Shuhao Fu has joined the Santa Fe Institute as a Program Postdoctoral Fellow, focusing on artificial intelligence research. Fu’s work aims to address current limitations in AI systems, particularly their difficulty with flexible and relational reasoning, which humans perform naturally.
Fu brings expertise in both machine learning and cognitive science. He designs models that attempt to replicate human abilities such as analogy-making and compositional understanding. During his doctoral studies, he used behavioral experiments alongside structural AI models to investigate how both humans and machines perceive and reason about relationships. “I'm more into finding out why models fail,” he says, “not just making them score better on some metrics.”
His recent projects include developing methods for AI to represent relationships between elements beyond basic object detection. This involves examining how objects are arranged, interact, or how concepts can be applied across different contexts. Fu has also contributed to health-related applications of AI: at UCLA, he collaborated on using large language models for depression screening; at Johns Hopkins University, he worked on detecting tumors in CT scans.
At the Santa Fe Institute, Fu is beginning research related to the Abstraction and Reasoning Corpus (ARC) challenge—a benchmark designed to test whether AI models can generalize abstract concepts from limited visual input/output examples using grids. He is also interested in advancing structural representations and new training approaches so that AI systems can achieve more human-like generalization.
