X.Y. Han, PhD and Kirill Skobelev: “What will get us to Surgical AGI?”

SDSC’s Data Science Roundtable with X.Y. Han, PhD – Assistant Professor of Operations Management at the University of Chicago Booth School of Business – and Kirill Skobelev – Pre-Doctoral Researcher at the University of Chicago Booth School of Business’ Center for Applied AI – explored a key question from their recent paper on medical artificial intelligence (AI): how close are we to surgical artificial general intelligence (AGI)?¹

The discussion challenged the growing narrative that large AI models trained on internet-scale data are close to replacing human expertise in surgery. The speakers demonstrated that modern vision-language models perform well on medical question-and-answer benchmarks, however, these successes do not translate to real-world surgical understanding.

Using SDSC’s newly developed endoscopic pituitary surgery dataset alongside other public surgical datasets, the speakers evaluated whether state-of-the-art systems could reliably identify surgical tools in operative video frames, a foundational task for surgical intelligence. Their findings showed that even the largest general-purpose AI models performed poorly and often struggled to outperform simple statistical baselines despite containing billions or trillions of parameters.

 In contrast, smaller specialized computer vision models trained directly on surgical data significantly outperformed generalist systems while using far fewer computational resources. The results suggest that progress toward surgical AGI depends less on scaling massive foundation models and more on access to high-quality, domain-specific surgical data and expertise.

 The speakers emphasized the importance of “tacit knowledge” in surgery – practical understanding gained through experience rather than textbooks alone – and argued that organizations like SDSC play a critical role in advancing safe, clinically meaningful surgical AI research.

^{1. Skobelev K, Fithian E, Baranovski Y, Cook J, Angara S, Otto S, et al. A comparative study in surgical AI: Datasets, Foundation models, and barriers to Med-Agi. 2026; doi:10.2139/ssrn.6476260} 

‍

‍