O impacto ambiental da IA generativa – Cryptopolitan
The rapid growth of generative AI tools for text, images, and audio has ushered in a new era of technological innovation. However, beneath the surface of these remarkable achievements lies a hidden cost that demands our attention—environmental sustainability. This article explores the environmental footprint of generative AI, including energy consumption, water usage, and the quest for more efficient models that can minimize the ecological impact. Generative AI models, in their quest for excellence, have an insatiable appetite for energy. These models often rely on large-scale cloud providers equipped with energy-hungry chips containing a multitude of transistors. In fact, these advanced chips can demand at least ten times more energy than traditional versions. Unsurprisingly, models utilizing extensive training data and numerous parameters consume even more energy. For perspective, training OpenAI‘s GPT-4 is estimated to have emitted roughly 300 tons of CO2. To put this into context, that’s equivalent to the emissions from 300 round-trip transatlantic flights. Importantly, this calculation doesnt account for the ongoing emissions generated during the use of the product. The environmental impact of generative AI extends beyond energy consumption to water usage. It has been estimated that a simple ChatGPT conversation can deplete the equivalent of a 500ml water bottle. Particularly