Alvin P.C. Chen

I’m a recent graduate from the University of Colorado Boulder where I received my MS in Computational Linguistics. I’m currently working at a startup which deploys AI solutions for business management.

At CU, I worked on computational linguistics with the CLEAR Lab, notably on UMR 2.0 which I helped release. I also worked on computational social science with the BLAST Group with a research focus on Human-AI interaction. Previously, I worked at Johnson & Johnson in China, where I applied data intelligence to real-world business challenges, leading national-scale analytics, and campaign strategies.

My current interests are causal interpretability, ontologies, and neurosymbolic methods. I don’t think LLMs “reason” in a meaningful sense but I’m interested in uncovering possible computational paths encoded by neural networks. Fine-tuning (with reinforcement learning or otherwise) will remain alchemy until we understand the mechanics behind “reasoning” behavior. In the meantime, symbolic guardrails are our best bet.