The text “Turning Budget Proposals Into Faces (Noh Masks: Omote)” was published in mirror.xyz
In the summer of 2020, we were developing a play to generate “faces” based on the “meaning” of the budget proposal, using BERT, a large-scale language model at the time.
Budget allocations are a good indicator of the nature of a group. But it is difficult to grasp the whole picture and how they are connected only by numbers and graphs.
Wouldn’t it be interesting to summarize the complex and difficult-to-grasp nature of the budget plan in a human “face”? This play was born from the idea.
However, while we were thinking about how to control the “meaning” of the “face” and the constraints (what details of the face should be connected to what properties of the budget, how much reason and basis for the connection should be explainable, or whether it would be more interesting if it could not be explained at all, etc.), we were so busy that we missed the opportunity to make it public. We missed the opportunity.
Then we decided to try it again with Stable Diffusion, which is currently booming.
One of the reasons we decided to give the previous stuff another try is that we found that Stable Diffusion is able to create reasonably appropriate images from abstract prompts (sentences).
In other words, to some extent, our hope was that it might automatically create a correspondence between the appropriate meaning (budget plan) and the image (face).
And that brings us to the “Noh mask”….
(To continue, see embedded link below.)
https://mirror.xyz/vection.eth/m_ZZwrvLrCOyxie6ybAqDg_sBdy8KGSLP7BcP239PJo
