This is the submission version of my review of Acemoglu and Johnson’s Power And Progress in Economic Record. The published version is available here [paywall]. To buy the book, go here.
“Technological progress… leads to material abundance for everyone”, writes the entrepreneur turned venture capitalist Marc Andreesen in his Tech Optimist Manifesto, articulating a view that has become almost as popular among politicians as it is in Silicon Valley. Its theoretical underpinning, in so far as it has one, is the idea of a “productivity bandwagon”. Supposedly, new tools and new production methods drive productivity growth, which in turn drives “wage growth, and the creation of new industries and new jobs”, pulling forward not just entrepreneurs and the venture capitalists who fund them, but the whole of society (Andreesen 2023). It is this “simple and powerful idea” (p. 14) that MIT professors Daron Acemoglu and Simon Johnson have come to bury with Power And Progress.
The book’s central thesis is that improvements in living standards during the last millennium were not the inevitable result of advancements in science and technology. Rather, they required deliberate political choices and the wielding of state power to ensure the benefits of productivity gains were shared. Absent such interventions, the authors argue, technological progress consistently produces outcomes that are not in the interests of wider society – including inequality, the exploitation of workers, rent-seeking, and threats to public health – while benefitting only the narrow elite that controls the technologies of the day. There are obvious implications for our own time, when seven of the world’s ten largest corporations by market capitalization are tech firms. Where Andreesen wants “compromised and corroded and collapsing” institutions to get out of the tech industry’s way, Acemoglu and Johnson want to see them reinvigorate and reassert themselves, so that digital technologies are harnessed for public good, rather than just private gain.
Power And Progress is a long book and a large part of it is devoted to historical cases. For instance, we learn that in medieval England, productivity-enhancing agricultural innovations like water wheel-powered grain mills did not result in higher wages or living standards for peasants. Instead, the benefits were captured by elites – principally the church, which frittered them away on grandiose cathedral-building projects (Chapter 4). Similarly, the productivity-enhancing application of steam engine technology to coal mining in the 18th and 19th centuries led to more hazardous working conditions for tens of thousands of child labourers by increasing the depth to which mineshafts could be dug (Chapter 6). Workers saw no increase in real incomes at the same time as they lived with the externalities of the industrial revolution – pollution, slum housing, and tuberculosis among them. The authors remind us that these conditions did not improve automatically, but only through a “contested process of political and economic reforms.” (p. 196)
The book is lively, engaging and enjoyable, and it seems to be reaching the general readership it is written for – the audiobook is available for free to Spotify subscribers, and the New York Times technology editor Kevin Roose mentioned he was reading it on the paper’s tech podcast. I certainly hope Chapters 1-8 of Power and Progress will be widely read by journalists, policymakers, and civil society actors, as well as people who work in tech. The broad sweep of the cases across time and space is a valuable contribution to current debates about technology regulation, which tend to lack historical sensibility. With this context, techno-optimist assertions about the relationship between technology and progress sound decidedly less plausible on their own terms, and rather more like a narrative serving “the vision and interests of those who hold power” (p. 107).
Chapter 9, which deals with Artificial Intelligence (AI), is where the authors lose their way. A key distinction made earlier in the book is between applications of technology which increase worker marginal productivity, benefitting wider society, and automation, which replaces workers with machines, benefitting only elites. Acemoglu and Johnson treat AI as an overhyped form of automation and dismiss claims that it can significantly improve productivity as an “illusion” (pp. 299ff). This analysis already looks dated: AI “co-pilots” – software applications designed to augment human capabilities – have demonstrable potential to deliver real improvements to worker marginal productivity in ways the authors ought to approve of. 2022 research by the developer platform GitHub found that developers using its co-pilot completed coding tasks 55% faster (GitHub 2022), while a working paper by Eric Brynjolfsson and the authors’ MIT colleagues Danielle Li and Lindsey R. Raymond reported similar benefits for customer support agents using a generative AI-based conversational assistant (Brynjolfsson et al 2023).
