Key points are not available for this paper at this time.
Regarding the issue of using copyrighted works as training data, there are two conflicting claims. One is that free use should be legalized for the development of the AI industry, and the other is that copyrighted works should not be used without permission due to reducing incentives for creation. Generative AI provides literary and artistic outputs much faster and cheaper based on learning from human works, putting human authors at risk of being replaced. In this situation, it is time to consider whether it is reasonable to justify free riding on the works that the creator achieved with a great deal of effort in the name of ‘the social benefit’. Rather than applying fair use doctrine or legislating TDM exception, an substantial solution is required that can maintain the balance between copyright holders, AI developers, and users, such as finding appropriate rewards system. In this paper, I suggest to introduce ‘Extended Collective Licensing (ECL) System’ as a new solution to resolve problems caused by use of copyrighted works without permission and establish a copyright trading system based on blockchain technology. Copyright licensing is an effective mechanism to ensure legitimate rewards to copyright holders, provide legal stability to users and developers of AI services, and accelerate industrial development. Because there are numerous copyright holders worldwide for works that have been used or may be used for machine learning purposes, the adoption of an Extended Collective Licensing System can be said to be the most logical response in the AI era. This system will manage transparently copyrighted works used in AI machine learning and distribute fair compensation guaranteed by copyright law.
Sangmi Lee (Sun,) studied this question.