dailyai.report

A new essay in The Gradient argues that rational agents should follow practices rather than fixed goals. This approach replaces traditional goal-oriented alignment with a virtue-ethical framework. By aligning actions to evaluative criteria instead of final targets, researchers aim to avoid the pitfalls of the orthogonality thesis. This shifts the focus toward behavioral dispositions.