For years, artificial intelligence (AI) was seen as a direct threat to the radiology profession. However, three years after the rise of ChatGPT, reality has disproven the most pessimistic forecasts: today there are more radiologists than ever, and their salaries continue to rise. The story of radiology perfectly illustrates how technology can transform sectors without necessarily destroying jobs. In 2016, Geoffrey Hinton, a pioneer of AI, stated that "we should stop training radiologists" because deep learning would surpass humans in five years. However, a decade later, the demand for radiologists continues to grow. In 2025, U.S. hospitals offered a record 1,208 radiology training positions, a 4% increase from the previous year. Additionally, radiology has solidified its position as the second highest-paying medical specialty in the U.S., with average earnings of $520,000 per year, a 48% increase since 2015. According to the American College of Radiology, the upward trend in employment will continue at least until 2055. In Europe, the situation is similar: the demand for radiologists is increasing, and 80% of professionals see AI as an ally, not a threat. A recent study by Vanguard Group, one of the largest fund managers in the world, debunks the idea that AI destroys jobs en masse. Analyzing 140 professions with high automation potential, the report reveals that employment in these sectors grew by 1.7% between 2023 and 2025, compared to 1% recorded between 2015 and 2019. Salaries also rose by 3.8%, compared to 0.6% in the previous period. Why is this happening? "AI rarely automates a complete job, but it does optimize specific tasks," explains Jaime Sevilla, founder of EpochAI. In radiology, AI streamlines image diagnosis, allowing professionals to focus on other critical functions, which increases their productivity and, consequently, their salaries. This phenomenon, according to Sevilla, is being replicated at a macroeconomic level. Miguel Otero from the Real Instituto Elcano agrees: AI is replacing tasks, not entire jobs. Since the emergence of ChatGPT, it has been shown that those who learn to integrate AI into their work are the ones who benefit the most. The case of radiologists could become a paradigmatic example of how technology, far from eliminating professions, can enhance them. The reality is more complex than anticipated by tech giants like Microsoft or OpenAI. A similar example is the autonomous car, which still faces significant challenges for mass adoption. Other factors also play a role: insurers remain cautious about AI-based diagnoses due to error rates, and health regulations are advancing slowly. Additionally, AI specializes in a limited number of pathologies, while the work of a radiologist encompasses much more than image analysis. However, not all sectors are faring the same. Jobs that can be done remotely and involve well-defined tasks—such as data analysts, lawyers, programmers, or translators—are the most susceptible to automation. Junior positions, in particular, have been affected: in the U.S. tech sector, hiring for entry-level profiles fell by 50% compared to pre-pandemic levels, and starting salaries have decreased by up to 6.3% in companies exposed to AI, according to data from SignalFire and IESE. Despite these changes, Sevilla argues that AI will drive economic growth, with rates between 2 and 20 times higher than the historical average of 3% per year. This view contrasts with that of economists like Daron Acemoglu, who predicts a much smaller impact. Sevilla argues that the most recent data shows a positive and tangible effect on productivity. Nevertheless, the relationship between AI and unemployment is still unclear. The unemployment rate in the U.S. reached 4.6% in November, the highest since 2021, but experts attribute this increase to economic and political factors rather than automation. "If someone claims that AI is already affecting unemployment, they are speculating," concludes Sevilla. The true impact of AI on the global economy will only become evident over time. Source: elconfidencial.com