In AI vs. Humans, AI Can ‘Surpass’ and By Far | AAO 2025

Many people worry about artificial intelligence (AI) replacing humans. Alfredo A. Sadun, M.D., Ph.D., has a different take.

“It’s not replacing humans. It's surpassing us by far,” said Sadun, chief of

ophthalmology at UCLA’s Doheny Eye Institute, said in an interview with Managed Healthcare Executive (MHE).

Sadun was the chair and moderator of a session on artificial intelligence this morning at the American Academy of Ophthalmology meeting in Orlando, Florida.

In this interview with MHE, Sadun gave two examples of AI surpassing humans and, more particularly, human ophthalmologists. Doheny is a major center for the reading of fundus imaging, so Google approached the institute about using its huge supply of images, Sadun explained. After working with the images, Google’s engineers said they could establish the gender of the patients from the fundus image 97% of the time. Sadun said he could only hope to be accurate 50% of the time if faced with the same task. There are some caveats: the patients have to be past puberty, and the eyes couldn’t have pathology; the Google AI didn’t work if there was a lot of macular degeneration, Sadun said.

AI can “do things we can't do, do it instantly, and do it very well,” Sadun said.

The other example that Sadun discussed with MHE was the detection of papilledema in the emergency room, which can be an indication of brain swelling and therefore the need for a CT or MRI to look for, for example, bleeding in the brain or even a tumor. “Typically, in an emergency room, they can't tell whether there's papilledema or not, but maybe AI can,” Sadun said.

A recent retrospective study compared the ability of AI to differentiate between true cases of papilledema and those that were just “funny-looking [optic] discs” to the ability of two skillful neuropathologists at Emory University, Nancy Newman, M.D., and Valerie Biousse, M.D., Sadun explained. AI did a little bit better, but the big difference was speed, with AI sorting through the images in a matter of seconds compared to the hours spent by Newman and Biousse.

Neuro-ophthalmologists are in short supply. Sadun spoke about AI filling that void so patients with serious brain conditions don’t get turned away “because their headache is just a headache because there’s nobody there who is a neuro-ophthalmologist.” But he also conceded that there are practical issues, such as who is going to dilate the pupil so a good photograph of the eye is obtained. There are also social, political, and legal issues that need to be worked out, he said.