OpenAI’s announcement of new image generation features in ChatGPT this week was quickly met with an onslaught of Studio Ghibli style anime images.

What started as an attempt to make new ‘My Neighbor Totoro’ out of selfies quickly became the ‘anime-zation’ of movie stills, cultural moments and darker images. The move not only reportedly melted the company’s GPUs, it also highlighted the need for anchors in how users handle AI’s limitless potential with limited imagination. 

OpenAI’s policy blocks specific artists in prompts, but allows references to studios (which is particularly sensitive in this case due to Ghibli’s director and his opinion on AI). This policy gives users enough reference to start to shrink the limitless potential of AI by all making the same thing.

Human behavior has lots of examples of our similar tastes when presented with unlimited options. Highly customized products often simplify down to pre-set packages of options. Spotify has shown that when given the world’s music, we often choose the same songs until heavily promoted to explore. In travel, we say we want something different, but we often go to the same destinations.   

Does this indicate how we will use AI long term? Is the limitation on AI’s creative potential not technological, but psychological? Or will users’ imaginations and vocabulary develop alongside the technology – finding a day where the GPUs melt outside of creative fads? We blame AI for the looming ‘enshittification’ of content, but is it what many of us actually make when given the choice?