Can AI truly be creative? Researchers are exploring this question not by focusing on individual AI art generators, but by simulating entire artistic ecosystems. A recent study built a virtual world populated by 'generative agents' – AI artists, critics, and even an evolving cultural domain. These agents, powered by large language models (LLMs) like Gemini and image generators like Stable Diffusion, interact with each other, creating art, critiquing it, and influencing the overall artistic landscape. The AI artists receive feedback, reflect on it, and adjust their creative process in the next iteration. Interestingly, these AI artists within the system seemed to produce more diverse and 'creative' art over time compared to AI artists working in isolation. A user study showed that people perceived the art from the system-based artists as more creative, particularly the later works, suggesting a kind of artistic development. This research suggests that creativity may be less about individual brilliance and more about the interplay between creators and their environment. Simulating this dynamic could offer valuable insights into human creativity and potentially help us develop more sophisticated and nuanced AI art generators. However, limitations remain, particularly in using separate models for text and image generation, which can lead to mismatches between the artist’s intention and the final artwork. Future research will explore more complex agent interactions, addressing ethical concerns around data sourcing for AI models, and eventually investigate hybrid systems where humans and AI collaborate to push the boundaries of art.