FEATURE ADDRESS

Technological progress is not automatic, and depends on institutions that balance exploring new ideas with scaling them.

Sustained progress requires both exploration and exploitation, which thrive under different institutional conditions. Systems focused only on execution can scale existing ideas efficiently, while excessive experimentation can make coordination and production difficult. Prof. Frey argued that forecasters fall into an “extrapolation trap” when they assume past growth will continue without accounting for whether institutions can generate the next wave of innovation.

Centralised systems can scale existing technologies effectively but often struggle to innovate at the frontier. The Soviet Union achieved rapid growth by adopting and scaling technologies developed elsewhere but contributed little to the personal computer revolution once mass production had run its course. By contrast, America’s decentralised financial system allowed rejected ideas to find alternative backers, as Google did after being turned down by Bessemer Venture Partners in 1999.

Flexible and competitive institutions help economies renew themselves as technologies and industries change. Japan became a global leader in cars and electronics before stagnating in the 1990s, while the United States renewed its technological leadership through global supply chains, competition and antitrust enforcement. China’s success in areas such as EVs, solar panels and batteries similarly reflects intense private-sector competition and room for local experimentation.

AI does not overturn the lessons of history, and making the most of it still requires experimentation, new entrants and the right incentives.

Scaling existing AI models alone is unlikely to sustain technological progress without new ideas and continued human experimentation.Prof. Frey noted that advances beyond existing scaling laws require further breakthroughs, while examples such as amateur Go players defeating leading programs illustrate AI’s continuing difficulty with unfamiliar situations. He compared AI today to the steam engine before James Watt’s separate condenser, suggesting that a critical breakthrough may still be needed to unlock its full potential.

New technologies generate the greatest gains when people and firms have the freedom to experiment and create new products. Economies such as the United States benefited strongly from the computer revolution because firms gave employees greater autonomy, while younger firms were often more willing than incumbents to create entirely new businesses. Prof. Frey stressed that if technological change since 1800 had focused only on automation, societies would have gained efficiency but missed innovations such as cars, vaccines, computers and aircraft.

Rising incumbent power and poorly designed incentives can weaken entry and reduce breakthrough innovation. Greater market concentration, lobbying, incumbent-friendly patent systems and “killer acquisitions” can make it harder for new firms and independent inventors to challenge established players. Prof. Frey also warned that “publish or perish” incentives can encourage researchers to use productivity tools to “drill more holes rather than dig deeper”, contributing to declining research productivity.

National power depends on the capacity to innovate and constrain vested interests, while AI creates new opportunities that must be used wisely.

Long-term prosperity depends on preventing incumbents and vested interests from blocking technological and economic renewal. Prof. Frey linked Britain’s post-war decline partly to cartelisation and contrasted it with Britain’s earlier success in dismantling guilds that had resisted disruptive technologies. His broader point was that progress is never inevitable, as illustrated by the fact that humanity existed for a long time before experiencing an industrial revolution.

AI could make service-led growth more achievable for middle-income economies such as Malaysia. As manufacturing becomes less labour-intensive, services may offer a more important route to growth, while machine translation can widen participation in cross-border services beyond English-speaking economies. AI may also narrow productivity gaps between less and more experienced workers, supporting the relocation of higher-value work from centres such as London and New York to economies including India and the Philippines.

Closer ties between governments and AI firms, together with excessive reliance on AI, create risks for competition and human capability. Prof. Frey was less concerned about competition among AI labs than about governments taking stakes in leading firms, warning that weaker rule of law makes proximity to political power increasingly important. He also cautioned that using AI as a shortcut can weaken learning, since evaluating increasingly convincing AI outputs still requires people to retain their own ability to read, write and think critically.

Quotes

“The greatest potential promise of AI, in my view, is that it will empower a new generation of independent inventors as we had in the 19th century to build those new industries to drive renewal.”

“National power depends on the capacity to innovate, to permit businesses to scale, but it also depends on checking the power of incumbent interests when they try to thwart the forces of competition.”

“Over the long run, hard military power ultimately depends on productive capacity…and uneven technological growth from time to time has led to reshuffles in the great power rankings.”

– Prof Carl Benedikt Frey