Abstract
Under current copyright law, the criteria for determining copyright infringement of generative AI are assessed the same as for human work. This paper analyzes the criteria for determining copyright infringement of AI-generated work, focusing on 'technical alternatives' for expressing 'human creative contribution'.
The results of a comparative analysis based on cases between the work of linguistics and imagery suggest that only areas 'creatively contributed by humans' can be protected as work. However, for representative examples like AI-generated audio work faces limitations in determining 'human contribution' due to voice modulation, synthesis, and tone changes among various instruments.
Moreover, given that audio works are limited in parts preferred by humans and can be formalized through mathematical algorithms, it is necessary to distinguish AI-generated work from human work. Therefore, to clarify the limitations in determining copyright infringement of AI audio work, it is necessary to introduce technical elements as an improvement.
Firstly, as a technical alternative, it is possible to use technology that identifies the AI area for AI-generated work. However, this identification system is focused on text detection, limiting its application to audio works. As an additional technical alternative, it is possible to convert various attributes of audio into text words and analyze the meaningful structural similarity between words.
Lastly, to utilize this, it is necessary to apply the technology by integrating an identifier and a similarity analyzer as a 'plug-in', a concept of linked service with ChatGPT, to predict the possibility of infringement. This paper aims to improve the limitations in determining copyright infringement criteria for AI work.
The results of a comparative analysis based on cases between the work of linguistics and imagery suggest that only areas 'creatively contributed by humans' can be protected as work. However, for representative examples like AI-generated audio work faces limitations in determining 'human contribution' due to voice modulation, synthesis, and tone changes among various instruments.
Moreover, given that audio works are limited in parts preferred by humans and can be formalized through mathematical algorithms, it is necessary to distinguish AI-generated work from human work. Therefore, to clarify the limitations in determining copyright infringement of AI audio work, it is necessary to introduce technical elements as an improvement.
Firstly, as a technical alternative, it is possible to use technology that identifies the AI area for AI-generated work. However, this identification system is focused on text detection, limiting its application to audio works. As an additional technical alternative, it is possible to convert various attributes of audio into text words and analyze the meaningful structural similarity between words.
Lastly, to utilize this, it is necessary to apply the technology by integrating an identifier and a similarity analyzer as a 'plug-in', a concept of linked service with ChatGPT, to predict the possibility of infringement. This paper aims to improve the limitations in determining copyright infringement criteria for AI work.
| Translated title of the contribution | Limitations and Improvement Tasks in judging Copyright Infringement of Artificial Intelligence(AI) Generated Works - Focusing on AI Generated Audio - |
|---|---|
| Original language | Korean |
| Pages (from-to) | 149-182 |
| Number of pages | 34 |
| Journal | 선진상사법률연구 |
| Issue number | 106 |
| State | Published - 2024 |