Donnerstag, 27.02.2025
New publication by Prof Dr Dornis and Prof Dr Stober in English
Generative AI, TDM and the phenomenon of training data memorization
This article highlights key findings on generative models, addresses copyright concerns related to training data memorization, and suggests how ISMIR can promote fair practices in generative AI for all stakeholders.
Prof Dr Tim Dornis and Prof Dr Sebastian Stober:
"In a recent interdisciplinary tandem-study, we have argued in detail that this is actually not the case because generative AI training fundamentally differs from TDM. In this article, we share our main findings and the implications for both public and corporate research on generative models. We further discuss how the phenomenon of training data memorization leads to copyright issues independently from the "fair use" and TDM exceptions. Finally, we outline how the ISMIR could contribute to the ongoing discussion about fair practices with respect to generative AI that satisfy all stakeholders."
"Parts of this article are based on an interdisciplinary study on generative AI and copyright law in Germany and the EU conducted by the authors in 2024 and funded by the German Authors‘ Rights Initiative (Initiative Urheberrecht). The contents of the resulting report (Dornis and Stober, 2024) as well as this article exclusively reflect the views and assessments of the authors."
The text was submitted as an overview article to the Transactions of the International Society for Music Information Retrieval.
The TISMIR paper was published as a pre-print on arXiv.
You can find it here: https://doi.org/10.48550/arXiv.2502.15858
picture taken in September 2024 during the presentation of the study
Pressekontakt: info@urheber.info