Superintelligence timelines disagree partly on the bar: automating AI research is not the same as months-long judgement and coordination.

The claim

Part of the disagreement over superintelligence timelines is a disagreement over the bar: if superintelligence requires stability over months-long tasks, consistently superb judgement, balancing conflicting goals and coordinating like a military or corporation, then automating the research behind algorithmic progress can arrive well before it, and "ASI within a year" forecasts and their critics may be measuring different things.