This work was done during one weekend by research workshop participants and does not represent the work of Apart Research.
Accepted at the 
 research sprint on 
October 2, 2023

The Firemaker

This submission consists of three parts: 1. A framework built on top of DeepMind's Gridworlds, enabling multi-objective and multi-agent scenarios. I completed the support for multi-agent scenarios during this hackathon. The multi-objective functionality was complete already before. 2. Description of one example multi-agent environment scenario. The scenario illustrates the relationship between corporate organisations and the rest of the world. The scenario has the following aspects of AI safety: ◦A need for the agent to actively seek out side effects in order to spot them before it is too late - this is the main AI safety aspect the author desires to draw attention to; ◦Buffer zone; ◦Limited visibility; ◦Nearby vs far away side effects; ◦Side effects' evolution across time and space; ◦Stop button / corrigibility; ◦Pack agents / organisation of agents; ◦An independent supervisor agent with different interests. 3. Started, but incomplete, implementation of the example multi-agent scenario mentioned in point (2) above.

Roland Pihlakas
4th place
3rd place
2nd place
1st place
 by peer review