Computer Says Maybe

Computer Says Kill: The AI Safety Circus W/ Heidy Khlaaf

In this episode of "Computer Says Kill," AI safety expert Heidi Kloth, chief scientist at the AI Now Institute, explains how the concept of safety in critical systems—originating from the nuclear arms race—has historically driven rigorous risk assessment and reliable military technology. She contrasts those century‑old safety frameworks with today’s AI companies, which often use the language of safety without adhering to the same quantitative standards, creating conflicts of interest and overlooking public values. Kloth highlights the danger of technocratic decision‑making that sidelines societal input and warns against the naive belief that large language models can replace specialized expertise in fields like nuclear power or defense. The discussion underscores the need for genuine, interdisciplinary safety processes to prevent catastrophic failures as AI becomes more integrated into high‑stakes domains.