The authors have presented a promising approach to using multi-modal LLMs in dental diagnosis and treatment, but the main issue lies in the lack of clarity on how these systems will be validated and their real-world applicability. I was wondering if the authors could share their thoughts on the challenges of data privacy, validation of synthetic data, and the accuracy of LLM-based diagnoses, especially in complex cases?
Also, the feasibility of integrating these systems into existing dental practices, particularly in low-resource settings, and the computational costs involved, have not been discussed well in my opinion.