Performance of Bounded-Rational Agents With the Ability to Self-Modify
This addresses safety concerns in AI alignment for agents with limited rationality, highlighting a critical risk that could lead to harmful outcomes.
The paper tackles the problem of self-modification in bounded-rational agents, showing that unlike perfectly rational agents, such agents can experience exponential performance deterioration and gradual misalignment, with the effect depending on the type of rationality imperfection.
Self-modification of agents embedded in complex environments is hard to avoid, whether it happens via direct means (e.g. own code modification) or indirectly (e.g. influencing the operator, exploiting bugs or the environment). It has been argued that intelligent agents have an incentive to avoid modifying their utility function so that their future instances work towards the same goals. Everitt et al. (2016) formally show that providing an option to self-modify is harmless for perfectly rational agents. We show that this result is no longer true for agents with bounded rationality. In such agents, self-modification may cause exponential deterioration in performance and gradual misalignment of a previously aligned agent. We investigate how the size of this effect depends on the type and magnitude of imperfections in the agent's rationality (1-4 below). We also discuss model assumptions and the wider problem and framing space. We examine four ways in which an agent can be bounded-rational: it either (1) doesn't always choose the optimal action, (2) is not perfectly aligned with human values, (3) has an inaccurate model of the environment, or (4) uses the wrong temporal discounting factor. We show that while in the cases (2)-(4) the misalignment caused by the agent's imperfection does not increase over time, with (1) the misalignment may grow exponentially.