A tongue-in-cheek shorthand for the estimated probability that advanced artificial intelligence will cause catastrophic or existential harm to humanity, typically expressed as a percentage or subjective Bayesian guess. Borrowed from the notation “P(x)” in probability theory, the term originated within AI safety and rationalist communities to quantify one’s personal expectation of “doom” scenarios such as loss of control, misaligned superintelligence, or irreversible societal collapse. A higher P(Doom) signals greater pessimism about safe technological trajectories, while a lower score reflects confidence in governance, alignment research, or human resilience.
The phrase began circulating in online AI-risk and Effective Altruism circles in the late 2010s and became more mainstream during the early 2020s with the public rise of large language models and broader tech discourse. It has since evolved into semi-ironic cultural shorthand, sometimes appearing in bios or prompts on dating apps as a personality signal or compatibility filter—e.g., “P(Doom) < 10% only” or “high P(Doom) realist”—where it functions as a playful indicator of one’s worldview, tech skepticism, or appetite for existential debate.