In the past, major generative AI tools were primarily trained on data that was freely available on the internet. However, the landscape is changing as sources of training data become more restricted and require licensing agreements. This shift has led to the emergence of new licensing startups that aim to provide a steady flow of source material for AI development.
A significant development in the AI industry is the formation of the Dataset Providers Alliance (DPA), a trade group consisting of seven AI licensing companies. These companies, such as Rightsify and Pixta, advocate for an opt-in system where data can only be used with explicit consent from creators and rights holders. This represents a departure from the opt-out systems used by many major AI companies, placing the ethical responsibility on data owners to give permission.
The shift towards an opt-in system has been praised by industry experts for its fairness to creators. Alex Bestall, CEO of Rightsify, believes that selling publicly available datasets without consent can lead to legal issues and damage the credibility of AI companies. Ed Newton-Rex, of the nonprofit Fairly Trained, agrees and emphasizes the importance of considering the rights of creators in AI data licensing practices.
While the DPA’s opt-in approach is praised for its ethics, there are concerns about its practicality. Shayne Longpre, from the Data Provenance Initiative, highlights the challenge of sourcing large volumes of data ethically while also adhering to the opt-in standard. He warns that this could lead to data scarcity or high costs, potentially limiting access to data for smaller players in the AI industry.
In its position paper, the DPA rejects government-mandated licensing in favor of a free-market approach where data originators and AI companies negotiate directly. The alliance also offers guidelines for fair compensation structures, including subscription-based models and outcome-based licensing, to ensure that creators and rights holders are appropriately compensated for their data. This approach is seen as a way to foster a more equitable relationship between data creators and AI developers in the era of rapidly evolving technology.
The evolution of AI data licensing represents a complex and rapidly changing landscape. The emergence of the Dataset Providers Alliance and its advocacy for ethical and fair data practices marks a significant step towards ensuring that creators are respected and compensated for their work. While challenges remain in implementing opt-in systems and negotiating fair compensation, the industry is moving towards a more transparent and ethical approach to utilizing data for AI development. It is crucial for all stakeholders to work together to cultivate a sustainable and equitable future for AI data licensing.
Leave a Reply