← 返回 Avalaches

2020年,在圣地牙哥,斯坦福数位经济实验室负责人Erik Brynjolfsson与西北大学经济学家Robert Gordon 打了场赌局:预测2020至2030年间美国劳动生产力年均增长是否超过1.8%。Brynjolfsson押注「高于1.8%」,Gordon押注「低于1.8%」,赌注400美元将捐给慈善机构。核心在于AI是否能显著推升经济。美国非农工人每小时产出指数(2017=100,已季节调整)在2020年后加速上升,表明生产率增幅正在回升,且经济福祉与此高度相关。

文章以美国总体生产率与主观体感为例,指出生产力是最能反映长期生活水准上升的指标。农业案例中,Arizona Surprise的Blue Sky Organic Farms经营者David Vose过去以人力防鸟:人工时薪约22美元、班次可达10小时,生长季每月防鸟支出可超过10,000美元。ASU工程师团队开发的AI驱鸟系统由Padma AgRobotics首席执行官Raghu Nandivada主导,设备约30,000美元,采订阅试用;在少量投入下可在数个月内回本,并以自走移动、避开幼苗的机制取代大量巡场人力。文中也提到现场还有人员以5加仑(约18.9升)水桶敲击作为传统驱鸟方法。

这一轮扩散同时揭示替代与再分配:Gordon指出1985年Excel上线曾「淘汰」大量簿记员,虽后续衍生金融分析等职位。当前Tech裁员环比同期上升40%(截至三月),且初阶职位聘用几乎停滞;他预估未来五年约三分之一美国白领工作将被改造或消失。2010年代后期以来的AI浪潮虽推升总体生产力:2020到2026年已高于2%年均,但MIT研究显示95%的企业在大规模AI投入后仍无可测量ROI,显示「技术到生产率」存在时间差;而在Meta、Microsoft、Amazon短期内仍出现裁员,显示获利与就业调整仍在重构期。

b1608f116a97.png



In 2020, in San Diego, Erik Brynjolfsson, head of the Stanford Digital Economy Lab, and Robert Gordon, an economist at Northwestern University, made a bet on whether U.S. labor productivity would grow above or below 1.8% annually on average from 2020 to 2030. Brynjolfsson bet above 1.8%, Gordon below, and the $400 stake was donated to charity, but the real significance is whether AI can materially lift the economy. The context highlights that the U.S. nonfarm workers’ output-per-hour index (2017=100, seasonally adjusted) has accelerated since 2020; productivity is central for long-run living standards, linking the macro debate to household welfare.

The article gives a concrete example from agriculture. In Surprise, Arizona, David Vose of Blue Sky Organic Farms previously relied on labor-intensive bird control: wages around US$22 per hour, shifts up to 10 hours, and seasonal bird-management costs above US$10,000 per month. A team from Arizona State University and Raghu Nandivada, founder and CEO of Padma AgRobotics, introduced an AI-powered self-driving scarecrow-style system that can navigate seedlings and reduce field labor needs. The system costs about US$30,000, is offered by subscription first, and is expected to pay for itself in a few months for smaller farms. The story also notes older deterrent methods using workers banging 5-gallon (about 18.9 L) buckets.

The tension is between displacement and redeployment. Gordon compares the current AI wave to Excel in 1985, which reduced bookkeeping roles while later enabling new analytical occupations. Tech layoffs were up 40% year-over-year as of March, with entry-level hiring flat; Gordon projects that about one-third of U.S. white-collar jobs may be transformed or eliminated within about five years. A Massachusetts Institute of Technology study reports that despite heavy AI spending, 95% of firms had no measurable return on investment, suggesting a productivity delay similar to the earlier IT paradox. From 2020 to 2026, U.S. productivity has averaged above 2%, yet major firms like Meta, Microsoft, and Amazon still announced major layoffs, indicating that while AI increases scale gains, labor outcomes still require policy and organizational adaptation.
2026-05-12 (Tuesday) · fc73e865b54b84dad4f365bd606f5d16f500c62c