Robust AI Security and Alignment: A Sisyphean Endeavor?

This article argues that there are fundamental mathematical limits to how secure and well-aligned (following intended behavior) AI systems can be, and that understanding these limits is important before deploying AI widely. The research also shows that AI systems have basic reasoning limitations that stem from these same information-theoretic constraints.