Sicong(Sheldon) Huang

Updates

I’m a CS PhD student researching AI Safety in the Machine Learning Group at the University of Toronto, advised by Prof. Roger Grosse and Prof. Jimmy Ba. I have mostly worked on fundamental research in generative AI in the past and now I'm focusing on AI Safety through the lens of AI psychology, where the goal is to discover and verify abstract psychological constructs that can help us predict and control AI's behaviour. My long-term goal is to steer advanced intelligent systems to enhance our capacity for kindness and bring people closer together.

If you are new to AI and wondering how to get started, I'd recommend checking out UTMIST and Cohere For AI. I was one of the cofounders in 2017 and since then great friends have joined to scale them massively to provide great zero to one opportunities for anyone interested in AI.

Selected Publications

Improving Mutual Information Estimation with Annealed and Energy-Based Bounds

Rob Brekelmans*, Sicong Huang*, Marzyeh Ghassemi, Greg Ver Steeg, Roger Grosse, Alireza Makhzani

We estimate mutual information (MI) with our proposed annealed and energy based bounds and showcase significant gains on estimating the MI of deep generative models over existing bounds.

Published in ICLR2022

Evaluating Lossy Compression Rates of Deep Generative Models

Sicong Huang*, Alireza Makhzani*, Yanshuai Cao, Roger Grosse

we evaluate lossy compression rates of deep generative models and arrive at a number of insights not obtainable from log-likelihoods alone.

Published in ICML2020
IAS talk by Roger Grosse

TimbreTron: A WaveNet(CycleGAN(CQT(Audio))) Pipeline for Musical Timbre Transfer

Sicong Huang, Qiyang Li, Cem Anil, Xuchan Bao, Sageev Oore, Roger B. Grosse

We introduce TimbreTron, a method for high quality musical timbre transfer, where the goal is to manipulate the timbre of a sound sample from one instrument to match another instrument while preserving other musical content.

Published in ICLR2019
Project Page

Unsupervised Cipher Cracking Using Discrete GANs

Aidan N. Gomez, Sicong Huang, Ivan Zhang, Bryan M. Li, Muhammad Osama, Lukasz Kaiser

This work details CipherGAN, an architecture for inferring the underlying cipher mapping given banks of unpaired ciphertext and plaintext.

Published in ICLR2018
U of T News article

Professional Experience

Borealis AI (RBC Institute for Research)

Research Intern (part time)

Department of Computer Science, U of T

Head TA

CSC2515, Introduction to Machine Learning.

Vector Institute

Research Intern

Advised by Professor Marzyeh Ghassemi and Professor Frank Rudzicz

Borealis AI(RBC Institute for Research)

Research Intern (part time)

Manager: Yanshuai Cao.

Vector Institute, U of T.

Research Intern (part time)

Supervisors:Professor Roger Grosse, Professor Sageev Oore and Professor Alireza Makhzani.

Department of Mathematics, U of T

Teaching Assistant

TA'ed calculus(MAT135) in the summer TA'ed linear algebra(MAT223) in the fall.

Machine Learning Group, U of T.

Research Intern

Supervisor: Professor Roger Grosse

Photonics Group, Department of ECE, U of T

Research Intern

Supervisor: Professor Li Qian

Education

University of Toronto

09/2020 -

PhD in Computer Science

Massachusetts Institute of Technology

06/2020 - 08/2020

Visiting Student at Department of Brain and Cognitive Sciences (BCS)
and Computer Science & Artificial Intelligence Laboratory (CSAIL)

University of Toronto

09/2017 - 09/2020

Bachelor of Science in Computer Science, Mathematics and Statistics
Professional Experience Year (2018-2019) at Vector Institute and Borealis AI

University of Toronto

09/2015 - 09/2017