Autonomous Learning

This project takes a canvas as the input to a neural net, we run a model and the output dictates the behavior of the entity, for example entities sending a chain of messages that persist on the previous state. It is possible that AI can decipher, learn and predict the meanings of the world state. Zero Knowledge proofs allow for a way to verify computations done by ML models and are now possible with ezkl.

We do the inference step in zero knowledge otherwise other players would be able to predict the behaviour of the entities.

We use the vanilla template of mud, and the ezkl cli to perform the neural net new entities to train Autonomous Models by creating by combining their evm world state and calculations in zero knowledge, that is not yet fully integrated but will potentially allow games with persistent storage using MUD to form machine learning interactions in directly from their game contracts and spawning zk interactions from that.