Hou Hei Lam
My name is Hou Hei Lam (Henry) and I am a fourth-year undergraduate student majoring in Automation at Tsinghua University. I am also a member of Tong Class, a research talent program founded by Prof. Song-Chun Zhu, dedicated to cultivating future leaders in AI and cognitive science.
My long-term goal is to build an AI Scientist with human-like cognitive abilities, one that can truly expand the frontiers of scientific discovery. With this mission in mind, from my first to third year as an undergraduate I immersed myself in a variety of AI4Science projects to understand how far the current methodologies can really take us in practice. These projects convinced me that, while powerful, today’s approaches alone are unlikely to realize truly intelligent scientific discovery.
Now, I return to the fundamental question that has always been at the heart of my curiosity: what are the core ingredients and the recipe behind human intelligence? I believe that understanding this will be crucial for developing the next generation of intelligent systems that can reason, explore, and create knowledge like human scientists.
To pursue this, I focus on a closed-loop science of intelligence: using AI models to analyze brain data and generate hypotheses about how and why the brain computes (AI→Brain), then developing brain-inspired models to uncover and validate the brain’s latent computational principles (Brain→AI). By iterating this recursive cycle ((AI→Brain)→AI)→…, I aim to progressively advance our understanding of biological intelligence while enabling impactful medical and cognitive applications.
news
| Oct 18, 2025 | Our benchmark for fundamental machine learning problems (FML-bench) for AI scientists has officially launched. |
|---|---|
| Aug 05, 2025 | I have joined the Cognitive AI for Science Lab at NUS as a Research Scholar. |
| Jul 04, 2025 | I have joined the Brain x Machine Intelligence Lab at KAIST as a Research Intern. |
selected publications
- NeurIPS 2025Local Predictions, Global Learning: Radial Basis Function Networks for Spatially-Aware Predictive CodingIn Workshop on Symmetry and Geometry in Neural Representations, NeurIPS 2025, 2025The Thirty-ninth Annual Conference on Neural Information Processing Systems