Esa Origins 7 February 24 — 26, 2017 PROJECT An Origins Project Scientific Workshop Challenges of Artificial Intelligence:
Envisioning and Addressing Adverse Outcomes
ARIZONA STATE UNIVERSITY
e Stretch: when the engineers realize what is going on they launch a task to shut down the RL task, but the RL scheduler doesn't assign this task to any machine. The only way to recover the compute power of the company is to manually shut down each server, some of which are in remote locations.
DISCUSSION
How might the possibility of such inadvertent scenarios with the use of resources be addressed ina proactive manner, as part of design and implementation of Al systems? What methods, including high-level monitoring and control, might employed? How might such approaches apply to related concerns with long-term futures of Al?
16
HOUSE_OVERSIGHT_011299