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

To improve accuracy, we labeled 150,000 ophthalmology images with input from seven ophthalmologists, reducing variance in scoring.

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

Deep learning applications at Google include Google Photos, which effectively utilizes computer vision to understand the content of images.

There is significant progress in AI applications in healthcare, particularly in image recognition tasks, which are becoming increasingly effective and reliable.

Innovative technologies, like AI for diabetic retinopathy screening, face regulatory hurdles due to the lack of appropriate coding, which delays their implementation in healthcare.

In the medical field, Deep Research can help doctors quickly find relevant literature and clinical trials, saving them time and improving patient care.

Our approach to AI in medicine involves using smaller, specialized models trained on peer-reviewed literature that outperform larger models on specific medical tasks, which was a novel insight at the time of development.