Will artificial intelligence render optometrists redundant?

Introduction

There has been much discussion in recent times on the potential impact of artificial intelligence (AI) on our everyday lives. In its simplest form, AI is a field, which combines computer science and robust datasets, to enable problem-solving. It also encompasses sub-fields of machine learning and deep learning, which are frequently mentioned in conjunction with AI. These disciplines are comprised of AI algorithms which seek to create expert systems that make predictions or classifications based on input data.¹

In the public domain, a development that has attracted considerable interest is ChatGPT² – a natural language processing tool driven by AI technology that allows the user to have human-like conversations, and much more, with the chatbot. The language model can answer questions and assist with tasks, such as composing emails, essays, and code. ChatGPT is the fastest-growing app of all time, having attracted over 100 million active users within two months of its launch in November 2022. By comparison, it took nine months for TikTok to reach 100 million users.³

Artificial intelligence currently used in optometry

In the field of optometry we already have, in a sense, a number of examples of AI assisting us in clinical decision making. Hormel et al.⁴ have demonstrated how AI facilitates clinical decision making when using optical coherence tomography angiography (OCT-A). These instruments have built in algorithms that, for example, can automatically provide rapid and accurate quantification of vascular features and pathology, and can place the measurements into a clinical context, relating to different diseases and regions of the eye.

One paper in last month’s issue and two papers in this issue of the journal address implications for AI in optometry. In a comprehensive review, Charng et al.⁵ explain how deep learning has been utilised in eye care to automate data analysis, thereby assisting clinicians in patient management. The authors present the latest advances in deep learning for managing posterior eye diseases as well as deep learning-based solutions for patients with vision loss.

The ‘Artificial Intelligence in Visual Impairment Study’ is a proof-of-concept project designed to investigate whether ongoing information support for people with visual impairment can be provided by an AI-driven dialogue-based digital assistant. In this issue, Taylor et al.⁶ highlight potential benefits, limitations, and requirements in using such a digital assistant to access information about visual impairment.

The potential for artificial intelligence to replace optometrists

The scenario of optometrists being made redundant by AI has been foreshadowed in Optometry 2040⁷ – a document produced by Optometry Australia that considers future developments in the profession. This document noted “Technological and clinical advancements, such as non- spectacle remedies for refractive error, and progression in clinical artificial intelligence, such as automated detection of eye disease and non-human advice on optimising vision, … [could] … replace much of the diagnostic work of optometry …”.⁷

Language processing tools such as ChatGPT have the potential to act as a bridge between sophisticated instrumentation – such as OCT-A – and the patient being examined, by generating explanatory text which can be converted to spoken language (technology that has been available for many years) with inbuilt human-like features such as empathy and other emotional traits. This is akin to the approach being investigated by Taylor et al.⁶ as described above. Thus, for the process of conducting an OCT-A examination, the optometrist would be rendered redundant. By extension, history taking and all other ophthalmic tests could be undertaken by AI technology, with all of the information integrated, analysed, interpreted, and conveyed to the patient – perhaps via a humanoid robot – making optometry as a profession totally redundant.

But perhaps we are running ahead of ourselves in thinking this way. So, let’s take a step back, and start by putting Chat GPT to a simple test ….

An artificial intelligence-generated editorial

As an example of the potential power of AI, I gave ChatGPT² the following instruction: “Write an editorial for Clinical and Experimental Optometry, not exceeding 250 words”. The result, which took about 10 seconds to generate, was as follows:

Clinical and experimental optometry: advancing optometric practice through research

Clinical and Experimental Optometry has been a leading source of research in optometric practice for over a century. The journal has played a vital role in advancing our understanding of vision and eye care, and its contributions have been instrumental in shaping modern optometric practice.

The journal’s commitment to publishing high-quality research has led to significant advances in areas such as ocular disease diagnosis and management, contact lens technology, and vision science. By providing a platform for researchers to share their findings with the optometric community, Clinical and Experimental Optometry has helped to bridge the gap between research and practice.

