Expression of Interest, AI Misalignment Bounty
Opportunity type
Funding
Cause areas
AI safety & policy
Routes to impact
Skill-building & building career capital
Learning about important cause areas
Testing your fit for a certain career path
Direct high impact on an important cause
Location
Remote, Global
Description
- Bounty of up to $1,000 per unique submission of AI agent misalignment cases.
- Submit a prompt for an AI agent, an environment for the agent to run in (Docker container), and a description of observed misaligned behaviour.
- The organisation will reproduce experiments to confirm the behaviour before paying bounties.
- Authors of best submissions may be invited to work on a contract basis.
- Pre-register to be notified when the program is live.
Join 60k subscribers and sign up for the EA Newsletter, a monthly email with the latest ideas and opportunities