AI Will Never Fit Into A Licensing Regime
The article argues that traditional licensing models are fundamentally mismatched with AI training processes, rendering current attempts to compensate artists through data licensing ineffective and potentially distracting. Because AI models extract abstract concepts rather than storing or reproducing individual images, the contribution of any single artist is infinitesimally small. This dilution makes it nearly impossible to calculate fair royalties or assign ownership based on influence, especially since the "black box" nature of neural networks prevents precise attribution of how specific artworks shaped the final model. Furthermore, establishing new rights for artistic influence or style is legally and practically problematic, as it could inadvertently hinder legitimate creative expression and imitation. The author contends that focusing on these complex licensing frameworks overlooks the primary threat to artists: the displacement by AI tools that are cheaper, faster, and more accessible. Instead of trying to force traditional copyright structures onto a system where the medium of exchange is computational patterns rather than direct copies, the conversation is misdirected away more pressing issues regarding the value of human creativity. Ultimately, the most viable path forward involves artists adapting to AI as a tool while leveraging copyright protections against low-effort, fully automated commercial outputs. By denying copyright to works lacking significant human authorship, the industry can prevent large corporations from monopolizing AI-generated content and undercutting professional creators. This approach encourages artists to integrate AI into their workflows, maintaining their competitive edge through skill and human intent, rather than relying on unsustainable micro-licensing agreements that fail to address the systemic economic shift caused by generative technology.
Source: techdirt.comPublished on 2023-06-03