Survey of AI researchers

Summary

This Wikipedia article covers 'P(doom)', the informal term used by AI researchers and safety advocates to describe their estimated probability that advanced AI leads to human extinction or civilizational catastrophe. It aggregates survey data and public statements from prominent AI researchers, capturing the wide variance in risk estimates across the field.

Key Points

• •P(doom) refers to the subjective probability assigned by researchers to AI causing catastrophic or existential harm to humanity.
• •Estimates vary widely among experts, from near 0% to over 90%, reflecting deep uncertainty and disagreement in the field.
• •Prominent figures like Geoffrey Hinton, Eliezer Yudkowsky, and others have publicly stated their personal P(doom) estimates.
• •The concept is used as a shorthand in AI safety discourse to signal alignment with concern about long-term catastrophic AI risk.
• •Surveys of AI researchers (e.g., AI Impacts surveys) provide aggregated data on how the broader ML community views existential risk.