Scott Alexander reviews competing methodologies for predicting AI timelines, focusing on Ajeya Cotra's biological anchors approach and Eliezer Yudkowsky's critique. Longer summary

Scott Alexander reviews Ajeya Cotra's report on AI timelines for Open Philanthropy, which uses biological anchors to estimate when transformative AI might arrive, and Eliezer Yudkowsky's critique of this methodology. The post explains Cotra's approach, Yudkowsky's objections, and various responses, ultimately concluding that while the report may not significantly change existing beliefs, the debate highlights important considerations in AI forecasting. Shorter summary

Scott Alexander explores how straight-line trends on graphs might mask the true impact of interventions, using examples like the Clean Air Act and Moore's Law to illustrate the complexity of interpreting such data. Longer summary

Scott Alexander discusses the difficulty of interpreting trend lines on graphs, particularly when evaluating the effectiveness of interventions or policies. He uses examples like the Clean Air Act, OSHA's impact on workplace safety, and Moore's Law to illustrate how straight-line trends can persist despite significant interventions or technological advancements. The post suggests that these trends might be maintained by control systems, where various factors adjust to keep the trend consistent. This perspective complicates the assessment of policy effectiveness and technological impact, as their effects might be visible in other areas rather than directly on the graph. The author expresses uncertainty about how to distinguish between scenarios where interventions truly don't matter and those where they're part of a complex control system. Shorter summary

Scott Alexander analyzes a paper suggesting scientific progress is slowing relative to researcher numbers, arguing this trend is expected and possibly beneficial. Longer summary

Scott Alexander discusses a paper by Bloom, Jones, Reenen & Webb (2018) that suggests scientific progress is slowing down relative to the number of researchers. The paper shows that while progress in various fields (e.g., transistor density, crop yields) remains constant, the number of researchers has increased exponentially. Scott argues that this constant progress despite exponential increase in inputs should be our null hypothesis, as expecting proportional increases would lead to unrealistic outcomes. He suggests that the 'low-hanging fruit' explanation is most plausible, where early discoveries were easier to make. Scott also warns against trying to 'fix' this trend, as it could lead to dangerous consequences if scientific progress accelerated too quickly. Shorter summary

Scott Alexander examines a paper suggesting scientific progress is slowing relative to researcher numbers, arguing this trend is expected and potentially beneficial. Longer summary

Scott Alexander discusses a paper by Bloom, Jones, Reenen & Webb (2018) that suggests scientific progress is slowing down relative to the number of researchers. The paper shows that while progress in various fields (like transistor density, crop yields, and economic productivity) has remained constant or grown linearly, the number of researchers has increased exponentially. Scott argues that this constant progress despite exponential input growth should be our null hypothesis, as the alternative would lead to unrealistic scenarios like immortality or 50% annual GDP growth. He suggests that the 'low-hanging fruit' explanation is most plausible, where easier discoveries are made first, making further progress increasingly difficult. Scott also cautions against trying to 'fix' this trend, noting potential dangers of accelerated scientific progress. Shorter summary