Recent AI developments have made it possible for AI to auto-generate content—text, image, and sound. Highly realistic auto-generated content raises the question of whether one can differentiate between what is AI-generated and human-generated, and assess its origin and authenticity. When it comes to the processes of digital scholarship and publication in the presence of automated content generation technology, the evolution of data storage and presentation technologies demand that we rethink basic processes, such as the nature of anonymity and the mechanisms of attribution. We propose to consider these issues in light of emerging digital storage technologies that may better support the mechanisms of attribution (and fulfilling broader goals of accountability, transparency, and trust). We discuss the scholarship review and publication process in a revised context, specifically the possibility of synthetically generated content and the availability of a digital storage infrastructure that can track data provenance while offering: immutability of stored data; accountability and attribution of authorship; and privacy-preserving authentication mechanisms. As an example, we consider the MetaScribe system architecture, which supports these features, and we believe such features allow us to reconsider the nature of identity and anonymity in this domain, and to broaden the ethical discussion surrounding new technology. Considering such technological options, in an underlying storage infrastructure, means that we could discuss the epistemological relevance of published media more generally.
No takes yet. Share an insight, caveat, or question.
Israel et al. (2022) studied this question.