The Development of AI-Generated Videos in 2023 and the Challenges Faced by NFT Marketplaces

The Development of AI-Generated Videos in 2023 and the Challenges Faced by NFT Marketplaces

In recent years, there has been significant progress in the development of AI technology that can generate videos based on text inputs. This has raised questions about the potential applications and limitations of this technology. In this article, we will explore the current state of AI-generated videos and the challenges faced by NFT marketplaces in relation to copyright liability.

AI-generated videos are primarily created using autoregressive models or diffusion models. Autoregressive models, such as Phenaki, TATS, and NUWA-Infinity, generate long videos by training models on short video clips and using inference to generate the rest of the video in a left-to-right manner. On the other hand, diffusion models, like MCVD, FDM, and LVDM, use a different approach to generate videos. These models employ multiple local diffusion models, which allow for parallel inference and faster generation of long videos. Additionally, the hierarchical structure of these models enables training directly on long videos, eliminating the gap between training and inference.

However, there are challenges associated with generating coherent and realistic videos using these methods. The reliance on short video clips for training and inference can lead to unnatural and distorted transitions between scenes. The lack of training data for long videos can result in incoherent storylines and inconsistent frames. These limitations highlight the need for further advancements in AI technology to improve the quality of generated videos.

While AI-generated videos pose technical challenges, NFT marketplaces face legal challenges in relation to copyright liability. The Digital Millennium Copyright Act (DMCA) offers safe harbors to internet service providers, protecting them from copyright liability if certain conditions are met. However, it is unclear if NFT marketplaces can invoke these safe harbors. Secondary liability, which includes contributory liability and vicarious liability, is the primary way traditional marketplaces can be held liable for copyright infringement.

Some NFT marketplaces claim to fall under the protection of DMCA safe harbors, while others do not. To invoke the safe harbor, a marketplace must have language in its terms of service or use a similar method to remove infringing material. By explicitly expressing compliance with the DMCA safe harbor rule, NFT marketplaces can potentially protect themselves from copyright liability. However, the specific safe harbor relevant to NFT marketplaces, section 512(c) for passive storage, poses challenges. If a marketplace exerts control over the sale and storage of NFTs or selects specific tokens for promotion, it may not qualify for protection under section 512(c).

Even if NFT marketplaces do not fall under DMCA safe harbors, it does not automatically make them liable for copyright infringement. Potential plaintiffs must still prove contributory liability, vicarious liability, or both. When purchasing an NFT, buyers are acquiring the digital token and a content license that outlines their rights and permissions related to the associated content or artwork. Therefore, the ownership of an NFT does not grant ownership of the underlying intellectual property.

In conclusion, the development of AI-generated videos in 2023 has made significant progress but still faces challenges in terms of generating coherent and realistic videos. NFT marketplaces, on the other hand, need to navigate the complexities of copyright liability and the applicability of DMCA safe harbors. To mitigate legal risks, NFT marketplaces should express compliance with DMCA safe harbor rules in their terms of service. Additionally, marketplaces should clarify that the purchase of an NFT does not grant ownership of the underlying intellectual property. By addressing these issues, both AI-generated videos and NFT marketplaces can continue to evolve and thrive in the digital landscape.

Actionable Advice:

1. AI developers should focus on improving the coherence and realism of generated videos by addressing the limitations of training on short video clips.
2. NFT marketplaces should ensure they comply with DMCA safe harbor rules, including the inclusion of relevant language in their terms of service.
3. Buyers of NFTs should be aware that owning an NFT does not grant ownership of the intellectual property associated with the artwork or content.