Generative AI and the Future of Work

It is well known that digital technological changes affect the labor market, but the impact of artificial intelligence (AI) will be more far-reaching. AI is a general-purpose technology that is touching nearly every sector and occupation. Unlike other digital technologies, it does not just automate routine, noncognitive tasks but in the form of generative AI is even capable of creating novel content (text, images, video, audio) that is indistinguishable from human-created content.

THE PROSPECTS FOR GENERATIVE AI TO HAVE AN UNEQUAL IMPACT ON DIFFERENT REGIONS

It is well known that digital technological changes affect the labor market, but the impact of artificial intelligence (AI) will be more far-reaching. AI is a general-purpose technology that is touching nearly every sector and occupation. Unlike other digital technologies, it does not just automate routine, noncognitive tasks but in the form of generative AI is even capable of creating novel content (text, images, video, audio) that is indistinguishable from human-created content. It does far more than merely describe or interpret existing information.

Some compare the ongoing AI revolution to the industrial revolution, which initially led to substantial job loss but ultimately proved to be beneficial once humankind adapted to it. AI too has spurred a lot of fears about the technology’s potential impact on work. Unions have expressed worries about the deployment of AI in the workplace to monitor workers. Most concerns, however, are related to potential job losses caused by AI.

Past narratives linked automation to disappearing bluecollar jobs. Research on AI, however, shows that so-called “knowledge work”, or labor done by well-paid professionals with a college education, faces the most upheaval. This is particularly true for generative AI. Concerns about AI’s impact on creative jobs is already reflected in the Hollywood strikes, in which unions of actors and writers protested the use of generative AI by film studios, before reaching a deal that – among other things – states that productions must get the informed consent of actors whose digital replicas are used.

The public discourse about AI on both sides of the Atlantic has so far focused on generative AI’s potential for job displacement. This essay argues that AI’s potential for job creation and productivity increases deserve more attention by, in particular, policymakers. They should focus on the unequal distribution of generative AI hubs across regions and on the potential of the workforce of the Global South that contributes to training and developing AI models.

SKEPTICISM IN THE U.S. ABOUT GENERATIVE AI’S IMPACT ON HIGH PAYING JOBS

A recent Pew Research Center survey found that few Americans believe that generative AI will have a major impact on their jobs. Fully 27% of employed adults who have heard of ChatGPT think chatbots will have no impact, while another 36% say it will have a minor impact. Only 19% say it will have a major impact. This perception is surprising, given that research suggests that generative AI will likely have a significant impact on the entire labor market. A Goldman Sachs study found that generative AI could expose the equivalent of 300 million fulltime jobs in the U.S. and Europe to automation over the next several years. Generative AI is also going to accelerate the changes that are already taking place. The McKinsey Global Institute estimated that automation could take over tasks that account from 21.5% of the hours worked in the US economy by 2030, the institute concluded that when that includes generative AI, the figure jumped to 29.5%.

A recent report by the Organization for Economic Co-operation and Development (OECD) found that the occupations most at risk of being automated account for an average 27% of the workforce across OECD countries. These include Germany and the U.S.

However, research also shows that generative AI is likely to augment jobs rather than destroy them. A recent International Labour Organization (ILO) study found that the impact of generative AI on job quality (e.g., work intensity and autonomy) might be larger than its capacity to replace jobs. The study concludes that most jobs and industries are only partly at risk of the effects of automation and are more likely to be complemented rather than replaced by the latest wave of generative AI, which comprises chatbots.

The potential of generative AI to augment jobs is corroborated by other research. A study by researchers at the Massachusetts Institute of Technology on the impact of generative AI on highly skilled workers found that it can improve a worker’s performance by as much as 40%compared with workers who don’t use it. The impact of generative AI on productivity has also been analyzed by other studies. A Google-commissioned study by IW Consult found that generative AI tools could save an employee in Germany an average of 100 hours per year in the future and therefore compensate a large portion of the loss of work that Germany will face due to retirements through 2030.

An area in which an increase in AI-led productivity can already be observed is coding, as software engineers have been early adopters of generative AI. One of the experts with whom the Bertelsmann Foundation fellows had the chance to talk, explained in a virtual session that coding camps have lost their relevance since ChatGPT is now able to produce code in the programming language Python with great accuracy. Despite the large potential of generative AI to produce code and the high uptake of AI tools by software engineers, many people – among them many who work in the field of software engineering - argue that generative AI is not going to replace the jobs of programmers, but augment them. Besides positively affecting software engineering, generative AI is reportedly increasing productivity in sales.

AI can also contribute to job creation, an issue that has so far received limited public attention. According to news reports, AI job posts on the platform Upwork increased by more than 1,000% year-over-year in the second quarter of 2023.

