人工智能对产出和通货膨胀的影响.pdf

1、The Impact of Artifi cial Intelligence on Output and InflationAldasoro,Doerr,Gambacorta and ReesDiscussion by Yongseok Shin*2025 Bank of Korea International Conference*Washington University in St.Louis and Federal Reserve Bank of St.Louis.The views and opinions ex-pressed here do not necessarily rep

2、resent those of the Federal Reserve Bank of St.Louis or the Federal ReserveSystem.I thank Mark Wynne for useful comments.SummaryImportant,timely questionVery nice paper,full of thoughtful analysis,leveraging existing tools“Useful tool to ground our thinking while remaining openMain pointsInflation r

3、esponse depends on anticipationWeak link between industrys initial AI exposure and output growth(linkages,GE effects;AI exposure similar across industries)AI in the ModelAI:Industry-level productivity shock(permanent level shift)Aggregate TFP assumed to rise by 15%over 10 years(asymptotic BGP only)U

4、se AIOE/AIIE(Felten et al.)to determine industry-level productivity change(more exposure=bigger productivity rise;cardinality?)AIIE turns out to be unimportantIs AI different from ICT?Or from any other technology?Effect of Technology Shock in the Model(1)Effect of productivity shock larger for consu

5、mption goods sector“Policies that foster the adoption of AI.should focus on sectors producingconsumption goods.”(p.33)Hultens(1978)theorem(for undistorted economies)dlogTFPdlogai=Gross Output of Sector iAggregate Value Added|zDomar weightDownstream sectors:Gross output value added“each sector accoun

6、ts for the same share of value added before AI”(p.20)For a fair comparison,each sector should have the same Domar weight.Effect of Technology Shock in the Model(2)Are technology shocks contractionary?(Gal,1999;Christiano,Eichenbaumand Vigfusson,2004;Francis and Ramey,2005;Basu,Fernald and Kimball,20