Michael Hu: studying the connections between machine learning and communications

Wednesday, Sep 29, 2021

Michael Hu, 21, Class of 2021

Background:

Hu earned a bachelor’s degree in computer science and two undergraduate certificates: one in robotics and intelligence systems and one from the Center for Statistics and Machine Learning (CSML).

Undergraduate Work:

Both his junior and senior independent projects involved machine learning, Hu said. In his junior project, which fulfilled the requirement of the CSML certificate, Hu programmed reinforcement agents to understand natural language constraints, such as “Don’t step in puddles.” He also developed HazardWorld, a new open-source reinforcement learning environment, to test agent performance. This work resulted in a 2020 research paper, Safe Reinforcement Learning with Natural Language Constraints, which was submitted to the 2021 International Conference on Learning Representations. Hu was listed as second author in the paper.

His senior project, which dealt with machine learning and communications, had two components. The first part was probing the framework where a machine learning agent “teaches” another machine learning agent via a piece of text. The second part was using machine learning algorithms to understand human intelligence, specifically how humans process word analogies.

In both his independent work and his development as a software engineer, the CSML certificate has been particularly helpful, said Hu. It not only enabled him to connect to a community of like-minded scholars, but it also gave him opportunities to explore machine learning more deeply.

“When it comes to machine learning, there are profound connections to life,” said Hu. “If you look at my junior work and senior thesis, a lot of it has been driven by the symbiosis between humans and designed intelligence. I wanted to understand how we talk to each other and learn from each other and transfer that process to machines. I'm interested in machine learning because I think there's a connection to human cognition, and I think that connection is something that's a lot of fun to explore.”

Hu also credited CSML’s summer internship program, which funded his work in 2020 that dealt with reinforcement learning with constraints.

After he graduated, Hu became a software engineer intern at Roblox this past summer. In August, he assumed the position of software engineer at Yobi, a new tech start-up. One of the founders is Tom Griffiths, the Henry R. Luce Professor of Information Technology, Consciousness, and Culture of Psychology and Computer Science. Yobi’s mission is to understand and predict human behavior via datasets and provide these tools to smaller companies that don’t have the resources of Facebook or Google.

Hu said he plans on working in industry for several years and then eventually getting his doctoral degree so he can dive deeper into machine learning, deep learning and artificial intelligence.

Extracurricular Activities:

Hu was a member of the Princeton Tower Club, an eating club, and Sympoh Urban Dance Crew.

For Fun:

Hu enjoys doing yoga, running and hanging out with his friends.