In a new blog-post on the lesswrong forum, AIXI labs fellow Mikhail Mironov explores the failure of the coding theorem for randomized stopping machines.
This post explains a technical separation result in algorithmic information theory, which was derived during Mikhail Mironov’s Summer 2026 PIBBSS fellowship.
The result contributes to AIXI Labs‘ research program on how Solomonoff induction generalizes from past observations in the face of novel events.
It also builds on Aram Ebtekar and Michael K. Cohen’s ‘Golden Handcuffs (GH)’; a method of making a universal agent safer by incentivising it to delegate control to a mentor in special cases which prevents the agent from exploring novel high-reward schemes.
Leave a comment