The provided text comprises excerpts from Roman V. Yampolskiy's 2024 book "AI," focusing heavily on the AI control problem and its associated challenges. Dr. Yampolskiy, a tenured associate professor and director of the Cyber Security Lab at the University of Louisville, outlines the core issue as humanity's struggle to maintain safe control over increasingly superior artificial intelligence, a central concern in AI Safety and Security. The text explores concepts like AI personhood and consciousness in machines, discussing the potential for AIs to gain legal rights and experience subjective states. It also details various impossibility results in AI safety, including the unpredictability, unexplainability, and unverifiability of advanced AI systems, suggesting inherent limits to human oversight and understanding. The author further categorises potential pathways to dangerous AI, stemming from intentional malicious design, engineering mistakes, or independent emergent properties of recursively self-improving systems.
Oct 18, 2025 · 16 min
Sep 25, 2025 · 17 min