Neural Network Quine

TL;DR

This paper introduces methods for building and training neural networks capable of self-replication, learning to output their own weights, and explores the trade-offs between replication and task performance.

Contribution

It presents a novel approach to self-replicating neural networks, including a regeneration training method and a design that balances replication with auxiliary tasks.

Findings

Self-replicating networks can be trained using regeneration and optimization.

There is a trade-off between replication ability and task performance.

Training biases towards task specialization over replication.

Abstract

Self-replication is a key aspect of biological life that has been largely overlooked in Artificial Intelligence systems. Here we describe how to build and train self-replicating neural networks. The network replicates itself by learning to output its own weights. The network is designed using a loss function that can be optimized with either gradient-based or non-gradient-based methods. We also describe a method we call regeneration to train the network without explicit optimization, by injecting the network with predictions of its own parameters. The best solution for a self-replicating network was found by alternating between regeneration and optimization steps. Finally, we describe a design for a self-replicating neural network that can solve an auxiliary task such as MNIST image classification. We observe that there is a trade-off between the network's ability to classify images and its ability to replicate, but training is biased towards increasing its specialization at image classification at the expense of replication. This is analogous to the trade-off between reproduction and other tasks observed in nature. We suggest that a self-replication mechanism for artificial intelligence is useful because it introduces the possibility of continual improvement through natural selection.