Published in Non-Clinical

Artificial Intelligence for Optometrists: A Breakdown of Terms

This is editorially independent content
14 min read

Familiarize yourself with commonly used terms relating to artificial intelligence (AI) in eyecare and what optometrists should know about AI.

Artificial Intelligence for Optometrists: A Breakdown of Terms
Artificial intelligence (AI) has been a popular topic over the past several months. From replacing search engines to writing scripts and even diagnosing diseases, AI offers a plethora of possibilities.
Unfortunately, AI is not without risks, and some have warned that, unless we tamper with our AI usage now, we may find ourselves living in our favorite science fiction movies in the not-so-distant future.
AI has several potential applications in eyecare, including research, clinical diagnosis, professional training, and even marketing for private practices. In fact, some practices are already using artificial intelligence to make the grueling task of marketing a lot easier and faster.
Whether you’re an AI enthusiast, skeptic, or have simply glossed over the AI-related conversations that have been taking place in every industry, including eyecare, here is what you need to know about AI in eyecare.

What’s AI anyway?

Artificial intelligence is a branch of computer science that deals with the simulation of human intelligence, including visual perception, decision-making, speech recognition, and language translation. The goal of AI is to approach, and even surpass, the human level of intelligence and decision-making.
There are two proposed classifications for AI to date; one is based on functionality, while the other is on intelligence. The functionality classification breaks AI down into four types: reactive machines, limited theory, theory of mind, and self-awareness.1
Reactive machines or reactive AI respond to specific tasks. For instance, a spam filter that decides whether a piece of email belongs in your inbox is an example of reactive AI. Limited theory AI can store information for future use; semi-autonomous vehicles that can map routes are an example of limited theory AI. Theory of mind takes this one step further and proposes an AI model that can understand the behavior of humans and animals.
Some have hypothesized that ChatGPT is a theory of mind model, but ChatGPT, when queried, responded that it does not have such understanding. Self-awareness in AI is currently a hypothetical concept, but one that is portrayed in science fiction.2

The intelligence model of artificial intelligence

Meanwhile, the intelligence model breaks AI down into narrow AI, general AI, and super AI. Narrow AI, or weak AI, is task-focused and cannot perform outside its task range. Google Translate is a great example of narrow AI; it can translate words and sentences, but it cannot help you generate a conversation. Other examples of narrow AI include image recognition software, Google’s page-ranking system, and ChatGPT.
General AI, which is also known as strong AI and artificial general intelligence (AGI), can learn and understand any tasks that a human can learn and understand. There are no verified examples of true strong AI, but researchers are hard at work towards developing an AI model capable of keeping up with human intelligence. Science fiction has plenty of examples of autonomous AI that can learn, adapt, and integrate with human society.
Like general AI, super AI, or artificial superintelligence (ASI) is also a hypothetical concept. Super AI would go beyond mimicking human behavior and learning. ASI would be self-aware and potentially capable of outperforming humans.1

How does AI work?

AI works by combining data, computing power, and algorithms. The algorithm is a set of step-by-step instructions and rules that a program or computer must follow. The algorithm can be simple, like the one used by autocorrect, or more complex like ChatGPT’s. Computing power is required to process the data in accordance to the algorithm.
The type of data that is fed into the algorithm is likewise important; data that is full of inaccuracies will lead to a poorly functioning AI even with a good algorithm.3

The history of artificial intelligence

Artificial intelligence is not a new concept. One of the first people to broach the topic of artificial intelligence was Alan Turing, a British polymath who explored the possibility of building intelligent machines in his paper Computing Machinery and Intelligence. Written in 1950, Turing used this paper to propose that machines can be taught to reason and make decisions just like humans.
Unfortunately, the computers of Turing’s time were not advanced enough to allow Turing to test his theory; it was also extremely expensive to rent a computer, which was not a household item in the fifties. However, the idea did not go unnoticed, and in 1956, the Logic Theorist, which was founded by the Research and Development (RAND) Corporation, became the first AI research program.4

As computers changed between 1957 and 1974, AI continued to flourish. However, technology at the time was not able to keep up with the theory.

