Artificial Intelligence, in question due to its high energy consumption

La Artificial Intelligence It is not the future, It is already the present, but it is not without controversy, and it is not about data analysis and security, the ability to predict the future or the way in which human beings are already normalizing it in our days as a tool for help, but for the side effects that could cause on the planet. And according to a recent study, AI could require as much electricity consumption as a small nation whole. The study, which has significant implications for the future of AI and energy sustainability, comes by Alex de Vries, a PhD candidate at the University of Amsterdam School of Business and Economics.

The history of AI dates back to the 1950s, but In recent years, we have witnessed impressive growth. Vries' study tells us warns that this progress is not free in terms of energy consumptiona. In fact, AI consumes a significant amount of electricity both in its training phase as in its inference phase, and this could increase exponentially as AI becomes more integrated and normalized in our lives and we demand it. Not only as isolated processes, but for automations that can be programmed indefinitely., which means generating an incalculable amount of data just for that individual demand, Can you imagine multiplying that by the hundreds of millions of individuals in the world who use AI on a daily basis?

The AI ​​training process is one of the biggest energy consumers in the use of this technology. A telling example comes from Hugging Face, an artificial intelligence company, which reported that Just one of the available possible AI models requires 433 megawatt-hours (MWh) of electricity for training.This figure is enough to supply approximately 40 average households for one year.

You ask, AI answers

A study by research firm SemiAnalysis noted that OpenAI needed 3.617 NVIDIA HGX A100 servers, with a total of 28.936 graphics processing units (GPUs) so that ChatGPT can automatically respond, which implies an energy demand of 564 MWh per day.

«By 2027, global AI-related electricity consumption could increase to between 85,4 and 134,0 terawatt-hours (TWh) of annual electricity consumption.»

These data are alarming, and it is important to note that although some companies developing AI have expressed plans to improve the energy efficiency of their tools, Alex de Vries warns that this could lead to a increase in the use of AI, ultimately negating any energy efficiency benefits.

Smart and sustainable use of AI

The study also highlights that AI is not a universal solution and should not be used for everything, especially in consideration of its privacy implications and high energy consumption. Rather than abusing this technology, suggests that companies should use it consciously, reserving it only for cases where its application is truly necessary. and beneficial despite the associated energy costs. It also proposes that policymakers consider implementing environmental disclosure requirements for AI and its supply chains, in order to create greater transparency and accountability in the use of this technology.