PARASITES OR BOOSTERS OF HUMAN CREATIVITY?: A MODEL TO SOLVE COPYRIGHT ISSUES OF TRAINING DATASETS OF ARTIFICIAL INTELLIGENCE BEFORE ALPHAART DEFEATS HUMAN ARTISTS

摘要

Advances in AI have been affecting almost all industries, including the creative industry. The current creative AI technology now generates high quality creative content, some of which has achieved significant financial success. As creative AI gets more sophisticated and produces better quality outputs, human creators may lose not only income, but also the incentive to create. A legal problem with creative AI is that in order to generate creative content, the AI system must be fed with similar or identical creative content that may be under copyright protection; however, even if creative AI is fed with copyrighted content without authorization of copyright holders, they cannot claim their rights because it is extremely difficult for them to check the data of creative AI under the current system.The Good Data Model proposed in this Note can help resolve both data opacity and copyright infringement issues with creative AI's training datasets. Via data certification, a license for gathering online content, and a suspicious AI reporting system, this Model would successfully ease tensions and build trust between copyright owners and creative AI developers, leading them to achieve super-creativity and furthering the objectives of copyright law.