Welcome to the “Artificial Intelligence” column, where various OD experts in this area will answer common questions being asked by their colleagues.
Thanks to an onslaught of recorded data and technological advances in interpreting such data, every industry is clamoring to take advantage of Artificial Intelligence (AI). The eyecare industry is no different.
Here, I define AI and how it relates to optometry.
What Is It?
AI is a “machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions influencing real or virtual environments,” according to the National Artificial Intelligence Initiative Act of 2020.
There are 3 main categories of AI: (1) Narrow AI, (2) General AI, and (3) Super AI.1 Narrow AI performs a specific and restricted activity. A mainstream example of this category is OpenAI’s
ChatGPT, whose task is exclusively text-based. Narrow AI is the only AI that currently exists.
General AI is the idea for AI to learn from its previous activities to perform new activities that are related to a different circum-stance without any human intervention.
Super AI is the concept of AI having cognitive abilities, such as reasoning, that outperform humans.1
Under the Narrow AI category are 2 types (1) Reactive Machine AI, and (2) Limited Memory AI. The former performs a specific and restricted activity sans any memory. The second type uses a designated amount of memory to make decisions to arrive at a desired result.1
How Does It Relate to Optometry?
There are several examples of both Reactive Machine AI and Limited Memory AI in optometry. Regarding the former, think autorefractors, frame-customizing technology, and some teleoptometry platforms. An example of the latter is corneal topography, which quantifies the different colors to compare them to a patient database to, for example, tell the OD whether a patient is at an elevated risk for keratoconus. Other examples of Limited Memory AI include screening/monitoring devices for diabetic retinopathy, and a virtual assistant.
Why All the Excitement?
The technology has substantially evolved and expanded into other technologies. Specifically, deep-learning algorithms, or those that tap into multiple-layered neural networks to assess data and prognosticate, have improved diagnostic accuracy, sometimes rivaling that of human ODs. Additionally, AI has become a part of teleoptometry to facilitate remote screenings, and electronic health records to expedite administrative tasks, as examples.
Wanted: Interconnectivity
With all of AI’s advances, interconnectivity is needed. Specifically, technology is required to enable the sharing of millions of patient databases among various technologies to best predict disease risk and the most appropriate treatment.
Reference
- IBM. Understanding the different types of artificial intelligence. https://www.ibm.com/think/topics/artificial-intelligence-types. (Accessed May 15, 2025)
