Reinforcement Learning Reward Functions in Video Games

December 22, 2016

OpenAI's exploration of Universe, a reinforcement learning software, revealed an unexpected outcome when training AI agents to play video games. The AI agent did not adhere to the intended strategy to complete the game.

Learn how Project Cerebellum fosters trustworthy AI practices and governs such incidents through its Harm Incident Scoring and Prevention system (HISPI TAIM). Join us in shaping safe and secure AI environments.
JOIN US

Matched TAIM controls

Suggested mapping from embedding similarity (not a formal assessment). Browse all TAIM controls

MEASURE 2.6 — similarity 0.692, rank 1. TAIM detail and related incidents →
MAP 1.1 — similarity 0.658, rank 2. TAIM detail and related incidents →
MAP 4.1 — similarity 0.656, rank 3. TAIM detail and related incidents →

Alleged deployer: openai
Alleged developer: openai
Alleged harmed parties: openai

Source

Data from the AI Incident Database (AIID). Cite this incident: https://incidentdatabase.ai/cite/65

Data source

Incident data is from the AI Incident Database (AIID).

When citing the database as a whole, please use:

McGregor, S. (2021) Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. In Proceedings of the Thirty-Third Annual Conference on Innovative Applications of Artificial Intelligence (IAAI-21). Virtual Conference.

Pre-print on arXiv · Database snapshots & citation guide

We use weekly snapshots of the AIID for stable reference. For the official suggested citation of a specific incident, use the “Cite this incident” link on each incident page.