Position: AI Capabilities May Not Increase Exponentially
Abstract
Rapidly increasing AI capabilities have substantial real-world consequences, ranging from AI safety concerns to labor market consequences. The Model Evaluation & Threat Research (METR) report argues that AI capabilities have exhibited exponential growth since 2019. In this position paper, we argue that the data is insufficient to support exponential increase in AI capabilities. We propose an alternative hypothesis that existing exponential growth has been driven by (1) model and data scaling, and (2) the development of reasoning models. Barring major new breakthroughs, our hypothesis is that AI capabilities have already exhibited an inflection point or will do so in the near future. We call for more rigorous evaluation methodologies for AI forecasts and better academic discussion on this topic.