The Unexpected Consequences of GenAI for Globalization

Services as the future of trade

Globalization is changing as it moves from manufactured goods to services, which has been observed for a long time. Amid this trend, generative AI (GenAI) has emerged as a new form of AI and is affecting the changing world economy. “The future’s already here, it’s just not evenly distributed,”—and this quote by science fiction writer William Gibson illustrates the situation quite well. The question is which avant-garde projects will become mainstream and which will be dropped.

While AI is definitely not a new thing, GenAI is a relatively new technology. For this presentation, GenAI can be defined as a technology that helps humans generate content, find, summarize, and analyze information. The nature of globalization has already changed and shows us the direction it will be going in. The central fact is that goods trade has already peaked, and services trade has not peaked yet. The goods export ratio has fallen since 2008, but the service export ratio has not. There was explosive growth in globalization through outsourcing and offshoring in the 1990s and early 2000s, but that is no longer the case as globalization is no longer driven by goods. Meanwhile, while world trade has mostly been trade in goods to date, trade in services keeps increasing and can be expected to become the most important part of global trade in a couple of decades. In value-added terms, almost half of world value-added trade is comprised of services, and more than half of jobs connected to exports are in services because they are also more labor intensive than manufacturing.

What are these services? WTO data shows that they are divided into transportation (17%), travel (20%), and other digitally enabled commercial services (63%). Digitally enabled services include financial services, insurance and pensions, IP charges, computer and information technology, and others. Other business services make up almost 40% and this is the part of trade in services that is expanding the most. If globalization was characterized by goods crossing borders and offshoring was characterized by factories crossing borders, then this is characterized by offices crossing borders, where people will do telework for overseas companies.

For providers of digitally enabled service exports the developing world’s share is low but growing fast. Developed countries account for about two-thirds of all exports. India and China are also important players. However, the dominance of the developed countries is changing. A part of the future of trade is going to be emerging markets, and exporting services will be a very large part of that trade. While this development is not new, researchers are starting to notice it now.

GenAI’s effects in the new age of globalization

How will GenAI affect this new form of globalization? It is likely that it will make foreign teleworkers better substitutes for domestic ones, and this will accelerate trade in intermediate services, which are mostly back-office B2B services that are not subject to heavy regulation. To assess GenAI’s role, anchor insights about the technology is useful. First of all, GenAI levels up skills, meaning that everybody gains. Interestingly, the least skilled workers gain the most from using it but not all tasks improve. And unlike many types of productivity enhancing technologies, GenAI fills in gaps. For example, letting ChatGPT write code instead of struggling to write it yourself is a form of professional education that fills in the gap between weak and strong workers. A famous study published in the Harvard Business Review targeting approximately 750 consultants from Boston Consulting Group illustrates this. They were assigned typical consulting tasks and some of them used AI for the work while others did not. GenAI use, in this case ChatGPT, reduced the time on average of all the workers by 25% and increased the quality of the output by 40%. The biggest boost was given to the least skilled workers. Why? I came up with a phrase to explain this phenomenon. “GenAI is trained on the best and given to the rest.” Technologies based on machine learning screen through the data and give weight to better performances—the output and resulting skills can then be used almost for free by workers around the world, which will have implications for global trade in intermediate services. The study also showed that while GenAI helps with brainstorming and first drafts, more time is spent editing the output generated, because the consultants essentially needed to verify the work as they didn’t produce the beginning work. Additionally, for tasks such as solving business problems where reflection is needed, using GenAI actually harmed performance.

In this sense, GenAI does not automate work, it augments workers. GenAI will have economy-wide effects on jobs, employment, and income, but the technology will not operate on its own. Most estimates say that 40% to 45% of the workforce in G7 countries will be affected in some way, especially in jobs related to content creation.

The nature of creating services and GenAI’s potential role

Service production processes are very different from industrial processes. Some researchers approach the topic as if services were an industrial process. In manufacturing, every product has exactly the same quality standards. There may be different models of a product, but production for each model is optimized and predictable. You may need more or fewer workers depending on the degree of automation, but the quality itself remains the same.

Services, on the other hand, are customized output. Almost every service is customized to at least some degree. Utilization of GenAI will not change what workers are doing, but the output quality will be higher. Since such technologies will be widely used, it will not be severely disruptive to the number of jobs or lead to explosive growth in output. Furthermore, GenAI is not a technology that companies purchase as capital expenditure—it is a mere service input, and firms all around the world will use it to stay competitive, not to gain competitiveness.

