MNIST to Memory

The project aims to research & reconsider NFTs as not only static assets, but as dynamic generators of new token assets.

In this new model, the NFT provenance chain representation can begin to create points or “structures of geometric relations” that aligns user contribution of model updates, towards increasing technical individuation from generative model preferences. NFT models can become personalized, in order to construct a type of globally accessible intelligence or brain.

In the future, the project concept might be used with personal data off-chain to train, learn, and mint new tokens from decentralized stored models.

This project uses react and a forked version of an in-browser generator network to create images from a pre-trained model (found here: https://github.com/mwdchang/tfjs-gan.git). The model can be updated, and minted to Zora protocol, to be transferred to another person, so they can mint additional tokens from the model.