Low AI-doom expectations coexist with support for safety work

A personal P(doom) estimate is described as “very, very low,” while extensive AI safety work—including cybersecurity-related work—is supported. The rationale is that sustained attention and incentives across the AI ecosystem could help identify and address catastrophic risks.

The same view warns that safety language may be co-opted by parties with different goals, creating a narrow line between funding safety and alignment research and generating political pressure to stop new data centers from being built. The AI policy debate is expected to remain messy for the coming years and become a dominant subject of the 2028 presidential election, with jobs, GDP and safety at issue. Conflicting industry positions are illustrated by support for both an open-weights initiative and a pacing letter.