Hey there 👋
I am Vyom, from the community team at AIcrowd.
I wanted to share with everyone that AIcrowd in collaboration with MIT is organising Neural MMO Challenge.
In this challenge, you will design and build agents that can survive and thrive in a massively multiagent environment full of potential adversaries. Explore Neural MMO's procedurally generated maps, scavenge for resources, and acquire equipment to protect yourself while preventing other participants from doing the same.
You can click here to participate in the challenge.
Neural MMO represents a new frontier for agent-based artificial intelligence. Taken together, its large agent populations, long time horizons, and multitask objective structure create a challenge unique among currently available research environments. Solving it will require significant adaptation of current methods and, quite possibly, the development of entirely new algorithms.
Despite these complexities, research on Neural MMO is not much more complicated than research on simple single-agent environments. Quality infrastructure through deep integration with RLlib provides sane defaults for what would otherwise be major inconveniences -- structured observation and action spaces, representing long trajectories in memory, and bootstrapping across segments, to name a few.
I am hopeful that you would give this email a thought and consider participating in this challenge.
If you have any questions around the same, I’d love to answer them.
Best,
Vyom
Community Team
AIcrowd