Meet My Artificially-Intelligent Virtual Self: Creative Avatars, Machine Learning, Smart Contracts and the Copyright Conundrum
by Eugene C Lim
Artificial intelligence (AI) technologies have triggered a dramatic paradigm shift in how we conceive of authorship and creation. Intelligent machines, such as those powered by the new GPT-3 neural network technology, are capable of generating human-like creative expressions, composing text, performing translations and producing other creative outputs once thought to be beyond the ability of computers.
This article focuses on intelligent applications (or “creative avatars”) that are programmed to replicate the style of a human author (such as Shakespeare, Rembrandt or J.K. Rowling), and the regulatory challenges flowing from the generation of such works. The challenges surrounding the production of such content relate not only to questions of ownership and authorship, but extend to issues of how, if at all, these works can be treated as copyrightable subject matter.
While much of the literature in the field has proposed significant modifications to traditional copyright rules to accommodate the emergence of AI-generated content, this article highlights the limitations of relying on copyright law in regulating rights in AI-generated derivative works. In developing this argument, the article adopts a novel approach by suggesting that interim solutions, in the form of alternative business and technological models, can be found outside of the “copyright box”. In particular, the article proposes a consent-based contract framework, featuring digital watermarks, Creative Commons licences and blockchain technologies, as part of an interim solution to regulate rights in relation to works generated by “creative avatar” programs. It is suggested that existing contractual and technological tools, drawn from current practices in the software licensing and cryptocurrency industries, can be useful to stakeholders in the AI industry, especially in the early years of emerging neural network technologies. In this regard, the article offers a utilitarian justification for the proposed consent-based framework by explaining how it can help to facilitate the dissemination of AI-generated derivative works in the absence of clearly-defined copyright rules, and promote the eventual enrichment of the public domain.
[This is an Authors' Take post, which provides readers with an insight into current IP scholarship, featuring preliminary comments and thoughts from authors of articles accepted for publication in forthcoming issues of the Journal of Intellectual Property Law & Practice (OUP). The full text of this contribution will be made available on Advance Access soon]