Longterm Wiki

Expert Positions10 topics

Topic	View	Estimate	Confidence	Date	Source
P(doom)	Very high	>90%	high	2022	—
How Hard Is Alignment?	Extremely hard	5%	high	2022	—
Current Approaches Scale	Very unlikely	5%	high	Jun 2022	AGI Ruin (2022)
Inner Alignment Solvability	Very hard	Requires major theoretical breakthrough	high	2022	—
Likelihood of Deceptive Alignment	Very likely	85%	high	2022	—
Would Misalignment Be Catastrophic?	Near-certainly catastrophic	95%+ → extinction	high	2022	—
P(AI Catastrophe)	Very high	>90%	high	Jun 2022	AGI Ruin (2022)
Takeoff Speed	Very fast	Weeks-months	high	2022	—
Will Advanced AI Be Deceptive?	Likely deceptive	85%	high	2022	—
Will We Get Adequate Warning?	Unlikely	10%	high	2022	—

Self-taught; no formal university education

Dimension	Assessment
Primary Role	Senior researcher and co-founder at MIRI; public advocate for AI risk policy
Key Contributions	Early articulation of the AI AI Alignment problem; Coherent Extrapolated Volition (2004); founding of LessWrong; extensive writing on decision theory and rationality
Key Publications	"AGI Ruin: A List of Lethalities" (2022); "Pausing AI Developments Isn't Enough. We Need to Shut it All Down" (TIME, 2023); Harry Potter and the Methods of Rationality (2010–2015); the Sequences (2006–2009)
Institutional Affiliation	Machine Intelligence Research Institute (MIRI), Berkeley, California
Influence on AI Safety	Among the earliest researchers to systematically formulate the AI alignment problem; his writings and community-building through LessWrong influenced a generation of AI safety researchers and, through them, major funders including Coefficient Giving (formerly Open Philanthropy)