Considering generative artificial intelligence (AI), such as ChatGPT, as a technology that will eliminate jobs is a mistake, says Swedish-German economist Carl-Benedikt Frey, professor at the University of Oxford, in the United Kingdom. By helping workers with a range of tasks, AI tools could facilitate access to more skilled jobs and increase productivity, he says. On the other hand, they must reduce average wages, which will have positive and negative effects.

— I am optimistic about the potential of technological change to boost productivity and increase living standards, on average, but, in addition, technological change has redistributive effects — says Frey.

Impacts of automation

Frey and engineer Michael Osborne, also a professor at Oxford, are authors of one of the most cited academic works on the impacts of automation on the job market, published in 2013. The study estimated that 47% of workers in the United States were in jobs that could be eliminated within two decades.

At the time, the technology buzz was all about self-driving cars, and professional drivers were among those most at risk.

If today this projection seems exaggerated, the success of ChatGPT from the end of 2022 has once again put the future of work under the spotlight.

In March last year, a study by Goldman Sachs estimated that 300 million jobs worldwide could be automated with generative AI. In June, the McKinsey consultancy estimated, in a report, that the benefits of generative AI could have an impact of US$4.4 trillion on the global economy, but would mainly affect professionals who work with knowledge.

Now, Oxford researchers have returned to the topic, in an article recently published in an academic journal at Brown University, in the USA. For them, ChatGPT and its peers should not change perspectives much, because they simply recombine content already produced by Humanity. More than replacing, they help workers.

Greater competition

According to Frey, in the last decade the focus of technological advances has shifted from “the ability to perform routine activities based on rules, technically specified in computer code”, to “the ability to allow more qualified professionals to become more productive and export their ideas and services more easily.”

— Generative AI, in its current form, is essentially lowering the barriers to entry into different activities. If you’re a great writer, you won’t benefit much from it, but if you’re not that good, the tool can make you an average writer — argues Frey.

This help could be positive, allowing less qualified workers to become able to take on jobs previously beyond their reach. This positive point has already been highlighted by American economist David Autor, from the Massachusetts Institute of Technology (MIT), who suggests, in recent work, that generative AI could help middle-class professionals.

On the other hand, the greater number of workers capable of taking on a specific qualified job will tend to reduce the average wage for the occupation in question — the greater the supply of labor, the lower the wages. Professionals who already work in the affected jobs will lose out.

In the article, Frey and Osborne make a comparison with the creation of Uber. Greater competition caused taxi drivers’ hourly earnings to fall, on average, by 10% in the cities where the app started to operate, they point out.

— With Uber and Google Maps, anyone with a driver’s license can become a taxi driver and they don’t even need to know the name of every street in São Paulo or Rio. This means more people driving, as more people can make a living this way, but it also means more competition — says Frey.

For him, the same could happen with generative AI, although the effect on average wages depends on the demand for labor. The quality of the technology’s translation capabilities and the help it provides with some tasks, such as writing programming codes, could allow workers from emerging countries to gain employment in companies in developed countries.

Work without borders

In the information technology (IT) sector, this is already a reality, intensified by the increase in remote work during the pandemic. Programmers who live in San Francisco, in the USA, a city with one of the highest housing costs in the world, could lose their jobs to professionals from Brazil, who may accept to earn less.

— After the pandemic, more people are working remotely. This means work can be done from anywhere. And we have big differences in labor costs, depending on the location — explains Frey.

At this point, generative AI can be “an equalizing technology”, by increasing the productivity of workers in the lowest income brackets, especially in emerging countries.

— There will be a very powerful incentive for firms to outsource more tasks, jobs and activities abroad. Then the company can hire someone in Manila (in the Philippines), in Cairo (in Egypt) or in Bangalore (in India). This could be really positive for developing countries with a highly developed IT infrastructure and a reasonably educated population — says the Oxford professor.

In this scenario, explains Frey, job creation in emerging countries will occur “at the expense of workers in advanced economies.” In other words, there is the potential to level workers’ incomes when comparing one country to another, but not necessarily within each national labor market.

Economist Leonardo Monasterio, researcher at Ipea and professor at IDP, estimates, however, that the impact could be greater than Frey and Osborne predict.

First, because the technology is in its infancy — problems like “hallucination” must be solved.

Monasterio also disagrees with the assessment that generative AI only serves to help or increase productivity. The recombination of already produced content is part of human activity, and tools can do this faster and better.

Finally, the economist highlights, the possibilities for applying technology should increase.

Invest in education

But whether the impact of generative AI in eliminating jobs is greater or lesser, for workers in emerging countries to benefit, basic education will need to be improved, reminds Frey.

Monasterio draws attention to another aspect related to education and professional training: will the increase in productivity resulting from the help of AI tools in carrying out a series of tasks be the same if the professional does not know how to do the job?

— The professional can start to have the report already generated, looking nice, without having to sit down and write the report alone. This can take away a skill from him. It is a tool that, to some extent, will not develop skills that are important. To draw very well using AI, the professional may have to have a period of learning how to draw without AI — says the IDP professor.

The low qualification of workers is often cited as a particularity of emerging economies when it comes to replacing jobs with automation technologies.

Vulnerable workers

Telemarketers, car, taxi and truck drivers, collectors, food service clerks, waiters, administrative assistants and office assistants are among the jobs most threatened by automation, according to Frey and Osborne’s classification.

Researchers from the universities of Princeton and New York, in the USA, included university professors, from different areas, in the list of most exposed professions, when updating in March last year, due to the advent of ChatGPT, an index of job exposure to AI.

According to economist Bruno Ottoni, a researcher at FGV, the speed of current technological advances, based on AI, coupled with inequalities in worker qualifications, could have negative impacts in Brazil if there are no advances in education and professional training. The risk is that less educated workers will face structural unemployment, while there will be a shortage of labor in more technological activities.

— There will not be enough new entrants to the job market to fill the more qualified vacancies that will open up. Somehow, it will be necessary to catch those who lost their jobs and retrain them — says Ottoni.