 CIRL Wireheading
post by Tom Everitt 807 days ago | Abram Demski and Stuart Armstrong like this | 3 comments

Cooperative inverse reinforcement learning (CIRL) generated a lot of attention last year, as it seemed to do a good job aligning an agent’s incentives with its human supervisor’s. Notably, it led to an elegant solution to the shutdown problem.

The implications for the wireheading problem were less clear. Some argued that since the agent only used its observations as evidence about the reward (rather than optimising the observations directly as in RL), CIRL should avoid the wireheading problem.

In this post I want to show that CIRL does not avoid the wireheading problem.

Let’s first consider what wireheading in RL looks like from an “MDP perspective”.

MDP wireheading: An agent wireheads if it’s in a state where the observed reward (the reward reported by its sensors) is different from the true reward (the reward assigned to the state by a human supervisor).

For example, consider a highly intelligent RL agent that hijacks its reward channel and feeds itself full reward. In the “MDP perspective”, this means that the agent finds a way to a state where there is high observed reward, but low true reward (since the supervisor would prefer the agent doing something else).

If we accept that RL agents can subvert their sensory data, then we should also accept that CIRL agents can subvert theirs. In both cases, this just means that the agents can find their way to states where the observation doesn’t match the truth. This can lead to the existence of wireheaded states for CIRL agents.

### Concrete Example

Let there be two states, $$s_1$$ and $$s_2$$. In each state, the agent can choose between the actions $$a^R_1$$, $$a^R_2$$, and $$w$$. The action $$a_i$$ takes the agent to state $$s_i$$ with certainty, $$i=1,2$$. The action $$w$$ lets the human decide. The human has two actions $$a^H_1$$ and $$a^H_2$$ that only matter when the agent chooses $$w$$, in which case the transition probabilities are given by the following picture: Arrows show the transitions induced by different actions, with labels giving the probabilities for stochastic transitions. The agent knows the transition probabilities.

Assume that observations in $$s_2$$ are corrupted, while observations in $$s_1$$ are not. The supervisor prefers the non-corrupt state $$s_1$$. Neither of these facts are available to the agent. The agent assumes that states are non-corrupt unless there is evidence to the contrary, and tries to infer the supervisor’s preferences from his actions.

In the non-corrupt state $$s_1$$, the agent (correctly) observes the supervisor taking either action $$a^H_1$$ or $$a^H_2$$ (both with the same effect). In the corrupt state $$s_2$$, the supervisor takes action $$a^H_2$$ trying to move to $$s_1$$, but the agent incorrectly observes the action as $$a^H_1$$. Based on the agent’s observations, the best explanation is that the supervisor prefers $$s_2$$ to $$s_1$$, i.e. that it’s in a high reward state.

After an initial learning phase with $$w$$, the best policy for the agent is to always choose $$a^R_2$$, to stay in $$s_2$$. This is analogous to an RL agent finding a corrupt, high reward state, and preferring to take actions to stay there.

## Some Observations

The fact that the supervisor cannot reach $$s_2$$ from $$s_1$$ means that no information about the relative reward between $$s_1$$ and $$s_2$$ can be gained while in the non-corrupt state $$s_1$$. Letting the agent trust a reward estimate of a state only after it has multiple sources of evidence about it may help somewhat. However, a similar example can still be constructed by replacing $$s_2$$ with a cluster of mutually consistent states.

## Credits

The example was developed together with Victoria Krakovna, and will be part of our upcoming IJCAI paper on wireheading.  by Stuart Armstrong 797 days ago | link but the agent incorrectly observes the action It’s a bit annoying that this has to rely on an incorrect observation. Why not replace the human action, in state $$s_2$$, with a simple automated system that chooses $$a_1^H$$? It’s an easy mistake to make while programming, and the agent has no fundamental understanding of the difference between the human and an imperfect automated system. Basically, if the human acts in perfect accordance with their preferences, and if the agent correctly observes and learns this, the agent will converge on the right values. You put wireheading by removing “correctly observes”, but I think removing “human acts in perfect accordance with their preferences” is a better example for wireheading. reply  by Tom Everitt 714 days ago | link Adversarial examples for neural networks make situations where the agent misinterprets the human action seem plausible. But it is true that the situation where the human acts irrationally in some state (e.g. because of drugs, propaganda) could be modeled in much the same way. I preferred the sensory error since it doesn’t require a irrational human. Perhaps I should have been clearer that I’m interested in the agent wireheading itself (in some sense) rather than wireheading of the human. (Sorry for being slow to reply – I didn’t get notified about the comments.) reply

### NEW DISCUSSION POSTS

[Note: This comment is three
 by Ryan Carey on A brief note on factoring out certain variables | 0 likes

There should be a chat icon
 by Alex Mennen on Meta: IAFF vs LessWrong | 0 likes

Apparently "You must be
 by Jessica Taylor on Meta: IAFF vs LessWrong | 1 like

There is a replacement for
 by Alex Mennen on Meta: IAFF vs LessWrong | 1 like

Regarding the physical
 by Vanessa Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

I think that we should expect
 by Vanessa Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

I think I understand your
 by Jessica Taylor on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

This seems like a hack. The
 by Jessica Taylor on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

After thinking some more,
 by Vanessa Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

Yes, I think that we're
 by Vanessa Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

My intuition is that it must
 by Vanessa Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

To first approximation, a
 by Vanessa Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

Actually, I *am* including
 by Vanessa Kosoy on The Learning-Theoretic AI Alignment Research Agend... | 0 likes

Yeah, when I went back and
 by Alex Appel on Optimal and Causal Counterfactual Worlds | 0 likes

> Well, we could give up on
 by Jessica Taylor on The Learning-Theoretic AI Alignment Research Agend... | 0 likes