These jobs, however, are not equally distributed geographically. A Brookings Institution study found that more than 60% of generative AI jobs posted in the U.S. in late 2022 and most of 2023 were clustered in only 10 metro areas, with the San Francisco Bay region and New York City leading the list. There seems to be no similar analysis for Europe, but it is fair to assume that a similar geographic concentration of AI jobs is taking place there as well, with cities such as London already identified as hotbeds for AI companies.

CHALLENGES FOR POLICY MAKING: BLIND SPOTS WITH REGARD TO THE FUTURE OF WORK?

While little attention has been focused on AI’s potential for creation of new jobs, there has been much attention focused on the possibility of regulation to minimize the risks associated with AI. In meetings in Brussels and Washington, DC, the Bertelsmann Foundation fellows heard from experts about approaches to regulating AI now under discussion. These include the EU’s AI Act.

Policy makers so far seem to be focused on instruments that would address potential displacement of jobs due to AI, including prohibiting the use of AI; establishing new instruments such as a universal basic income; or giving more bargaining power to groups of workers.

The potential of generative AI to augment jobs does not get much attention from policy makers, even though several scholars have proposed innovative approaches to support augmentation, such as changing tax laws, which currently treat income that uses labor less favorably than income derived from capital.

The geographic distribution of generative AI jobs and the potential to incorporate AI-trained workers in the Global South are two other aspects of generative AI jobs that deserve more attention from policymakers.

The aforementioned Brookings Institution study noted that generative AI jobs will continue to cluster in a limited number of hubs unless policymakers intervene. The study’s authors suggest public-sector investment to help disseminate AI activity in the U.S. through:

• expanding the National Science Foundation’s National Artificial Intelligence Research Institutes program, as universities constitute a widely spread network of hubs
• establishing and enlarging the proposed National AI Research Resource to make essential data and computational capacities more accessible

While the geography of AI seems to be less skewed in the European Union than in the U.S, European policy makers would be well advised to take geographical distribution of jobs into account when deciding how to promote the AI innovation ecosystem in the EU and its Member States.

Moreover, developments in Europe might even provide some inspiration to policymakers on the other side of the Atlantic. In Germany for instance, AI competency does not cluster in a few hubs, but is spread across the country. One example is the network of centers of excellence for AI research. Furthermore, it can be observed that with public-private partnerships, AI can contribute to the emergence of new AI ecosystems such as the under construction large AI innovation park, a collaboration funded by the state of Baden-Württemberg and the private Schwartz Foundation in the city of Heilbronn. According to the developers, this will be the largest ecosystem for AI in Europe. Similarly, public private partnerships to create AI hubs could also be used in other regions that are not traditional homes to the tech industry.

POTENTIAL TO CREATE GENERATIVE AI JOBS IN COUNTRIES OF THE GLOBAL SOUTH

Workers in countries in Africa have long been used to train AI systems, for instance by labeling data that is used to train self-driving cars. Many workers in countries in Africa and elsewhere in the Global South are hired to train the models that form the basis of generative AI applications. Because of this work, these employees are described by some authors as a hidden workforce behind AI.

Western media reports have criticized the low wages that these workers receive but offered scant coverage of an expanded role that they may play in the development of AI ecosystems in their countries. Bertelsmann Foundation fellows had the opportunity to speak with a Tanzania-based social entrepreneur who showcases Africa’s AI potential. He has founded a service that uses AI for tutoring and teaching. Seeing examples like this, where individual founders create innovative AI-based solutions suggest that African countries can be a breeding ground for creative AI applications.

For policy makers, the AI workforce in the Global South does not always have a voice in the political process, such as the G7’s Hiroshima process, in which the G7 members have developed international guiding principles on AI and a voluntary code of conduct for AI developers. Policy makers from countries in the Global South have already expressed frustration about being excluded from decision making and conversations on AI that will have an impact on these countries. For instance, Tanzanian politician Neema Lugangira has demanded that African voices are included in the increasingly complex global discussions on digital policy.

In the absence of a membership in multilateral groups such as the G7, where AI regulation is under discussion, bilateral dialogue between G7 members and countries of the Global South on the future of AI could fill this gap. In the European Union, the Trade and Technology Council (TTC), established in February 2023, could be a forum for this dialogue. The German government’s international digital dialogues with a few governments, including countries from the Global South, can serve as a platform for an exchange of ideas about AI and the future of work. Moreover, the German government is working on its first strategy for international digital policy. Some members of the German Parliament have already publicly demanded that this strategy account for the interests of countries in the Global South. While it has not been announced yet whether the future of work will be a topic that is going to be addressed in the above-mentioned bilateral dialogues, they would certainly provide a good platform to do so.

The opinions expressed in this essay are those of the author and do not reflect the opinions or views of any organization with which the author is affiliated or has been affiliated.

Print