工作的未来：PSE-CEPR政策论坛2025

PSE - CEPR Policy ForumJune 18-20 2025 Keynote Lectures byBarbara PetrongoloDavid Autor CEPRPRESS Centre for Economic Policy Research187 boulevard Saint-Germain 2 Coldbath SquareLondon, EC1R 5HL, UK Tel: +44 (0)20 7183 8801Email: cepr@cepr.orgWeb:www.cepr.org Copyright ©CEPR Press 2025 Introduction The third PSE-CEPR Policy Forum took place on the Paris School of Economics campus from June 18 to 20,2025. This unique event is dedicated to discussing emerging issues with leading researchers and policymakers,reaching a broad audience that includes government officials, legislators, media, and academia. Jointly organisedby the Paris School of Economics (PSE) and the Centre for Economic Policy Research (CEPR), the Forum features The three days of the 2025 Policy Forum focused on “Artificial Intelligence and Labor Reallocation” (Day 1),“Working Conditions and Remote Work” (Day 2), and “Gender Inequality in the Workplace” (Day 3). This documentcontains the transcripts of four keynote lectures given on those days by Barbara Petrongolo, David Autor, Steven Contents 1Opening Lecture,Barbara Petrongolo............................................................................................12 Expertise, Artificial Intelligence, and the Work of the Future,David Autor................................ 11 1Opening Lecture Barbara Petrongolo University of Oxford and CEPR Barbara Petrongolo’s keynote explores three central questions shaping today’s labour markets. First, sheanalyses the disruptive potential of AI, noting that its unprecedented pace of adoption and capacity fortask automation and augmentation raise deep uncertainties about winners and losers, and emphasisesthat technology’s impact depends on human and institutional choices. Second, Petrongolo examines thetransformation of work organisation since COVID-19, highlighting the shift toward hybrid work (25–30% ofworking time now from home) and the resulting trade-offs for productivity, career progression, well-being, I will provide a brief overview of some of the main questions and challenges facing today's labour markets.Let me start with some very broad and simple data, some aggregate trends for the US labour market. I’llbegin with some relatively good news. You can see in Figure 1 the real median wages of full-time employeesin the US since 1979. You can see a period of stagnation between the early 1990s and 2015. This stagnation workers? Figure 2 looks at inequality. There are two series here — again, these are very rough measures. Theblue line shows the wage differential between college and high-school graduates. After the secular increasein the 1980s, this gap has been essentially stable for more than 20 years. Inequality in the middle-to-toppart of the distribution hasn’t been rising. I’m not saying anything here about what is happening at the Finally, let’s look at employment-population ratios in Figure 3. The red line refers to prime-age workers(25–54). After the declines associated to the Great Recession and COVID-19, the employment-populationratio for this group appears to have returned roughly to its 2000 level. But for the overall working-age The first theme of the 2025 Policy Forum concerns the question:What does the future of work look like,given the labour-market impacts of new technologies?This naturally brings us to the central topic of AIand the future of work. Discussions of AI invariably combine both excitement and apprehension. This is notnew; every major wave of technological innovation has been characterised by that same mixture. We canthink back to the power loom, the computer revolution, or the spread of automation. If we were to create aword cloud of terms associated with AI and labour markets, we would find expressions such as “enhancinghuman knowledge,” “augmenting expertise,” and “expanding capabilities.” But we would also see terms like“job displacement,” “surveillance,” “misinformation,” and “manipulation.” So where are we heading? Howare these hopes and concerns taking shape? The idea that “this time is truly different” has accompaniedmany previous technological transitions. We can ask the same question again today:Is this time really Something that we have seen time and again in the discussion about technological progress is that thereare always winners and losers. Throughout history, new technologies have replaced some tasks, augmentedothers, and created entirely new ones. The computer revolution is a classic example. The same applies to AI:it can either reduce or increase the demand for human expertise. We can think of routine or quasi-routinetasks in manufacturing or data-intensive sectors, as opposed to more intellectual or semi-intellectual tasksin healthcare, finance, and consulting. What is particularly concerning in the case of AI is that, at thisstage, our understanding of who will ultimately benefit or lose is largely speculative. There is substantial Another feature that AI shares with previous wave