At the moment generative AI is impossible to ignore on the internet. An AI-generated summary might randomly appear at the top of search results every time you do a Google search. Or you might be asked to try AI Meta Tool while browsing Facebook. And that the ubiquitous sparkle emoticon still haunts me in my dreams.
This rush to add AI to as many online interactions as possible can be traced back to OpenAI’s groundbreaking release of ChatGPT in overdue 2022. Silicon Valley soon became obsessed with generative AI, and nearly two years later, AI tools based on gigantic language models have become an integral part of online user experiences.
One unfortunate side effect of this proliferation is that the computational processes required to run generative AI systems are much more resource-intensive. This has led to the advent of the era of hyperconsumption of the internet, a period defined by the proliferation of a fresh kind of computing that requires excessive amounts of electricity and water to build and operate.
“The backend algorithms that any generative AI model has to run are fundamentally very, very different from the traditional Google search engine or email,” he says. Sajjad Moazeniresearcher in computer engineering at the University of Washington. “For basic services, they were very lightweight in terms of the amount of data that had to be passed back and forth between processors.” By comparison, Moazeni estimates that generative AI applications are about 100 to 1,000 times more computationally intensive.
Energy requirements for training and deployment are no longer generative AI’s filthy secret, as experts last year predicted energy demand would spike in data centers where companies work on AI applications. Google, almost on cue, recently stopped considering itself carbon neutraland Microsoft can trample on it sustainable development goals underfoot in the ongoing race to create the biggest and best AI tools.
“The carbon footprint and energy consumption will be proportional to the amount of computation that is being performed, because essentially these data centers are being powered in proportion to the amount of computation that is being performed,” he says. Junchen Jiangresearcher in networked systems at the University of Chicago. The larger the AI model, the more computation is often required, and these frontier models become absolutely gigantic.
Even though Google’s overall energy consumption doubled between 2019 and 2023, company spokeswoman Corina Standiford said it would not be fair to say that Google’s energy consumption has skyrocketed during the AI race. “It’s incredibly difficult to reduce emissions from our suppliers, which make up 75 percent of our carbon footprint,” he says in an email. The suppliers Google blames are manufacturers of servers, networking equipment and other technical infrastructure for data centers — an energy-intensive process that’s required to create the physical parts for pioneering AI models.