We’re excited to launch the fifth edition of the AFRICAI Journal Club, presented by Vincent Zossou, part of our ongoing series that brings together researchers, students, and professionals to critically engage with cutting-edge scientific papers at the intersection of AI and healthcare.
📄Singhal, K., Azizi, S., Tu, T. et al. (2023).
“Large language models encode clinical knowledge.”
Nature, 620, 172–180.
DOI: https://doi.org/10.1038/s41586-023-06291-2
Why this paper? This paper addresses one of the most critical and high-stakes applications of large language models. What makes it especially relevant is its emphasis on trustworthiness, safety, and domain alignment—key pillars for deploying AI responsibly in clinical settings. By combining benchmarking, human expert evaluation, and alignment techniques, the paper offers a rigorous and thoughtful evaluation framework for clinical LLMs. It also highlights the gap between AI models and expert clinicians, revealing both the potential and current limitations of these systems. As the use of AI in global health expands, such responsible studies are essential for guiding adoption in real-world, resource-constrained environments.
Vincent Zossou is a data scientist and data engineer specializing in medical imaging, clinical NLP, and AI for public health. He holds a master’s degree in information systems and worked at Société Nationale de Mécanisation Agricole, a state-owned enterprise in Benin, where he led projects in satellite data analysis and predictive modeling. He later earned a PhD in data science and biostatistics from Université Paris-Saclay and INSERM, focusing on liver lesion segmentation and the extraction of clinical information. He is now a research engineer in bioinformatics at Assistance Publique – Hôpitaux de Paris (AP-HP), where he develops language models for oncology reports, supporting cancer research, decision-making, and integration with FHIR interoperability standards.
Event Details
• Date: Tuesday, August 26, 2025
• Time: 13:00 GMT | 14:00 WAT | 15:00 CAT | 15:00 CEST
• Duration: 60 minutes (30-minute talk + 30-minute discussion)
• Location: Online via Zoom (link shared upon registration)
• Register now: link
Full article Reference
Singhal, K., Azizi, S., Tu, T. et al. (2023). Large language models encode clinical knowledge. Nature, 620, 172–180. https://doi.org/10.1038/s41586-023-06291-2
