The recent investigation led by the University of Cambridge revealed that the Artificial Intelligence (AI)specifically the model GPT-4have demonstrated Outperform non-specialist doctors in evaluating eye problems and offering relevant advice. This finding suggests that GPT-4’s reasoning skills and clinical knowledge reach the level of specialized ophthalmologists. This marked a milestone in the integration of AI in the healthcare sector.
The study involved comparing GPT-4 competencies with physicians at different stages of their professional training. It also included young general practitioners with basic knowledge of ophthalmology to experts in the field.
These professionals were presented with 87 scenarios of patients with eye problems, where they had to diagnose or suggest a treatment from among four possible options. The results were revealing: GPT-4 not only outperformed non-specialized physicians, but also obtained scores comparable to those of training and expert ophthalmologists. However, they were below the most brilliant doctors.
These findings point to a promising future for the application of advanced language models in medicine, especially in areas where access to specialists is limited. However, the researchers emphasize that the purpose of AI is not to replace healthcare professionals, but to complement and improve healthcare within the clinical workflow.
AI could be particularly useful in classifying patients with eye problems by helping to determine which cases require immediate specialist care, which can be managed by general practitioners, and which do not need urgent treatment. He Dr. Arun Thirunavukarasu, lead author of the study, suggests that the use of AI in patient classification could follow already established algorithms to facilitate clinical decision making. He also said it could be used even in the absence of immediate advice from an ophthalmologist.
Furthermore, Dr Thirunavukarasu said: “We could implement AI to classify patients with eye problems and decide which cases are emergencies.”
The study revealed that models like GPT-4 could provide valuable advice to general practitioners faced with the growing demand for ophthalmological services and the long waiting times for specialized care.
This advance represents a great contribution to medicine, especially in contexts where resources are scarce or where clinical workflows are sought to be optimized.
Large language models, such as GPT-4 and GPT-3.5, are developed by training on vast data sets spanning hundreds of billions of words from articles, books, and various Internet content. These platforms represent only part of the spectrum of evolving AI technology, along with other large language models widely recognized and used in various applications, such as PaLM 2 (Google) and LLaMA 2 (Meta ex Facebook).
The study also used GPT-3.5, PaLM2, and LLaMA with the same set of questions as GPT-4. However GPT-4 gave more precise answers than all the others.
Although GPT-4 showed significantly superior performance even compared to other AI models in the evaluation of ocular problems, it is important to emphasize that the decision to integrate computer systems in patient care should be an individual choice, as Dr. Thirunavukarasu highlights. . The priority remains empowering patients to make informed decisions about their health to keep physicians at the center of healthcare.
Rapid advancement in the field of AI and language models promises significant transformations in medical practice. As more advanced models emerge, the possibility of AI reaching and even surpassing the level of expertise of specialists in specific areas becomes increasingly tangible. This opens up new options to improve the quality and accessibility of healthcare for all.
