AI for brainstorming or creative work could narrow human thinking over time. Unlike humans, AI lacks personal experiences, emotions, and unique perspectives—the core drivers of originality. As a result, while AI can assist creativity, overdependence on it may lead to a world where ideas become more uniform rather than diverse.
