Generative AI tools, which turn text prompts into images and videos, have a bigger carbon footprint than thought. Researchers at Hugging Face found that video generators use much more energy than expected, quadrupling their power demand when the video length doubles. This means a six-second clip consumes four times as much energy as a three-second one.
Experts warn that we’re launching these tools without understanding their environmental impact. The current design is inefficient, and new technologies are needed to reduce energy consumption. However, there are ways to slim down demands, such as intelligent caching and “pruning” from training datasets.
The scale of AI’s impact is substantial, with 20% of global datacenter power demands now attributed to AI usage. Tech giants like Google are investing billions in infrastructure, but their climate goals are being abandoned in the process. A recent study found that Google’s carbon emissions increased by 13% last year due to its generative AI tools.
The environmental impact of these tools is unknown, but it’s likely worse than thought. The widespread use of generative AI has significant implications for our planet.
Source: https://futurism.com/future-society/ai-power-usage-text-to-video-generator