> Microsoft AI Diagnostic Orchestrator (MAI-DxO) correctly diagnoses up to 85% of NEJM case proceedings, a rate more than four times higher than a group of experienced physicians.

> Clinicians in our study worked without access to colleagues, textbooks, or even generative AI, which may feature in their normal clinical practice.

1. As I understand, it's very common for doctors to fall back on reference material in their practice, especially for the most complex cases. If all access to resources was cut off (as seems to be implied by the second quote), the comparison seems somewhat unfair.

2. What were the publication dates of the case records? I can't find this information, and it makes a difference if the NEJM case studies were in the LLMs' training data.