In the past, smartphones or laptops were seen as innovations that would lead to workers saving time and becoming more effective in their work. However, the actual content of tasks that are done using these devices has not greatly changed and people are not working shorter hours. These technologies were transformative, but not disruptive, and GenAI can be expected to move in a similar direction.

The future of international trade in services

Regarding domestic and non-domestic teleworkers, those in G7 countries tend to be more productive, but also significantly more expensive. Since GenAI reduces productivity and quality differences, it has a leveling effect that will likely make offshoring of telework more promising. In an as yet unpublished study together with Michael Wade, we monitored four groups of online workers: 50 Americans using AI, 50 Americans not using AI, 50 South Africans using AI, and 50 South Africans not using AI. Each of them completed three tasks. They had to draft an email to a new client explaining a product in a culturally sensitive way, come up with a creative idea for a 90-second ad that is adaptable to different settings, and analyze a table of gross sales data for growth trends. The results clearly showed a leveling effect. The gap between the workers from the U.S. and South Africa was always smaller with AI. Furthermore, AI utilization raised the productivity of all participants. As the American workers were paid twice the amount of the South African workers more South African workers would likely be employed next time around. Thus, overall, GenAI made the South African workers better substitutes for the U.S. workers in terms of value. In the second-round of the experiment we took the 600 answers and paid U.S. online professionals to classify them. They had to answer whether the output for each task had been done by a U.S. worker or by someone from a foreign country. They classified 60% of output without GenAI use correctly. However, for output generated with GenAI use, they were unable to tell the difference between South Africans and Americans most of the time. This directly indicates that GenAI turned the South African workers into better substitutes for the U.S. workers by shrinking the gap between them.

GenAI’s potential to facilitate a global tidal wave of talent

Related to large language models and GenAI, one technology that is of particular interest is so-called simultaneous speech translation. This type of translation happens instantly, not sequentially. Many of us have used text translation, which is good in itself, but it is not very helpful when talking to people. While sequential translation is useful for talking, it does not feel natural. Simultaneous speech translation will occur via earphones. The translation will be done almost instantly and therefore feel more natural. The beta versions are already available, and the technology is expected to be widespread soon. This technology will turn foreign teleworkers into even better substitutes for domestic workers. For Japan, for example, it could unleash a large pool of Japanese talent onto the world market for services. Since Japan is well-known for its high level of services, the country has a significant advantage with this technology. Quality matters in the service industry, which is why Switzerland—one of the most expensive countries in the world—is a major exporter of services. It is often not the cheapest service that wins, but the one that provides the best quality.

In conclusion, GenAI contributes to leveling up in terms of productivity and quality while also reducing cultural differences and language barriers. Therefore, GenAI will make domestic and foreign teleworkers better substitutes and thus turbocharge trade in intermediate services. We can expect a global tidal wave of talent which means that for businesses, the pool of talent you can access will expand enormously. So, for countries that possess a world-class service industry, this development will potentially open up new markets.

Comment

IMAZATO Kazuyuki:
Discussions on GenAI often focus on negative aspects, such as potential loss of jobs or social unrest instead of focusing on its potential for applications in the business world. While statements were made on how GenAI functions only as a support for humans and has a limited disruptive impact on industries, especially in the service-sector, discussing the differences between GenAI and conventional automation technologies is necessary. As technology continues to develop at an ever-increasing pace, it is necessary to debate whether the role of GenAI will be limited to a supporting role, or whether it could entirely automate service processes as well as possible synergies with other technologies.

For Japan, these technological advances are not necessarily a threat to society, but rather present welcome opportunities. As Japan faces a population decline, further advances in automation technology will play an extremely important role in resolving social issues. Currently, discussions on an international regulatory framework for GenAI are underway. The EU passed its Artificial Intelligence Act which includes GenAI regulations last year. The potential increase in remote work in developing countries that was mentioned will lead to a decrease in demand for the same work in developed countries, which may stir up social unrest and lead to stronger regulations. In order to fully harness the positive effects of GenAI, establishing internationally harmonized regulatory frameworks is a crucial step.

Richard BALDWIN:
The impact of automation in manufacturing and GenAI in the service industry differs greatly. GenAI does not affect a huge number of low-skilled workers as they do not generate content. There may be some substitution with robots jn new areas, but we didn’t cover that today. In manufacturing, automation has been going on for a long time—more incremental, rather than radical—whereas automation has not really been implemented in offices. The impact will be much flatter. For example, AI has been extensively used in law firms. Around 20 years ago, law firms began using e-discovery to process and analyze legal documents electronically. This led to a reduction of young lawyers hired to read these documents but increased the number of employees without law degrees who are capable of assessing the output. Those with mid-level skills who receive training through AI-based applications will become more employable and productive.

