We present a suite of reinforcement learning environments illustrating various safety properties of intelligent agents. These problems include safe interruptibility, avoiding side effects, absent supervisor, reward gaming, safe exploration, as well as robustness to self-modification, distributional shift, and adversaries. To measure compliance with the intended safe behavior, we equip each environment with a performance function that is hidden from the agent. This allows us to categorize AI safety problems into robustness and specification problems, depending on whether the performance function corresponds to the observed reward function. We evaluate A2C and Rainbow, two recent deep reinforcement learning agents, on our environments and show that they are not able to solve them satisfactorily.
@article{arxiv.1711.09883,
title = {AI Safety Gridworlds},
author = {Jan Leike and Miljan Martic and Victoria Krakovna and Pedro A. Ortega and Tom Everitt and Andrew Lefrancq and Laurent Orseau and Shane Legg},
journal= {arXiv preprint arXiv:1711.09883},
year = {2017}
}