Joseph Suárez

First initial last name location mit suffix edu
GitHub / Twitter

About me

I am an MIT PhD, recently graduated from Phillip Isola's lab, and the creator of Neural MMO, an open-source and computationally accessible research platform that simulates populations of agents in procedurally generated virtual worlds. The focus of my work is enabling modern agent-based learning methods to scale to increasingly complex, cognitively realistic environments.

I have run three public competitons on Neural MMO in partnership with AICrowd. The most recent two were additionally partnered with Parametrix.ai and hosted at IJCAI 2022 and NeurIPS 2022, each receiving around 500 signups and 1600 submissions. I am currently organizing a fourth competition on generalization to novel tasks in partnership with Stability.ai and CarperAI.

Outside of competitions and pure development, I run the Neural MMO community Discord, which currently has around 700 members. This includes both competition participants and academics using the platform for other research. I also run the LIFE monthly group meeting (Learning in Foundation Environments) with invited talks from researchers and developers working on new environments, platforms, and associated tooling. Message me anywhere if you want an invitation.

Previously, I graduated from Stanford for CS+AI in 2019. I took a semester off from my degree in 2018 to intern at OpenAI for 6 months, where I formalized and released the first public version of Neural MMO. Before that, I worked with Justin Johnson in Fei-Fei Li's group on dynamic visual question answering architectures. Before that, I started on language modeling research in a program run by Andrew Ng's lab.


News



→ We are running a 4th Neural MMO competiton on generalization to new tasks in large-scale multiagent environments. Join the Discord for updates!

Thesis
PhD Thesis
Download Joseph Suarez's PhD Thesis.
[Download]
Competitions
The NeurIPS 2023 Neural MMO Challenge
In partnership with Parametrix.ai, Stability AI, CarperAI, and AICrowd.
Multitask reinforcement learning and curriculum generation
[Competition]
The NeurIPS 2022 Neural MMO Challenge
In partnership with Parametrix.ai and AICrowd. Sponsored by AWS.
Specialize and bargain in brave new worlds
[Competition]
The IJCAI 2022 Neural MMO Challenge
In partnership with Parametrix.ai and AICrowd
Survive and thrive together in large open worlds
[Competition]
The Neural MMO Challenge (2021)
In partnership with AICrowd
Robustness and teamwork in a massively multiagent environment
[Competition]
Neural MMO Core
Neural MMO 2.0: A Massively Multi-task Addition to Massively Multi-agent Learning
Joseph Suárez, Phillip Isola, Kyoung Whan Choe, David Bloomin, Hao Xiang Li, Nikhil Pinnaparaju, Nishaanth Kanna, Daniel Scott, Ryan Sullivan, Rose S. Shuman, Lucas de Alcântara, Herbie Bradley, Louis Castricato, Kirsty You, Yuhao Jiang, Qimai Li, Jiaxin Chen, Xiaolong Zhu
NeurIPS Datasets and Benchmarks 2023.
[Pre-print]
The Neural MMO Platform for Massively Multiagent Research
Joseph Suarez, Yilun Du, Clare Zhu, Igor Mordatch, Phillip Isola
NeurIPS 2021, Track on Datasets and Benchmarks.
[Pre-print][Website][Code][Competition]
Neural MMO v1.3: A Massively Multiagent Game Environment for Training and Evaluating Neural Networks
Joseph Suarez, Yilun Du, Igor Mordatch, Phillip Isola
AAMAS Extended Abstract 2020.
[Extended abstract][arXiv paper][Code]
Neural MMO: A Massively Multiagent Game Environment for Training and Evaluating Intelligent Agents
Joseph Suarez, Yilun Du, Phillip Isola, Igor Mordatch
arXiv 2019.
[Paper][Blog][Code]
Other Work
GAN You Do the GAN GAN?
Joseph Suarez
arXiv 2019.
[Paper][Code]
DDRprog: A CLEVR Differentiable Dynamic Reasoning Programmer
Joseph Suarez, Justin Johnson, Fei-Fei Li
arXiv 2018.
[Paper][Code]
Character-Level Language Modeling with Recurrent Highway Hypernetworks
Joseph Suarez
NIPS (now NeurIPS) 2017.
[Paper][Code]
Efficient Approaches to Batch Parallelization for Dynamic Neural Architectures
Joseph Suarez, Clare Zhu
arXiv 2017.
[Paper][Code]