On the topic of regulations, as these are new technologies that have incredibly large implications for our society it is a good idea to regulate them like the EU did. This kind of technology could potentially be quite disruptive to socioeconomic structures. While I am not an expert on AI regulation, the regulation implemented by the EU seemed reasonable. Some industry experts argue that Europe will never be able to catch up with the U.S. due to this regulation, but on the other hand, European consumers will be protected from some of the worst aspects of AI. The number of people working in the field of AI research and development is rather limited, but on the other hand everybody is a consumer. These regulations are put in place to protect consumers, not to promote the production of AI technology.

Q&A

Q:
Some forms of generative AI may actually replace humans—one example raised is AI actors replacing human actors.

Richard BALDWIN:
While this is true, the number of situations where GenAI can actually replace humans is relatively limited. Also, AI does not create such content on its own. For example, if a voiceover is done using AI, there will be a team of people in charge of checking and editing the output. In the past, theaters experienced a similar challenge when movies started to spread. Such changes occur and not every single job can be protected. Ultimately, GenAI will lead to higher demands regarding quality, not fewer people or more output.

Q:
Regarding the relationship with globalization, it was mentioned that GenAI is breaking down language barriers. Will GenAI, in the context of globalization and current changes, lead to a flatter world?

Richard BALDWIN:
The first wave of globalization was goods crossing borders. The second, which my book The Great Convergence was about, was factories crossing borders. This third wave is offices crossing borders. However, the world will never be completely flat. In particular, large economies are relatively closed. They are their own best customers, and this is unlikely to change. The third transformation will lead to higher integration but will not create a flat world.

Q:
What about the possibility of GenAI reducing cultural differences and creating a flatter world?

Richard BALDWIN:
Starting with an anecdote, at a lecture given to executives at IMD on the topic of GenAI I demonstrated the use of ChatGPT during my presentation. People who have never used it before are often impressed when they see it for the first time. I asked ChatGPT to write a two-minute speech to welcome new workers to a company and read the result out loud. One participant remarked that the speech sounded American and asked whether it understood my accent. I then asked ChatGPT to produce versions geared toward a South African and British audience, and the results were tailored somewhat successfully to the different regions I mentioned. GenAI is trained on an enormous database; of course, it has no knowledge about culture or language. But if given the right prompts, it can adapt the output to cultural narratives automatically. For example, this may allow remote workers in one country to be more culturally appropriate when dealing with clients in other countries. Call center employees in India are one example that comes to mind.

Q:
As AI becomes more widespread, what kind of education will be required in primary, secondary, and higher education?

Richard BALDWIN:
For the next five years, what is being done in offices will not change much. Only the means of how these tasks are carried out will change, for example by offshoring certain services. Machine learning is driven by data, so it systematically excels at things that are already being done. GenAI helps us do what we already do in a faster manner or with higher quality, but there is no huge transformation. Many people are thinking that AI will change everything, and the work of humans will become redundant. However, while AI augments the skills and potential of workers, it does not automate work itself. This is a transformative process, not an innovation of products or services.

Of course, a minimum level of digital fluency is needed, but almost all young people nowadays possess this skillset because they use different devices in their daily lives. Previously, I advised young people to try and do things that AI cannot compete with. This still rings true for AI in general, but GenAI is actually an assistant to workers, not a replacement. I encourage students to use GenAI tools because GenAI itself will not take their jobs, but other people using GenAI will. Education from high school on should help young people use these tools to generate higher quality services.

Q:
Regarding disparities in digital literacy, what impact will the digital divide have on the potential use of AI for offshoring?

Richard BALDWIN:
This was a bigger issue for old AI technologies which required skills like programming. GenAI on the other hand is very easy to use without reading an instruction manual first. Smartphone users are using AI features on their devices without being aware that they are using AI. It differs greatly from previous technologies that required a university degree to utilize them. Countries without sufficient electricity and internet connection are at a disadvantage, but as far as skills are concerned the digital gap is actually narrower in the domain of GenAI for advanced economies.

Q:
You mentioned that Generative AI functions as an equalizer. The serious rivalry between the U.S. and China marks a historic change in global trade, leading to a division into two blocks. How does this change affect your argument?

Richard BALDWIN:
The trade conflicts that geopolitics are concerned with are primarily about manufacturing. Specifically, Chinese manufacturing is often a topic in this debate. However, we do not hear similar concerns being voiced about back-office jobs being offshored to other regions—so far, there has been no backlash or high-level political traction opposing trade in services.