Generative AI using excessive energy 

Training artificial intelligence has led to big increases in carbon emissions, according to disclosures made by Google. 
 
According to an environmental report made by the tech industry giant, carbon emissions were higher than in the year 2019 by 48 percent, with AI being a large contributor.  

One recent study suggested that a generative AI system such as Chat GPT might use up to 33 times more energy than machines running task-specific software.  

Google’s disclosure comes after similar declarations were made by the likes of Microsoft. 

In its sustainability report released in May, Microsoft declared an emissions growth of 29 percent since 2020. Microsoft attributed this to the building of more data centers “designed and optimised to support AI workloads.” 

Jesse Dodge is a Senior Researcher at the Allen Institute for AI. He estimated that one query to ChatCPT uses about as much electricity as a light bulb used of 20.  

“So, you can imagine with millions of people using something like that every day, that adds up to a really large amount of electricity,” he said. 

The research comes after some economists from Goldman Sachs have questioned whether or not generative AI will prove to be worth the cost of investment, with Jim Covello arguing that the technology is not suited to solve the complex problems that would justify the cost of investmnent. 

 
Tech industry figures believe, however, that AI could also be used to curb carbon emissions, with Bill Gates among those who have suggested this.   

“The question is, will AI accelerate a more than six percent reduction? And the answer is: certainly,” he told the Financial Times.