# Curiosity Killed or Incapacitated the Cat and the Asymptotically Optimal Agent

arXiv: https://arxiv.org/abs/2006.03357 · HTML: https://mlsanity.com/paper/2006.03357/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 5 Jun 2020.

## Built on by (most cited first, 6 of 6)

- [Pessimism About Unknown Unknowns Inspires Conservatism](https://mlsanity.com/paper/2006.08753/) (2020)
- [Avoiding Catastrophe in Online Learning by Asking for Help](https://mlsanity.com/paper/2402.08062/) (2024)
- [Safe Learning Under Irreversible Dynamics via Asking for Help](https://mlsanity.com/paper/2502.14043/) (2025)
- [A Model-Free Universal AI](https://mlsanity.com/paper/2602.23242/) (2026)
- [Golden Handcuffs make safer AI agents](https://mlsanity.com/paper/2604.13609/) (2026)
- [Learning When Not to Learn: Risk-Sensitive Abstention in Bandits with Unbounded Rewards](https://mlsanity.com/paper/2510.14884/) (2025)

## Abstract (arXiv)

Reinforcement learners are agents that learn to pick actions that lead to high reward. Ideally, the value of a reinforcement learner's policy approaches optimality--where the optimal informed policy is the one which maximizes reward. Unfortunately, we show that if an agent is guaranteed to be "asymptotically optimal" in any (stochastically computable) environment, then subject to an assumption about the true environment, this agent will be either "destroyed" or "incapacitated" with probability 1. Much work in reinforcement learning uses an ergodicity assumption to avoid this problem. Often, doing theoretical research under simplifying assumptions prepares us to provide practical solutions even in the absence of those assumptions, but the ergodicity assumption in reinforcement learning may have led us entirely astray in preparing safe and effective exploration strategies for agents in dangerous environments. Rather than assuming away the problem, we present an agent, Mentee, with the modest guarantee of approaching the performance of a mentor, doing safe exploration instead of reckless exploration. Critically, Mentee's exploration probability depends on the expected information gain from exploring. In a simple non-ergodic environment with a weak mentor, we find Mentee outperforms existing asymptotically optimal agents and its mentor.