• Followers
About this project

OpenBrain's central goal is to introduce a new framework and philosophy for recurrent neurocomputation. By requiring that all neurons act asynchrynously and independently, we introduce a new metric for evaluating the universal intelligence of continuous time agents. We prove representation and universal approximation results in this context which lead to a set of learning rules in the spirt of John Conway's game of life. Finally evaluate this framework against different intelligent agent algorithms by implementing an approximate universal intelligence measure for agents embedded in turing computable environments in Minecraft, BF, and a variety of other reference machines.

Project Members