top of page

Generative AI - Exploring ChatGPT-3.5

ChatGPT 3.5: Unveiling its Performance and Attributes

ChatGPT 3.5 stands as an exemplary conversational AI system harnessed from the GPT-3.5 model, a creation of OpenAI. This innovation aims to mirror human-like dialogues by responding coherently and contextually to user inquiries. Anchored in natural language processing (NLP), a realm of artificial intelligence dedicated to facilitating machines in comprehending and generating human language, ChatGPT 3.5 emerges as a milestone in AI-enabled interactions.

An Insight into ChatGPT 3.5's Aptitude via an Ophthalmological Lens

The study titled "Evaluating ChatGPT‑3.5's Performance in Responding to Questions from the Brazilian Council of Ophthalmology Board Examination" meticulously scrutinized ChatGPT 3.5's efficacy in tackling questions from the Brazilian Council of Ophthalmology Board Examination. This endeavor aimed to ascertain ChatGPT 3.5's competence in addressing intricate and specialized queries related to the field of ophthalmology.

At the core of the ChatGPT system lies a robust large language model (LLM), meticulously trained to decipher data patterns and craft coherent retorts. The GPT-3.5 model underpinning ChatGPT boasts a staggering 175 billion parameters and has undergone training on an extensive array of textual sources, encompassing medical literature. This model receives fine-tuning through a combination of human-guided training and reinforcement learning, culminating in an optimized performance for conversational undertakings.

Within the scope of this study, ChatGPT 3.5 underwent rigorous testing against questions from the 2022 Brazilian Council of Ophthalmology Board Examination. These questions spanned theoretical dimensions of ophthalmology, encompassing fundamental sciences as well as clinical and surgical facets. This empirical exploration sought to gauge the precision of ChatGPT 3.5's responses and evaluate its prowess across diverse question categories.

Findings and Implications of the Study

The study's outcomes unveiled ChatGPT 3.5's varied performance when confronted with the ophthalmology board examination queries. Notably, it exhibited a 41.46% accuracy in providing correct responses, while 53.66% of its answers were inaccurately formulated, and 4.88% remained inconclusive. When delving into mathematical concepts, ChatGPT 3.5's accuracy dwindled to 23.8%. In theoretical examinations I and II, it managed to achieve correct responses for 43.18% and 40.83% of questions, respectively.

Inferences from the study emphasized the suboptimal performance of ChatGPT 3.5 in the context of the specialized Brazilian ophthalmology board examination. Challenges encountered in its performance were attributed to factors such as insufficient clinical training data and nuances in question formulation. Consequently, a note of caution was sounded, advocating prudence when considering the deployment of AI-driven chatbots like ChatGPT in intricate domains like ophthalmology.

Study Link:


If you are looking for any kind of help in machine learning, please contact us.

bottom of page