Using generative AI to create images could be using as much energy as it takes to charge your smartphone, a new study has found.
The study, via Hugging Face and Carnegie Mellon, is the first of its kind to measure the environmental impact of generative AI models such as ChatGPT’s Dall-E and Midjourney – and it found that using an AI model such as these may produce more carbon than driving four miles.
Here’s what Dr. Sasha Luccioni, who led the study, said to Gizmodo on the subject:
People think that AI doesn’t have any environmental impacts, that it’s this abstract technological entity that lives on a ‘cloud’. But every time we query an AI model, it comes with a cost to the planet, and it’s important to calculate that.
The study highlighted the fact that image generation – in other words, turning text into images – used significantly more energy than any other task for generative AI. It also found that larger, multipurpose models such as ChatGPT use a lot more energy than models that are task-specific.
The study carried out involved testing 88 models on 30 data sets, and measured the carbon and energy impact of generative AI models.
Interestingly, the study did not turn its attention towards OpenAI for one reason; they don’t share data. According to Dr. Luccioni, that’s problematic. She said:
I think that for generative AI overall, we should be conscious of where and how we use it, comparing its cost and its benefits.
To put things into perspective, ChatGPT-4 has 1.76 trillion parameters – now, that’s a hecka lotta computation every time someone goes on there and types an inquiry.
Although Dr. Luccioni acknowledges there are benefits of deploying generative models like ChatGPT in certain circumstances, she does does “not see convincing evidence for the necessity of their deployment in contexts where tasks are well-defined.”
She also rightly points out that web search and navigation could easily use smaller models which would save on energy.
Personally, I find this scary as all hell. Is it any wonder we’re not seeing enough change when it comes to tackling the climate crisis, considering millions of people are using these AI models every day? It feels like insanity.
It’s also worthwhile for me to point out Geek Native’s own AI content policy, which you can find right here.
What’s your opinion on generative AI? Did you realise how much energy they used, and now you’ve seen the study, will you be thinking and using it differently? Contribute to the conversation by leaving me a comment – I promise to reply if I can!
