I’ve made a new prototype for the Design Research Playground. PThinkTank explores using spatial elements to prompt an LLM, and then optionally visualising what the LLM produces. The question behind it is whether moving concepts (known here as ‘fish’) around inside a space (known here as a ‘tank’) could be an intuitive way of interacting with an LLM.
Fish are just keywords, and the higher a fish sits in the tank, the more it shapes what the LLM does with your starting thought. The most fun way to use it is ‘fisheye’ mode, where the output is fed into an image generator and turned into a visual. The work plays with the possibilities of near real-time AI models, and with spatial metaphors for interacting with them.
For want of a better word, it’s a bit bamboozling, but a bit of play and you should get the idea. There are a couple of easter eggs hidden in there too.
Try PThinkTank here. Good luck.