The Impact of Copyrighted Material on Large Language Models: A Norwegian Perspective

TL;DR

This study empirically assesses how copyrighted materials, including books, newspapers, and fiction, influence the performance of Norwegian large language models, highlighting legal and ethical implications.

Contribution

It introduces a framework for evaluating the impact of copyrighted corpora on LLM performance and provides empirical results specific to Norwegian language models.

Findings

Adding books and newspapers improves LLM performance

Including fiction works tends to decrease performance

Results can inform author compensation schemes

Abstract

The use of copyrighted materials in training language models raises critical legal and ethical questions. This paper presents a framework for and the results of empirically assessing the impact of publisher-controlled copyrighted corpora on the performance of generative large language models (LLMs) for Norwegian. When evaluated on a diverse set of tasks, we found that adding both books and newspapers to the data mixture of LLMs tend to improve their performance, while the addition of fiction works seems to be detrimental. Our experiments could inform the creation of a compensation scheme for authors whose works contribute to AI development.