OpenAI Researcher Calls Incentive Landscape Alignment's Outermost Loop
OpenAI researcher Jason Wolfe posts that lab incentives form the deepest alignment layer.
Jason Wolfe, a researcher at OpenAI working on AI alignment and the Model Spec, posted that the outer-outer-outer loop of AI alignment is the incentive landscape the labs are embedded in. The remark positions alignment work as reaching beyond technical fixes into the economic and competitive pressures around AI labs. It appeared in posts tagged AI Safety, matching Wolfe's documented focus on alignment topics.
Combined views
5.8K
2 posts, first seen 2d ago
OpenAI Researcher Calls Incentive Landscape Alignment's Outermost Loop
OpenAI researcher Jason Wolfe posts that lab incentives form the deepest alignment layer.
Jason Wolfe, a researcher at OpenAI working on AI alignment and the Model Spec, posted that the outer-outer-outer loop of AI alignment is the incentive landscape the labs are embedded in. The remark positions alignment work as reaching beyond technical fixes into the economic and competitive pressures around AI labs. It appeared in posts tagged AI Safety, matching Wolfe's documented focus on alignment topics.
Reactions
2 postsJW 
Jason Wolfe@w01feThe outer-outer-outer loop of AI alignment is the incentive landscape the labs are embedded in.
- likes: 77
- replies: 5
- bookmarks: 10
- reposts: 9
JL Jeffrey Ladish@JeffLadish@w01fe What’s the outer outer loop? Employee incentives? Internal company incentives? The selection process on the training runs, eg which ones work?
- likes: 2
- replies: 2
- bookmarks: 0
- reposts: 0