This work was done during one weekend by research workshop participants and does not represent the work of Apart Research.
ApartSprints
Governance
Accepted at the 
 research sprint on 
Accepted at the 
Governance
 research sprint on 
January 8, 2024

Boxing AIs - The power of checklists

Guidelines for managing risks during the development of medium to high risk models (from client facing AIs to AI for super alignment), aimed for ASL-4 and High-risks systems. In this post, we open the discussion about concrete AI-risks mitigations strategies during training and predeployment by proposing a non-exhaustive list of precautions that AGI developers should respect. This post is intended for AI safety researchers and people working in AGI labs.

By 
Charbel-Raphael SEGERIE, Quentin FEUILLADE-MONTIXI
🏆 
4th place
3rd place
2nd place
1st place
 by peer review