It's unlikely that artificial intelligence will steal our jobs

In the last two years, following the success of artificial intelligence systems, investments in the sector have increased significantly – in the order of several billion dollars. However, companies have not yet managed to generate as much profit from these services, which continue to have very high fixed costs. The main expense is related to servers and components necessary to develop AI systems. The tech news site The Information revealed that OpenAI spends about $700,000 per day on ChatGPT, whose operation requires immense computational power. According to Reuters, each request made to ChatGPT would cost the company four cents. The high expenses of these services are not a new problem in the industry: the collaboration between OpenAI and Microsoft, for example, began because the former needed a commercial partner to reduce management costs. One of the first products born from the alliance between the two companies was Copilot, a tool for writing and editing code with the help of AI. Today, this service has about one and a half million users who pay ten dollars a month to use it: according to the Wall Street Journal, however, the company lost an average of more than twenty dollars per user per month last year.

Why Making Money with AI is Complicated

One of the problems in the industry is that AI systems, as sophisticated as they are costly to maintain, are used in the majority of cases for relatively simple tasks: as the Wall Street Journal points out, using them to summarize an email is like using a Lamborghini to deliver a pizza. For this reason, Adobe, which develops well-known software such as Photoshop and Illustrator, has chosen to limit the use of AI in its programs. The company does this through a credit-based system: once depleted, the service speed is greatly reduced to discourage overuse. At a time when many companies that have integrated AI systems into their products struggle to turn these investments into profits, the prevailing sentiment is that – after the first months of widespread investment – it's time to proceed more cautiously. Essentially, we are at a point where it is necessary to translate the excitement and interest in artificial intelligence into actual adoption, capable of being economically sustainable. Some even think that by the end of 2024, the cascade of investments in the sector will almost completely cease.

AI and its Impact on the World of Work

In this context, the concern that artificial intelligence will take away jobs from people becomes less solid. Most likely, these technologies will influence work processes, in some cases revolutionizing them, but this transition will occur gradually – and therefore it is believed that there will be ample space to predict policies and measures to mitigate potential impacts on unemployment. A study by MIT – one of the most authoritative universities in the world – argues that to date, in many areas, replacing human workers with AI systems purely in economic terms is not at all advantageous, contrary to what one might think. The research team has calculated the cost-benefit ratio concerning the hypothetical automation of 800 different occupations. The results show that only in 23% of cases would it be worthwhile to adopt artificial intelligence systems. In all other cases, it is still less expensive to hire employees. Similarly, with the advent of ChatGPT, some believed that such technologies could profoundly change journalism, and more generally, publishing. However, this has not yet happened: it has been noted that the challenges in having software write content do not only concern reliability or glaring errors, but rather the inability to create interesting content for readers.