Artificial intelligence promises to revolutionize work, but studies revealed by The Economist shows that AI separates better from others, expanding inequalities and favoring few
At a meeting in Paris on February 10 and 11, technology chiefs competed to make the most grand statements on artificial intelligence. “AI will be the deepest change of our lives,” said Sunda Pichai, alphabet head. Dario Amodei, executive director of Anthropic, said this would lead to “greater transformation in the global work market in human history.” In a blog post, OpenAi’s Sam Altman wrote that “in a decade, perhaps everyone on Earth are able to do more than the most impactful person today.”
Altman’s forecast reflects a line of thought already established. When early language models began to gain popularity in the early 2020s, economists and executives were hopeful that these tools, along with others, would level the playing field, mainly benefiting less qualified workers. Software capable of dealing with tasks such as protein folding and poetry writing would certainly democratize opportunities. Jensen Huang, CEO of Nvidia, a chip design company, imagined a future in which workers “would all be ceos of AI agents.”
However, more recent discoveries have cast doubts about this vision. Instead, they suggest a future in which the most successful will continue to stand out even more-while the others will be left behind. In complex tasks, such as research and management, new evidence indicates that the best performances are better positioned to work with AI (see table). Evaluating the output of models requires expertise and good judgment. Instead of reducing disparities, AI is likely to expand divisions in the job market, much like past technological revolutions.
The argument that AI acts as an equalizer has been supported by research showing that technology increases the productivity of less experienced workers more. A 2023 study conducted by Stanford University Erik Brynjolfsson and Danielle Li and Lindsey Raymond of the Massachusetts Institute (MIT) found that generative AI tools have increased productivity by 34% for starting client support workers, helping them to solve queries faster and more effectively. On the other hand, experienced workers saw few benefits, as AI reinforced approaches they already used. This suggested that technology could reduce gaps by transferring good talented employees to those less skilled.
A similar trend has been observed in other knowledge intensive tasks. Research conducted by Shakked Noy and Whitney Zhang, both from MIT, revealed that less skilled writers have experienced the highest improvements in the quality of their work by using OpenAI chatgpt to write materials such as press release and reports. Many have become quality to improve simply using the unlimited IA exit, highlighting their ability to raise basic performance. Similarly, Jonathan Choi of the University of Southern California, and co -authors found that a general purposes of general purpose has improved the quality of legal work, such as the writing of contracts, especially among less talented law students.
The problem is that this effect is overcome by another. A job can be considered as a set of tasks that technology can both mercantilize and assist. For air traffic controllers, technology is an extension: it processes flight data while leaving decisions in the hands of humans, keeping salaries high. On the other hand, self -service systems simplify the role of boxes, automating tasks such as calculating change. This reduces skill requirements, causing salaries to stagnate.
Thus, despite the initial optimism, customer service agents and other poorly qualified workers may face a future similar future. Your repetitive tasks are susceptible to automation. ServiceNow’s Amit Zury, a business software company, estimates that more than 85% of customer customer service cases no longer require human involvement. As IA advances, this number will probably increase, leaving fewer agents to deal with only the most complex cases. Although AI can initially increase productivity, its long -term impact will be to market skills and automate tasks.
Unlike previous automations, which have replaced routine jobs such as working on assembly and accounting lines, AI can extend its reach to non -routine and creative work. It can tacitly learn, recognize patterns and make predictions without explicit instructions; Perhaps, over time, you can write fun scripts and design useful products. For now, it seems that in high -paying industries, it is junior employees who are most vulnerable to automation. At the A&E SHEARMAN law firm, IA tools now deal with much of the routine work previously done by members or legal assistants. The company’s software can analyze contracts, compare them with previous businesses, and suggest revisions in less than 30 seconds. The best performances have been the most able to use technology to make strategic decisions, says David Wakeling, the firm’s head of AI.
The change in recent economic research supports this observation. Although initial studies suggested that the worst performances could simply benefit from copying AI exits, more recent studies examine more complex tasks such as scientific research, business administration and investments. In these contexts, the best performances benefit much more than their less qualified colleagues. In some cases, less productive workers do not see improvement, or even lose ground.
Intelligent design
Aidan Toner-Rodgers of MIT, for example, found that the use of an AI tool to assist in the discovery of materials almost doubled the productivity of top researchers while there was no measurable impact on the lower third. The software has allowed researchers to specify desired characteristics and generate candidate materials planned to have these properties. Elite scientists, with broad technical knowledge, have been able to identify promising suggestions and discard the bad. Less effective researchers, on the other hand, had difficulty filtering useful exits from irrelevant (see chart 2).
Similar results have emerged in other areas. Nicholas Otis, from the University of California, Berkeley, and co -authors found that strong entrepreneurs in Kenya increased their profits by more than 15% with an AI assistant, while those facing difficulties saw their profits fall. The difference was how they applied AI recommendations. Low performance followed generic advice, how to make more publicity; High performance those used AI to find personalized solutions, such as ensuring new sources of energy during blecau (see chart 3).
In financial decision making, Alex Kim, from the University of Chicago, and co -authors conducted an experiment where participants used IA to analyze results calls before allocating $ 1,000 in a simulated portfolio. Sophisticated investors reached almost 10% larger with AI returns; Less sophisticated investors saw gains of only 2%. Experienced investors have taken better advantage of results of results calls, such as those related to R&D spending, stock repurchase and operating profit before depreciation and amortization.
As the AI reshape the work, new tasks are emerging. Atlassian’s Rajeev Rajan, a corporate software company, says AI tools release a few hours a week for engineers, allowing them to focus on creative work. Junior lawyers spend less time on housework and more on customers. “Really intelligent people who may be bored with routine results reports will benefit more,” says one boss of a large investment company. “The ability that will be most rewarded in the short term is the imagination to find creative ways to use AI.” The manual work of these industries is being automated, allowing junior employees to assume advanced tasks earlier in their careers.
Labor markets have always been defined by the destruction of ancient roles and the creation of new ones. David Author, from MIT, estimated that about 60% of work in the US in 2018 did not exist in 1940. The position of “aircraft designer” was added to the census in the 1950s; “Conference Planner” arrived in the 1990s. But who will occupy AI’s new jobs when they arise? History suggests that technological changes favor the qualified ones. In the Industrial Revolution, engineers who dominated new machines saw their salaries firing while routine workers lost space. The age of computers rewarded software engineers and made obsolete datylographers. AI seems to be following a similar path, benefiting those with judgment, agility and expertise to navigate in complex and information rich environments.
In addition, today’s AI tools are just the beginning. As technology becomes more sophisticated, semi-autonomous agents capable of acting independently-imagined by Huang-can turn workplaces. This can make each worker a type of CEO, such as the NVIDIA CEO predicted. But there will be no leveling: the most talented will be the best CEOs.
Source: https://www.ocafezinho.com/2025/02/20/the-economist-ia-separa-melhores-dos-demais-e-causa-desigualdade/
