Neural MMO Teaching AI About Agents' 'Overall Competence'
OpenAI is a California-based nonprofit group that released "Neural MMO", a "massively multiagent" game with the goal that "the inclusion of many agents and species leads to better exploration, divergent niche formation, and greater overall competence". Players enter a tile-based map where, according to VentureBeat, "they observe the square crops of tiles centered on their respective positions & make one movement & one attack per tick". They are engaged with foraging or are in combat with others.
From time to time, servers are merged to increase the server population. The "tournament style" evaluation kicks in that "allows us to directly compare policies learned in different experiment settings" the site's blog reads.
In the natural world, competition among animals can incentivize them to spread out to avoid conflict. We observe that map coverage increases as the number of concurrent agents increases. Agents learn to explore only because the presence of other agents provides a natural incentive for doing so.
Given a sufficiently large and resource-rich environment, we found different populations of agents separated across the map to avoid competing with others as the populations increased. As entities cannot out-compete other agents of their own population (i.e. agents with whom they share weights), they tend to seek areas of the map that contain enough resources to sustain their population. Similar effects were also independently observed in concurrent multiagent research by DeepMind.
If this sounds complicated, or even simply interesting, head to the Neural MMO blog to read more detailed information and / or to get involved with OpenAI.
What do you think the goals of the research into MMO "agent" competence are in the big picture of things? Are you willing to have your behaviors observed? Leave us your thoughts in the comments!