Companies like OpenAI, Meta, and Google are in a race to gather data for training their AI models. However, some industry leaders suggest that relying solely on publicly available data could lead to copyright lawsuits. Instead, they advocate for using synthetic data, which can be generated artificially and offers more control and privacy. Synthetic data is seen as a safer alternative that can facilitate the development of generative AI. However, it has limitations, such as the inability to create perfect data with no existing information and challenges in terms of privacy and accuracy with limited data.

Public data also presents challenges, including missing crucial specifics and regulatory restrictions on data access. The focus should shift towards privacy-first approaches and responsible data-driven design for sustainable AI development. Synthetic data can play a role in democratizing data access within organizations, providing value and fostering innovation. The future lies in smaller, specialized AI models targeted at specific tasks, as well as regulations that quantify AI risks. This shift aligns with the principles of differential privacy and synthetic data, emphasizing targeted and efficient development rather than relying on massive data collection.