After reading, I was reminded of the saying “technological maturity comes long before society is ready.” The two directions—open-world evaluation and reliable, orthogonal measurement—have indeed been overlooked, and the dual-track approach is worth following, with follow-up.

AIMPACT News, May 16 (UTC+8), Princeton University Professor of Computer Science Arvind Narayanan discussed adaptation strategies for the transformation of knowledge work at a seminar at the Stanford Digital Economy Lab. He said that the potential for AI to automate most cognitive labor deserves serious attention, but that the real bottleneck lies downstream in capability, and that AI’s impact will unfold gradually over decades. He criticized the current evidence infrastructure for overemphasizing capability layers and introduced his team’s efforts to measure characteristics of diffusion-related technologies, including “open world” assessments (testing AI’s ability to handle chaotic real-world tasks) and measuring AI reliability as a dimension orthogonal to capability. In addition, he proposed a forward-looking agenda for a world in which cognitive labor has been automated, aiming to predict changes in labor demand, the risks of institutional collapse, and new social, ethical, and political challenges, and he called for a dual-track approach: developing