How Progress Ends

Economic, social, and technological development occur on a knife’s edge. A pattern emerges among thousands possible in a world improbably sprung from the Big Bang, in societies unfathomably complex, with people whose bodies must be precisely calibrated along hundreds of dimensions just for human action to be possible, let alone interaction, let alone technologically mediated interaction, let alone valuable technologically mediated interaction. But, somehow, it happens. And, as quickly as a society innovates, so fast can its members forget whence and how.

Recent rapid developments in AI in particular have prompted diverse speculations about how it will shape, well, the very future of civilization itself. Will we ultimately be reduced to machine users, heavily reliant on AI to think for us? If so, will we welcome the reduction? Might machines end up using us more than we use them? Who, or what, will have ultimate control?

Carl Frey takes the long view in How Progress Ends, showing why economic, social, and technological advances have long been far from guaranteed. Frey’s book will be of interest to historians, technologists, technology ethicists, economists, entrepreneurs, and policymakers, among others. In impressive historical detail and breadth, Frey offers due caution against so-called Whig history, showing how economic and technological innovation often precede socioeconomic stagnation and failure. About AI in particular, he notes jarringly: “We’ve seen this movie before” (352). “Grand promises about the future of AI were made in the 1980s, which led to a funding boom and media hype quite similar to today” (352). Frey’s tome examines technological development in America, China, and far beyond, focusing largely on the theme that “decentralization is necessary for further progress as a country approaches the technological frontier,” but adding that “this is rarely in the interest of incumbents” (19).

Frey helpfully explains the rise of American economic and technological power. “[I]n corporate America,” we are told, “the coming of the computer age went hand-in-hand with decentralization” (xi). “By the mid-nineteenth century,” says Frey, “America had developed a market for technology unmatched anywhere in the world” (21). This enabled latecomers worldwide to adopt and deploy this technology, which they did with varying success (22, 154, et passim). Evidencing this claim is the vast increase in breakthrough inventions in the United States from 1776–1825 to 1876–1925, with the country far behind Britain and France at first but eventually developing as many breakthrough inventions as did those countries and Germany combined. The cotton gin that Americans invented in 1793, for instance, catalyzed the country’s shift from producing one percent of cotton globally in 1793 to half of cotton worldwide in 1830. Frey also zooms out to make the interesting observation that “while the voyages of Columbus and da Gama were surely important events, there was little or no price convergence before the transport revolution of the nineteenth century” (160). The latter, not the former, radically increased global market integration.

Now should AI-based and other technological innovation be centralized or decentralized? America has one answer. China has another. Frey compares the two, situating his analysis in an admirably comprehensive account of global history. Moving from American technological development to Chinese, he shows that even as state-based data collection in China has cemented its leadership in surveillance technologies, China’s data gathering might be more akin to an oil curse (351). For one thing, China has not concurrently established an “advantage in areas like drug discovery, autonomous vehicles, or large language models” (351).

Frey asks how a “market economy” with “individual actors and small firms,” as Hayek thought about it, might be transformed into “an economy of command-and-control organization in large companies connected by market exchange” (13). “In a pure market system,” says Frey, “there would be no managers overseeing production. Each step would be handled by individuals buying and selling to one another based on changing prices” (11). But today, throughout the world, we do not have pure market systems. Crony capitalism is largely responsible. Corporate lobbying, we are told, leads “to anti-competitive tactics and pressur[ing] the government for regulation that shields them from competition” (18–19). As evidence of this, “politically connected companies take out fewer patents” (19).

Frey’s assessment of the AI-based regulation merits particular attention. “If the present or past is any guide to the future,” he notes, “organisations and societies that embrace decentralization are more likely to push the technological frontier in AI” (355). This understated point about the predictive importance of all current and past data is rather important. For, among other reasons, we live in a world where a single company (Nvidia) maintains a 92% market on chips used to train AI models (368).

The scale and scope of, and public interest in, AI-based innovation depends on numerous considerations. At the end of the book Frey notes a 2023 Pew Research Survey showing that “52 percent of Americans are now more concerned than excited about the growing use of artificial intelligence, while only 10 percent say they are more excited than concerned” (398). Innovation requires both ideation and “converting ideas into practical and reliable products that are available and affordable to the masses” (12). Frey later observes that different cultures, such as the U.S. versus England, are differentially inclined to undistinctive, mass-produced products. Successful innovation and diffusion require the creation of a viable prototype followed by “scaling production, cutting costs, and increasing efficiency—and this is when centralization and corporate consolidation take over” (18).

Of course, not everything that can be said about technological progress should be said in a single book, even one as extensive as Frey’s. Even so, I would’ve been curious to hear more about whether AI-based central planning could itself be decentralized. I mean “decentralized” here in the specific sense of adopting a digitally decentralized system, even if it is located in one central place or within nodes across a single region, perhaps subject to the control of one government or other group. Either way, Frey’s impressive work has challenged me to think that the coming technological revolution could spread either much faster or much slower than many expect and be more or less powerful than many think it will be, including in its future impacts on economic planning. The historical record implies that, even if we can predict eventual stagnation, short-term prediction is no easy task in any immensely complex social world.

I would also be curious to hear more from Frey about whether AI might, put one way, intellectually dominate humanity along a number of dimensions. In 2016, AI roundly defeated Lee Sedol in Go, an extremely complex game, long after IBM’s Deep Blue beat Gary Kasparov in chess. None of this entails, however, that we should offload as much thought-work as possible to AI. For one thing, AI reasons discursively and explicitly, but so much of our thought is not algorithmizable. It is largely intuitive and based on know-how, and so inaccessible, perhaps, to AIs present and future. Moreover, there is counterevidence to consider. For example, an amateur Go player with a simple computer roundly defeated the Go-champion AI by using strategies it had not learned (354). Imagine, then, if a country’s military were reliant on AI for defense, the country seemed militarily secure, and then a second country relied on a pattern hitherto unrecognized by AI to throw the first country’s defenses into disarray.

Technological innovation occurs in fits and starts. Observed progress can be learned from and replicated, and observed failures avoided. AI and associated technologies might indeed speed economic and social development considerably, fueling real progress. Alternatively, they might yield a set of known and hidden costs, broadly construed to include rights violations, that swamp the benefits. What a blessing it is then, as I would put it, to have someone like Frey to turn to for insightful assessment of currently unpredictable, civilization-shaping developments in the years to come.