The hidden energy cost of Generative AI
Colin Bowdery
Chief Architect @ Ericsson | Wireless Networking Expert, Strategic Management
A recent article from MIT Technology Review discusses a study by researchers at Hugging Face and Carnegie Mellon University. They found that:
generating an image with a powerful AI model consumes as much energy as fully charging a smartphone.
The study also revealed that using an AI model for text generation is significantly less energy-intensive, with creating text 1,000 times using only as much energy as 16% of a full smartphone charge.
The research highlights that while training large AI models is energy-intensive, the actual use of these models contributes more to their carbon footprint. The study is the first to calculate the carbon emissions from using an AI model for various tasks. The team examined emissions from 10 popular AI tasks and found that image generation was the most energy and carbon-intensive task.
In contrast, the least carbon-intensive text generation model had emissions equivalent to driving 0.0006 miles in a gasoline-powered car.
The article emphasizes the importance of understanding these emissions to make informed decisions about using AI in a more environmentally friendly manner.
This article raises some good questions on the future of this nascent technology. As we debate the merits of Generative AI in academic or business settings we are missing a critical discussion: Should there be controls to the use of such technology in order to meet Net Zero targets and reduce our energy dependency? or is it an acceptable cost for the progress it may bring?
领英推荐
In either case, the energy consumption impacts of adopting technology like this are a critical impact in the TCO analysis of investments. I wrote about this a year ago in my blog. The concept is still as true today as it was back then, perhaps even more so.
The postings on this site are my own and do not necessarily represent the postings, strategies or opinions of my employer.
#GenAI #AI #NetZero #DALL-E #Chat-GPT #TechInnovation #Connectivity #InternetOfThings #DigitalTransformation #NetworkingTrends