Yes we use feedback and de-ID'ed data to improve ChatGPT– like every LLM company (xcancel.com)

In a recent update, OpenAI clarified its practices concerning user data and model improvement, specifically addressing concerns raised about the use of member data in their Navier-Stokes research. OpenAI confirmed that while no personal user data was reviewed during this particular study, they do utilize user feedback and de-identified data to enhance the performance of ChatGPT and Codex, a common practice across the large language model (LLM) industry. This transparency highlights the importance of user interactions in refining AI systems, aligning with ongoing discussions about ethical data usage in AI development. This announcement is significant for the AI/ML community as it sheds light on the complexities of data utilization in model training and improvement. By acknowledging the use of de-identified data, OpenAI reinforces the balance between leveraging user interactions for development while ensuring user privacy. Furthermore, the interplay of collaborative research efforts and the competitive landscape within AI development underscores the need for transparency in how models evolve, a critical factor for maintaining trust among users and contributing to the broader scientific dialogue in artificial intelligence.