In 1961, chemist Otto Wichterle realized that a disc of hydrogel material could be shaped with different front and back surface curvatures by spinning the mold during polymerization—and thus the soft contact lens was born. He did so by building a makeshift contraption from an Erector set and a record player’s motor.  In 1964, ophthalmologist Charles Kelman dreamed up phacoemulsification while at the dentist. The ultrasonic probe used to clean his teeth inspired him to create a similar device that could break up an opacified crystalline lens in a controlled manner.  In 1991 a laser engineer, a satellite communications expert and an ophthalmologist pooled their differing areas of expertise to collaboratively develop optical coherence tomography.

These are just three examples of the rich history of innovation that made eye care what it is today. In each, success was achieved through a combination of inspiration, ingenuity and just plain chutzpah. You know what doesn’t have any of those? Your computer.

As in so many other walks of life, eye care seems ready to be disrupted by artificial intelligence. The topic has exploded this year, with a series of journal articles showing ChatGPT and other AI tools generally did well in replicating the work of cornea, glaucoma and retina specialists  in diagnosis, clinical decision-making and patient education. This news fills some doctors with dread. The conversation around AI’s forthcoming impact in eye care is oddly bipolar: hype in some corners, hand-wringing in others. Reality surely lies somewhere in between.

I think we’ve all seen by now that AI-generated content is a bit wild and wooly. The most public recent example was Google Gemini’s heavy-handed push for diversity in its photo creation tools, leading to some hilarious and cringe-worthy results. But others can come from an AI process that scrapes data from the internet and uses it as a source. Last fall, a chatbot on the discussion board Quora inexplicably answered yes to the question, “Can you melt an egg?” and this foolish answer was picked up by Google, which started repeating it. The ensuing online discussion then lent it credence and gave it more reach. This kind of feedback loop is just one way that misinformation propagates online.

So, AI clearly has some growing pains to go through before it’ll be ready for prime time. Pop culture has trained us to expect AI to be either amazingly useful or unrelentingly evil.  “Gullible and inept” is kind of a new one. Any implementation of AI in eye care that uses public input, like queries about—and, good lord, users’ own explanations of—eye diseases and surgeries is going to be a bit of a mess. More useful will be tools built upon valid datasets to help simplify disease diagnosis for professionals. I could see those being hugely helpful in developing nations where there’s a dire lack of doctors and equipment. Here at home will be a different story. There’s too much institutional inertia in the American healthcare system for AI to run wild.

AI tools will eventually do amazing things in eye care. But they’ll never give us contact lenses, phaco or OCT. That takes human audacity. They also won’t be able to make eye contact with a worried patient and reassuringly explain a tricky diagnosis. Chatbots may be able to spit out a paragraph on a disease, but they can’t help a patient cope and move on. Only you can do that.