Marcus Ma

Hello! I am a PhD Student at USC and Chief Technology Officer at echo LLM. I am advised by Shrikanth Narayanan in the SAIL Lab and supported by the Andrew and Erna Viterbi Graduate Student Fellowship. I previously received my Undergraduate and Master’s degrees in Computer Science from Georgia Tech with concentrations in Intelligence and Theory, where I was fortunate to be advised by Wei Xu (Master’s), Jennifer Gahee Kim (Undergrad), and Lelia Glass (Undergrad).

I am broadly interested in making artificial systems that are capable of modeling language capable of modeling thought. I personally reject the theory of embodied cognition and believe that given the proper environment, completely digital systems are capable of intelligence; I believe that while this future is still very far off, agentic language models are our best attempts yet. The missing piece is grounding: only through grounding can language models develop intuitions for causality and interaction.

At echo LLM, we support linguistic grounding by generating insights derived directly from human preference. We collect real-time, genuine, egocentric beliefs from our users in order to build a better map of human intent. In my research at USC, I investigate the neuroscientific angle: how are fMRI signals interpreted by the human brain in relation to the neural networks behind larguage language models? At Georgia Tech, also have previously investigated how LLMs differentiate between the stylistics and semantics of written documents. Currently, I am using rt-MRI scans to reconstruct audio from vocal tract contours, building upon my research from Georgia Tech that created an automated pipeline for improving transcription speed and accuracy. While LLMs are an impressive piece of technology, I don’t think they the best vehicle for future development of artificial cognition. I hope to understand the core properties of human thought to inspire future research.

Feel free to take a look at my publications or my various lab involvements. I have also written descriptions of my various software internships along with my resume.