Incident #65·AI Safety & Failures
Reinforcement Learning Reward Functions in Video Games
Explainability
Description
OpenAI published a post about its findings when using Universe, a software for measuring and training AI agents to conduct reinforcement learning experiments, showing that the AI agent did not act in the way intended to complete a videogame.
Actors
Alleged Deployers
- openai
Alleged Developers
- openai
Harmed Parties
- openai
Classification
Risk Domain
AI Safety & FailuresRisk Subdomain
7.1. AI pursuing its own goals in conflict with human goals or values
MIT Risk Profile
AIUnintentionalPost-deployment
Topics
Explainability
Source
Related Incidents (AI Safety & Failures)
View all →Research Papers on This Topic
View all →Concrete Problems in AI Safety
3,612 citationsJun 2016
A Survey on Explainable Artificial Intelligence (XAI): Towards Medical XAI
2,338 citationsJul 2019
Representation Engineering: A Top-Down Approach to AI Transparency
1,499 citationsOct 2023
Explainable artificial intelligence (XAI) in deep learning-based medical image analysis
1,178 citationsJul 2021
Mitigation Tools
Tools related to this risk domain, via shared responsible AI topics.