Training Foundation Models as Data Compression: On Information, Model Weights and Copyright Law

TL;DR

This paper proposes viewing foundation model training as a form of data compression, analyzing its implications for information theory and copyright law, and exploring legal challenges and solutions.

Contribution

It introduces a novel perspective of training as data compression, linking model weights to data representation and discussing legal implications for copyright.

Findings

Model weights can be seen as a compressed form of training data

The approach highlights legal challenges related to copyright and model reproduction

Information-centric methods offer potential solutions to legal issues

Abstract

The training process of foundation models as for other classes of deep learning systems is based on minimizing the reconstruction error over a training set. For this reason, they are susceptible to the memorization and subsequent reproduction of training samples. In this paper, we introduce a training-as-compressing perspective, wherein the model's weights embody a compressed representation of the training data. From a copyright standpoint, this point of view implies that the weights can be considered a reproduction or, more likely, a derivative work of a potentially protected set of works. We investigate the technical and legal challenges that emerge from this framing of the copyright of outputs generated by foundation models, including their implications for practitioners and researchers. We demonstrate that adopting an information-centric approach to the problem presents a promising pathway for tackling these emerging complex legal issues.