Despite the setbacks, AI continued to evolve with government interest and funding. Over the next several decades, AI continued to thrive quietly and, in 1997, IBM’s Deep Blue, a chess program, defeated grandmaster Gary Kasparov in a highly publicized match.
Then, in 2001, Dr. Cynthia Breazeal, a postdoctoral associate out of MIT’s laboratory, created Kismet, a robot that was socially intelligent and capable of learning.5 Over the past several decades, AI has become a subtle part of our lives, from translating a sentence to using a program to help correct grammar or make suggestions for an email.
AI has recently exploded with the advent of ChatGPT, which has become a go-to for questions despite only having data until 2021. There are plans to update it in the future. Meanwhile, many other companies, including Apple, Meta, and Google, are working on developing their own versions of AI.

AI in the exam room

It should come as no surprise that AI is making advances in healthcare in general and in eye care specifically. Currently, there are already technologies in existence within the eye care industry.

Diabetic retinopathy screening

AI can be extremely valuable as a screening tool, especially for underserved populations. AI models that can detect diabetic retinopathy have been developed over the past few years and, despite some limitations, have performed well. In one study, AI even outperformed general ophthalmologists and retinal specialists alike as an early detection tool for diabetic retinopathy.
The promising results have encouraged further development of AI as a screening and diagnosis tool for other ocular conditions, such as age-related macular degeneration and glaucoma.

Scribing

Medical record keeping is both extremely important and very tedious. The American Medical Informatics Association has recognized the problem recently, reporting that the burden of clinical documentation can have a negative impact not only on the healthcare provider but also on the quality of healthcare that the provider can render.
To help alleviate the problem, several companies have developed AI-driven scribing tools for medical professionals. AI scribes have the potential to help ease the workload that doctors face by transcribing data in real-time.
Unfortunately, AI scribes aren’t perfect yet; one problem that’s been identified is the inability of AI to record clinically relevant non-lexical conversation sounds (NLCS) accurately.  Sounds like “uh-huh” or “uh-uh” can be an important patient response to a question, and as of right now, AI cannot recognize them.6

The future of AI in eyecare

With many technologies being developed, we can only guess what the future holds for AI in eyecare.
Here are some potential future applications of AI:
  • As a screening tool: If we can teach AI to detect diabetic retinopathy accurately, we can also hypothetically teach it to screen for a plethora of other ocular diseases.
  • In diagnostics: Future AI can provide diagnostic options for healthcare professionals based on a combination of symptoms, signs, and test results.
    • However, it is important to note that the final clinical decision will still fall to the clinician, and AI will not replace the critical thinking of an eyecare professional.
  • For patient education: AI may be taught to help patients learn more about their specific conditions in a way that is fun and conversational; it can also help create models, images, and videos to help patients better understand their ocular health.
    • If delivered well, this has the potential to increase compliance and decrease the need to deliver extensive patient education in the office.
  • To teach optometrists of the future: AI models can serve as an important learning tool for future optometry students and, when coupled with virtual reality, can even offer simulations for the practice of advanced techniques such as laser peripheral iridotomy and YAG capsulotomy.
    • If AI can mimic various patients, students stand to gain valuable experience to supplement their clinical rotations.
  • In marketing: Eyecare professionals who are also business owners might take advantage of AI for marketing their practice. AI in marketing is already being used by corporations and small businesses alike.
    • From designing a logo to helping create an engaging video, AI-driven tools are making “putting yourself out there” a lot easier.

Challenges of AI in eyecare

While AI has some definite benefits, it also presents several challenges, and some of these may affect you regardless of whether or not you are prepared to implement AI in your practice.

1. AI as the new “Dr. Google”

Patients looking up their symptoms online is not a new phenomenon. However, for the first time, patients can potentially query a single, “trusted” source and get a breakdown of potential conditions- as opposed to having to sift through Google results. While having a breakdown from one place is hypothetically easier, it also robs the public of the opportunity to check their sources.
When using a search engine, one can identify whether the information that they are reading is coming from a trusted source, like an academic article, or simply a popular blog. Unlike with Google, patients aren’t actually sure where the data is coming from. Of course, this goes back to the challenge of procuring quality data—the better the data powering the AI algorithm, the more likely it is to yield reliable results.

