We should consider Energy Consumption
Creating (“training”) models of the size and complexity of ChatGPT, Gemini, Claude, etc is VERY COSTLY. It is said that the energy cost of making ChatGPT corresponds to powering over 1,000 U.S. households for a year (these calculations are obviously not straightforward, also because we do not have access to all training details from Open AI, the company which made ChatGPT.) Using Large Language Models also costs energy, and it’s important that we are aware of such costs and use these tools responsibly. However, we should also put the use of these tools in the same box as using other digital tools which we often take for granted, but also consume massive amounts of energy, such as Netflix, Instagram, TikTok, etc.
