Reacting to the Environmental Encore of Creativity-Fueled Artificial Intelligence: A Light-Hearted Reflection
“Responding to the climate impact of generative AI”
“MIT researchers are suggesting that generative models, a kind of AI that’s used to make new content like deepfakes or AI art, could have as much of a climate impact in the future as aviation or concrete.” This statement sparks an amusingly overcast envisioning of a world where deepfakes become the newest culprits for climate change. Equating deepfake technology to belching airplane engines or smoke-spewing cement factories is indeed a fascinating comparison. Puts everything into a bit of an unusual perspective, doesn’t it?
Yet, there’s a robust scientific angle to these claims. Generative models, for those that aren’t in the know, are a type of AI that uses large amounts of data to predict and produce novel, convincing content that may as well be the next composition of Mozart or creation of Picasso. Or, in a more devious vein, it’s the mastermind behind the increasingly realistic deepfakes that have been causing quite a stir lately. While these models certainly aren’t spewing out smoke or guzzling gallons of aviation fuel, they do suck up significant amounts of electrical energy. This excessive energy consumption, unless fueled by renewable sources (and let’s be real, it most likely isn’t), could end up having a sizeable carbon footprint.
Interestingly, the MIT folks illustrated this with an example. Training an AI language model, OpenAI’s GPT-3, could potentially emit as much carbon dioxide as the average lifetime emissions of five cars. Five cars! That’s the stuff of nightmares for those who are environmentally conscious. Indeed, the environmental blowback of these AI models can be massive if not kept in check.
Nevertheless, the researchers haven’t just played the doom-and-gloom card. They’ve proposed potential solutions for this impending crisis. They assert the importance of reporting the energy use and greenhouse-gas emissions related to AI models. This sort of transparency could force companies to rein in their energy consumption voluntarily. It also could lead to the development of more energy-efficient AI training techniques. As the tech giants scramble to demonstrate how ‘green’ they are, this emphasis on transparency could likely become a win-win for everyone involved.
Furthermore, the push for using renewable energy sources for AI work comes across as a no-brainer. Couple this with the benefit of predictive abilities in AI – for example, the potential to accurately predict areas and times of excess renewable energy – and one can imagine a world where AI becomes not just environmentally neutral but possibly a net-positive influence on the climate.
So, here we are, holding up a magnifying glass to the scary monster called AI, which now supposedly adds climate villainy to its reputation. Yet, it’s also within this monster that we could find the solutions to the problem it poses. The irony, it seems, is not lost on AI either. Quite a charming world we live in, don’t you think?