Effective training compute required for AI systems capable of automating all economic tasks (FLOP, in 2025-equivalent effective FLOP)
2 events · 0 assessments · 1 decision
Structured and estimated
First pass. Set the quantity definition (2025-equivalent eFLOP, capability requirement only). Recorded the GATE figure (1e33 to 1e41, median 1e36.5) as an estimate on the existing instance. Added three arguments: a floor from the present shortfall (certain), anchoring plus the compute gap (slight), and rapid horizon gains (slight). Created three subclaims: the shortfall claim (a deferred stub), the 20%-to-100% compute gap (the crux), and the single-effective-compute framework (assumes). Linked the existing horizon-doubling claim b5ba920f. Estimate: median 3e33, 90% interval 3e28 to 1e40, with 15% outside for ill-definition. Importance 0.6, contestation 0.7. There are no dependents to notify (new quantity). A further pass could read the GATE parameter appendix and recent Epoch/forecaster updates; moderate marginal yield.
Claim entered the graph