ChatGPT-4.0 developed by OpenAI has reached an important milestone in the interface between artificial intelligence and medicine. In a recent proof-of-concept study, ChatGPT-4.0 was able to demonstrate its technology by passing a clinical neurology test by answering an astonishing 85% of questions correctly. This surprising result was achieved in an experiment conducted by researchers at Heidelberg University Hospital and the German Cancer Research Center.
The Clinical Neurology test was developed to assess the level of expertise an AI has in a specific specialty, using information obtained from the American Board of Psychiatry and Neurology and the European Council of Neurology. The performance of ChatGPT-4.0 was significantly better than its predecessor, ChatGPT-3.5, which received a score of 66.8%. It even surpassed the average human score of 73.8%. These achievements demonstrate that the model possesses advanced skills in areas related to behavioral, cognitive, and psychological components of neurology.
Although ChatGPT-4.0 was successful, it was found that there is still room for improvement in performance. Compared to activities that require only low-order thinking, the results of both variants of the model appear to be less effective when it comes to tasks that require higher-order thinking. This finding is consistent with a broader recognition in the field of artificial intelligence that acknowledges the possibility that the cognitive abilities of the models in question will improve in the future.
Large-scale language models (LLMs) such as ChatGPT-4.0 can potentially be incorporated into clinical neurology, and the results of this study serve as a guide for future integration. Researchers believe that these models have the potential to make significant contributions to the field of medicine as long as they undergo further modifications and special fine-tuning efforts. Moreover, while this study proposes applications in decision-making and documentation support systems, it currently warns about the limitations of these systems in terms of their ability to perform high-order cognitive tasks.
The results of this study are more likely to serve as a proof of concept for LLM capabilities rather than being immediately applicable to clinical settings. The way forward consists of creating and fine-tuning these models to exploit their full potential for practical use in clinical neurology. An innovative combination of football and digital technologies will be showcased during the FIFA Club World Cup 2023TM. Through a special series of NFTs that will be available on FIFA+ Collect in collaboration with Modex. The series will enhance fan interaction and memorabilia experiences.
Image source: Shutterstock