Strongly-Typed Agents are Guaranteed to Interact Safely
share
As artificial agents proliferate, it is becoming increasingly important to ensure that their interactions with one another are well-behaved. In this paper, we formalize a common-sense notion of when algorithms are well-behaved: an algorithm is safe if it does no harm. Motivated by recent progress in deep learning, we focus on the specific case where agents update their actions according to gradient descent. The first result is that gradient descent converges to a Nash equilibrium in safe games. The paper provides sufficient conditions that guarantee safe interactions. The main contribution is to define strongly-typed agents and show they are guaranteed to interact safely. A series of examples show that strong-typing generalizes certain key features of convexity and is closely related to blind source separation. The analysis introduce a new perspective on classical multilinear games based on tensor decomposition.
READ FULL TEXT
