The ubiquity of generative artificial intelligence (AI) (GPT 4, Bard, Copilot, etc.), the spectacular increase in the value of companies in the sector and the massive layoffs in large technology companies have generated concern about the impact that this technology may have on the labor market. But how founded is this fear? Let’s go in parts.
First of all, the great stir caused in November 2022 by the spectacular performance of ChatGPT has caused anything that smells of artificial intelligence to skyrocket in valuation. At the beginning of 2023, the company OpenAI, creator of ChatGPT, was valued at 29 billion. Last week its valuation was 80,000 million. Not bad as a revaluation in a year. This same trend is observed in any company that is dedicated, even marginally, to artificial intelligence. It even extended to the producers of GPUs (Nvidia, AMD), the most used processors for software based on artificial intelligence algorithms. Recall that on Friday Nvidia said that its net results had increased by 770%. However, investors almost always try to get ahead by betting on companies with high expectations of future profits and they are not always right.
In fact, there are many experts who talk about a huge bubble around the value of companies that produce software based on artificial intelligence algorithms. What is clear at this point is that, just as happened in the 1980s with computers, everyone is talking about this technology, but its effects on productivity are not observed, at least not yet. In fact, the cut of 112,000 jobs at Meta, Amazon, Microsoft, Google and Apple from 2022 cannot be attributed to the impact of artificial intelligence. Again it is a question of unmet expectations: all in all, these companies employ 71% more workers than before the pandemic.
Secondly, new technologies have three impacts on employment. There is a displacement or substitution effect of labor, which reduces employment. But there is also a positive effect on productivity that increases income and produces goods of higher quality and lower price with a positive effect on employment. Finally, there is another positive effect on employment derived from the new tasks created by new technologies. Calculating the net balance of employment lost and generated by these impacts is quite uncertain. The first estimates of the impact of computerization, in the early 2010s, indicated that in the United States 47% of all jobs were at risk of substitution, while in the European Union they rose to 54%.
But 10 years have passed since these calculations and this massive effect has not been seen. The initial studies based their methodology on occupations that could be replaced. However, it is not occupations but certain tasks that can be substituted. When the study is carried out by tasks, the proportion of automatable jobs in OECD countries falls to 9%. More recently, the World Economic Forum estimates that artificial intelligence will destroy 85 million jobs by 2050 but 97 million will be created. This vision is shared by economists surveyed by the CfM-CEPR in 2023. Almost 80% consider that artificial intelligence will increase growth to 4-6%. 60% of those surveyed indicate that it will not have a significant impact on unemployment in high-income countries and could reduce the number of hours worked and increase leisure hours.
Thirdly, it is important to consider the distributional consequences of new technologies. At this point there is a clear differential aspect between the impact of computerization and robotization of the past and the effect of generative artificial intelligence. Robotization produced an evident polarization of the labor market. The most affected workers were those in routine jobs with an intermediate level of skills and training. Lower-skilled jobs were partly protected: a robot took 20 minutes to fold a towel that a hotel worker would do in a few seconds. However, many of the tasks performed by staff with an intermediate skill level were clearly threatened: an ATM provided cash at a much faster rate and with fewer errors than a human teller, and a robotic arm successfully replaced a qualified factory worker. In this way the labor market became polarized, harming the intermediate stratum.
The spread of generative artificial intelligence means that increasingly cognitively complex tasks, and more creative jobs, can be automated. This means that the segments of the labor market most affected by substitution will be those with the highest qualifications and, in that sense, the impact on employment will be very different from the impact of previous technological revolutions. This could mean a reduction in polarization. In fact, in recent years in the United States there has been a rapid increase in the salaries of workers with lower incomes and a stagnation or fall in the incomes of workers who earned the most. However, it is also true that the factor of creating new tasks complementary to the actions of artificial intelligence will have a more positive impact on those more adaptable workers who tend to be those with the highest level of qualifications.
In summary, we can say that research on the impact of artificial intelligence on employment is increasingly voluminous but also quite inconsistent in its results, with a high degree of uncertainty associated with the fact that the benefits of these technologies are not yet known. totally visible and are difficult to predict. What there is no doubt is that the world of work is undergoing a revolutionary transformation.
