AI relies on data

Regulatory approval for vaccines was not possible without large-scale clinical trials, which relied on the availability of data. In the US, the horrendous spread of Covid generated abundant data for clinical trials. In China, the initially effective pandemic response made Covid too rare for clinical trials, forcing China to rely on data collected abroad in countries such as Brazil. We rightly worry about data privacy, especially when our personal information is in the hands of foreign entities. However, we must remember that data privacy regulation and data localization rules, while essential, come at a cost. In the case of Covid, the cost is our health.

AI and trade need trust

China’s inactivated-virus vaccines are rarely used in Western countries and Western mRNA vaccines are not approved in China. Despite the clear superiority of mRNA vaccines, according to some estimates, 75% of Chinese citizens do not trust them. Distrust has many causes, but in this case the consequences are clear. Distrust of foreign vaccines, and foreign AI-enabled products more generally, severely limits international trade.

Managing AI's risks

Big AI superstars such as Moderna, Google, and Tencent are almost all located either in the US or China. In the future, the rest of the world will pay the US and China hefty royalties for AI-based intellectual property. Restated, the US and China will reap the largest financial benefits from AI at the expense of all other countries. How this exacerbates global inequality will depend on how we manage the ownership of data, intellectual property regimes, and competition policy.

AI through the lens of international trade

My research leads to three sharp conclusions about the benefits of AI for international trade. First, AI has massively increased bilateral digital service trade: Firms that deploy AI increase foreign downloads of their apps by a factor of six. Second, AI deployment has doubled the number of apps used internationally. Third, AI has accelerated the process of creative destruction, meaning the creation of new products and the exit (destruction) of older products. This in turn raised the benefits from mobile app use by 10.6%. AI has resulted in new products, more trade, and greater consumer benefits.

Because AI straddles the worlds of ideas and goods, it is exceedingly difficult for the consumer to know which goods and services use AI. You might be surprised to know, for example, that Google search made minimal use of AI until 2019. However, my research shows that we can track a large swath of AI-related international trade by tracking the digital services on our phones. Each day we use AI to sift through the news, check out friends’ posts, shop online, search the internet, transfer money, and more. For many years now, my colleague Ruiqi Sun and I have been tracking the international use of mobile apps. This international trade in digital services is a trillion-dollar market and so represents a substantial chunk of all international trade.

Several observations about digital trade emerge. First, some of the largest firms in the world (including Apple, Amazon, Google, and Tencent) provide mobile apps. Second, these firms were and are far more internationalized than many others. For example, more than 80% of Google’s downloads come from users residing outside of the United States. Few goods producers attain this level of foreign penetration. Third, by contrast, China’s biggest apps (think WeChat) have much more limited penetration outside of China, in part because of distrust by foreign users. TikTok is a partial exception. Fourth, mobile app producers have a great deal of monopoly power. For example, digital advertising is the largest source of mobile app revenue and Google has used its dominant position to enrich itself. The implication for inequality both within and between countries is disheartening.

AI's costs: The reorganization of global value chains

AI comes with significant costs. We can expect a loss of privacy, reduced competition that leads to rising inequality both within and between countries, and an orchestrated degradation of trust in our domestic and international institutions. This will damage our ability to deliver sustainable and inclusive growth to citizens of the world. In my view, it is the degradation of inclusive institutions such as democracy that will ultimately do the most harm. This view stems from years of research indicating that while the proximate source of the wealth of nations is innovation, the ultimate source of wealth is the effectiveness of national and international institutions that support innovation and equitably distribute its benefits.

What of job losses? To date there is little evidence of widespread job loss. This is not for lack of studying the issue. Rather, it stems from two features of AI. The first starts with an understanding that a job is a collection of tasks. While AI can do many tasks, it is rarely capable of doing all the tasks needed to complete a job. As a result, even sectors such as retail have not seen a collapse in employment. Rather, jobs have shifted from bricks-and-mortar to e-commerce operations. The other feature of AI is that its adoption often requires major changes in how firms organize their activities. We know from past emergence of major technologies such as the steam engine, electrification, and information technology that firms take about a generation to make these changes.

We should not expect a major AI-induced change in the location of jobs or the connectedness of jobs within global value chains because AI displaces tasks rather than jobs and AI adoption involves slow organizational change. AI-induced re-shoring and loss of developing-economy jobs are storm clouds on a distant horizon.