Our model for detecting diabetic retinopathy now performs comparably to the median of eight US board-certified ophthalmologists, which is significant given the shortage of specialists in many regions.

In the medical domain, we applied deep learning to ophthalmology, specifically to detect diabetic retinopathy in retinal images.

In collaborations with hospitals, we focus on developing algorithms that maximize the quality of high-resolution images from limited data, particularly in medical imaging.

AI can analyze medical images, but establishing a reliable ground truth for diagnoses can be challenging due to differing opinions among experts.

We reduced the false positive rate from over 40% to 0.38% in our demo system, which is a significant improvement.

AI tools have been FDA-cleared and are already augmenting radiologists in tasks like diabetic retinopathy diagnosis.

AI can help create a new gold standard in medical diagnosis by providing consistent results where human experts may disagree.

In a sequential manner, we adapt our objective while processing images, rather than feeding all images at once, which allows for better optimization.

OpenAI's recent paper shows that models can effectively grade medical answers based on specific criteria, indicating their potential in other domains.