A further criticism I have of the latter part of Power And Progress is that it sometimes succumbs to what the Science and Technology Studies scholar Lee Vinsel has termed “criti-hype” – criticism which takes the claims of technology boosters at face-value, paradoxically amplifying their hype, with obfuscating effects (Vinsel 2021). A discussion in Chapter 10 draws on Shoshanna Zuboff’s The Age of Surveillance Capitalism (an influential but empirically-weak theory of digital dysfunction),blaming the advertising-based business models of big tech companies for everything from far-right online radicalisation to the ethnic cleansing of the Rohingya people in Myanmar (Zuboff 2018). And yet it is clear the authors have misunderstood fundamental aspects of digital advertising: they suggest Google ad revenues ($65.1 billion in 2021) are predicated on the large-scale collection of user data for use in ad personalization (p.367), when in fact Google’s ability to react to user search queries in real-time largely obviates the need for personalized targeting. (Crudely, when I use Google to search for air fryers, Google can simply show me ads for air fryers.) The issue with this kind of mistake is that it suggests Google’s market power could be tackled through stronger data privacy regulations, obscuring more serious problems of gatekeeping and vertical integration (Gilbert 2021).
Policy recommendations are sketched out in the concluding chapter. Some are both original and easy to agree with: Acemoglu and Johnson propose tax reforms to remove the accounting incentive for firms to invest in buying machinery instead of creating jobs, and government subsidies for public-interest applications of digital technology (similar to those available for clean tech). Others are more well-worn. The authors support the expansion of data privacy rights, and antitrust action against big tech companies on the model of the Sherman Act (1890), which was used to break up the robber barons’ oil, railroad and banking monopolies at the start of the 20th century.
Endorsing calls for the repeal of Section 230 of the Communications Decency Act, which protects tech companies from lawsuits when users post illegal material on their platforms, feels like another misstep. The authors claim removing Section 230 protections would disincentivize the promotion of “controversial and sensational material” on platforms like Facebook and YouTube (p. 412), but they do not seem to have considered its other predictable consequences: the over-policing of user-generated content, which they find distasteful in China (pp. 345ff), and the high costs of compliance, which would put smaller platforms, websites and blogs at an even greater disadvantage relative to big tech firms (Morrison 2023). As with the dismissive treatment of AI and the “criti-hyping” of digital advertising, I fear detachment from current technical and market realities will give techno-optimists an excuse for rejecting the book’s main arguments and recommendations. It is all too easy too imagine Marc Andreesen responding, you obviously just don’t get it.
SAM GILBERT
Bennett Institute for Public Policy, University of Cambridge, Cambridge, UK
REFERENCES
Andreesen, M. (2023), The Tech Optimist Manifesto, available at https://a16z.com/the-techno-optimist-manifesto/
Brynjolfsson, E., Li, D., and Raymond, L.R. (2023), “Generative AI at Work”, Working Paper 31161, National Bureau of Economic Research
Gilbert, S. (2021), Good Data: An Optimist’s Guide to Our Digital Future. London: Welbeck Publishing Group Limited.
Github (2022), #Quantifying GitHub Copilot’s impact on developer productivity and happiness”, available at: https://github.blog/2022-09-07-research-quantifying-github-copilots-impact-on-developer-productivity-and-happiness/
Morrison, S. “Section 230, the internet law that’s under threat, explained”, Vox, available at: https://www.vox.com/recode/2020/5/28/21273241/section-230-explained-supreme-court-social-media
Vinsel, L. (2021), “You’re Doing It Wrong: Notes on Criticism and Technology Hype”, available at: https://sts-news.medium.com/youre-doing-it-wrong-notes-on-criticism-and-technology-hype-18b08b4307e5
Zuboff, S. (2018), The Age of Surveillance Capitalism: The Fight for the Future at the New Frontier of Power. New York, NY: Profile Books.
Sam Gilbert // Good Data 