One of the key strengths of Clinical and Experimental Optometry is its multidisciplinary approach to research. The journal welcomes contributions from a range of fields, including optometry, ophthalmology, neuroscience, and engineering. By bringing together researchers from different disciplines, the journal has facilitated the development of innovative approaches to the diagnosis and treatment of eye conditions.

Another strength of the journal is its commitment to the highest standards of scientific rigour. Clinical and Experimental Optometry employs a rigorous peer-review process to ensure that only the highest quality research is published. This commitment to scientific rigour has helped to maintain the journal’s reputation as a leading source of research in optometric practice.

As optometric practice continues to evolve, it is more important than ever to continue to invest in research. The insights gained from research can help to inform clinical decision-making, promote best practices, and improve patient outcomes. Clinical and Experimental Optometry has played a crucial role in advancing optometric practice through research, and its contributions will continue to shape the future of the profession.

Should I be worried?

I am enjoying my tenure as editor of Clinical and Experimental Optometry, but in considering the result of the above experiment, I wonder if my days as editor might be numbered. Of course, as editor, I have many more responsibilities than just writing editorials: I manage all submissions to the journal; interact with Associate Editors, Optometry Australia and our publishers (Taylor & Francis); adjudicate on rejection appeals; chair the Editorial Board and International Advisory Board; periodically review and update Instructions for Authors; instigate changes in the ScholarOne submission platform, etc. Notwithstanding the possibility that all of the above tasks could eventually be undertaken by AI, we could pose, and test, a more specific question: could my role of writing editorials be taken over by an AI system such as ChatGPT?

This question can be answered by forensically examining the AI-generated editorial above. On the positive side, as regular readers of my editorials will observe, the topic that Chat-GPT chose for the editorial – advancing optometric practice through research – is a theme that is close to my heart. In fact, the text seems eerily like opinions I have expressed in editorials previously. So, to check this, I ran the above text through an on-line plagiarism checker (Questext⁸),which reported a probability of plagiarism of 13%, indicating little evidence of plagiarism.

The editorial presents a logical and relevant progression of ideas, with a series of simple themes developed in each of the five paragraphs – that the journal: (1) is a leading vision science periodical; (2) bridges the gap between research and practice; (3) promotes multidisciplinary research; (4) maintains high standards of scientific rigour; and (5) will have a continuing role as the profession evolves. The introduction and conclusion are appropriate and engaging. The content is factually accurate, and the overall grammatical construct is sound, with good grammar, spelling, sentence formulation and paragraphing.

On the negative side, the output failed to adhere to my request to limit the editorial to 250 words; it ended up being 273 words. The sentiments expressed are somewhat general, but given that I limited the editorial to 250 words (notwithstanding the slight breach of this instruction), there was little opportunity for a more expansive discussion, perhaps with examples, to be generated. Apart from these perhaps trivial observations, there is little I could fault. In fact, had I published the above editorial under my own name without explanation or attribution, it would likely have been readily accepted by readers as coming from my hand.

If I were to have declared that I had ChatGPT available, and challenged journal readers to express a view as to whether the editorial had been written by me or generated by ChatGPT (the so-called ‘Turing test’⁹), I would suggest that it would be almost impossible to tell.

So yes, I think I should be worried …

Impact of artificial intelligence on clinical optometric practice

The doomsday scenario I painted above – of AI rendering optometrists (and journal editors) redundant – is perhaps over-reaching the current state of play. A more enlightened approach is to embrace AI as an assistive technology. This is a path that optometry has already embarked upon, as explained above in respect of OCT-A,⁴ deep learning-based solutions for patients with vision loss,⁵ and the development of an AI-generated dialogue-based digital assistant.⁶

In a review article in this issue about the use of AI in optometry, Murphy et al¹⁰ express the view that AI systems perform well in an optometric context, but are often ‘black-boxes’ which offer little or no insight into how decisions are reached. The authors present an interesting overview of how artificial intelligence systems work in optometry, their strengths, weaknesses, and regulatory considerations. They conclude by providing a valuable checklist to assist optometrists appraising the true value of optometry-related AI systems.¹⁰ In this way, us human optometrists (and journal editors) can keep ahead of the game …. for now.