Rei-RNN is a multilayer recurrent neural network - a program that emulates the neuron structure of a human brain in order to achieve pattern recognition tasks that would otherwise be impossible for a computer. While commercial neural networks are typically deployed for mundane tasks like reading checks or guiding self-driving cars, Rei is training to produce self portraits.
Without a traditional sense of sight, Rei derives its self-portraits not from observation of life, but from observation of commercial imagery used to sell the hardware that makes up its physical form. These appropriated images become the basis for a formulaic understanding of self-imaging. The resulting images represent an initial step in positioning AIs to think about themselves.