When tested, ChatGPT can now pass the US Medical Licensing Exam. However, the current version only has data up to 2021, so some of the information can be outdated.7

2. Advertising within AI-utilizing platforms

It won’t take long before companies realize they can market their products within AI-based platforms. Although a system for paid AI marketing doesn’t currently exist, there will likely be one developed in the future as AI adoption continues to grow.

3. Information safety

The use of AI must be done in a HIPPA-compliant manner to help protect patient information. Since AI collects information to “learn,” privacy safeguards need to be developed to help safeguard patient information. Over time, we will develop ways to safeguard data better, but in the meantime, it is important to be vigilant.

4. AI fakes

Unfortunately, AI fakes are a threat to all. AI can be used to mimic human voices and even human likenesses. As a result, bad actors can use the technology to manipulate or even scam people.
The rise of deep fakes, where AI is used to create a video that looks and sounds like a real person to spread false information or to falsely portray them is also an issue. This can translate to scammers pretending that they are calling from a doctor’s office to collect a “payment” that the patient never owed.

5. A new challenge for optometry schools

AI has made it easier than ever to create emails, articles, and even reports. AI can provide an amazing framework and guidelines for students who may feel lost at the beginning of a written assignment. However, students may be tempted to simply copy and paste AI-generated text in order to save themselves time.
In January of 2023, a program was introduced to help detect AI-generated content, which can help teachers distinguish AI-written work from that of their students. However, this app has not been tested.8

Final thoughts

Though AI is not new, it has recently returned to the spotlight with the potential to help eyecare professionals. The popularity of AI has motivated tech companies to continue their advances.
Research and development are currently being conducted in order to make AI better and more accessible to doctors and patients alike. With substantial financial backing and the interest of major tech companies, it’s likely that AI within eyecare will develop quickly and that we will see many advancements in the coming years.
  1. Joshi N. 7 Types Of Artificial Intelligence. Forbes. Published June 19, 2019. https://www.forbes.com/sites/cognitiveworld/2019/06/19/7-types-of-artificial-intelligence/.
  2. Gillis A. 4 Main Types of Artificial Intelligence: Explained. TechTarget. Published August 7, 2023. https://www.techtarget.com/searchenterpriseai/tip/4-main-types-of-AI-explained.
  3. SAS Institute. Artificial Intelligence (AI): What it is and why it matters.” n.d. SAS Institute. Accessed August 31, 2023. https://www.sas.com/en_us/insights/analytics/what-is-artificial-intelligence.html.
  4. Anyoha R. The History of Artificial Intelligence. Science in the News. Published August 28, 2017. https://sitn.hms.harvard.edu/flash/2017/history-artificial-intelligence/.
  5. Massachusetts Institute of Technology. MIT team building social robot. MIT News. Published February 14, 2001. https://news.mit.edu/2001/kismet.
  6. Davidson Josh. AI-Powered Scribes Aren’t Perfect. Modern Optometry. Published February 13, 2023. https://modernod.com/significant-findings/ai-powered-scribes-arent-perfect.
  7. Lubell J. ChatGPT passed the USMLE. What does it mean for med ed?. American Medical Association. Published March 3, 2023. https://www.ama-assn.org/practice-management/digital/chatgpt-passed-usmle-what-does-it-mean-med-ed.
  8. Osborne M. Student Creates App to Detect Essays Written by AI. Smithsonian Magazine. Published January 17, 2023. https://www.smithsonianmag.com/smart-news/student-creates-app-to-detect-essays-written-by-ai-180981463/.
Irina Yakubin, OD
About Irina Yakubin, OD

Irina Yakubin, OD, is a primary care and low vision optometrist currently practicing in Los Angeles, California. She graduated from the InterAmerican University of Puerto Rico in 2020. Her areas of interest include dry eye, ocular disease, and contact lenses. In addition to seeing patients and writing, she also co-produces My Vision Show.

Irina Yakubin, OD
Eyes On Eyecare Site Sponsors
Astellas LogoOptilight by Lumenis Logo