A human wrote this, and a human edited it. A human laid out this page too.
I could have asked one of the popular generative artificial intelligence models to compose a 600-word essay, in this case about the concerns a middle-aged writer and editor holds about a flood of environmentally destructive new technology that threatens his industry and livelihood.
But I have always been a late adopter. I prefer to see how other people misuse new technology so that I can avoid their mistakes, and I like to take some time to weigh the costs versus the benefits. Yet try as I might, it is becoming difficult to avoid the new wave of AI as tech companies thrust chatbots and enhancements into every app and interface.
You’ve probably heard and read about the latest existential crisis that publishers like Red Flag Media, Grid’s parent company, find themselves in thanks to the generative AI boom. Red Flag Media hasn’t licensed the use of its Grid opus to any AI companies, and the same is true for thousands of other publishers and writers. But that hasn’t stopped AI companies from using our articles to develop their models, nor does it stop them from using our content to deliver summaries to readers who then might not bother to pick up a copy of Grid or visit the website. The question of whether this is theft or fair use is working its way through the courts, but, as Jordan Teicher’s article in this issue explores, the environmental impacts of the new models are devastatingly clear.
Generative AI seems low cost, like everything else we do in the “cloud,” the harmless-sounding name for the vast networks of computers housed in data centers owned by Amazon, Alphabet and other huge corporations. I don’t have to pay to see Gemini’s answers to my questions, and the same would be true if I asked Claude or ChatGPT to lend me a hand — although there are paid tiers available that boast unlimited use and the best AI models on offer. But one “free” ChatGPT query uses 10 times as much energy as a traditional Google search, and tech companies are building data center complexes covering hundreds of acres to handle the anticipated increase in traffic. Those vast banks of computers, however, aren’t fueled solely by the aspirations and egos of tech entrepreneurs and venture capitalists.
Data center complexes can suck up as much electricity as hundreds of thousands of homes. And as Grid has reported, the regional electrical system does not expand at the speed of AI. New power plants — whether powered by renewables or fossil fuels — take years to plan and then build. PJM Interconnection, the entity managing the grid that connects power suppliers with consumers, can take years to approve the connections that allow new producers to bring their electricity to market. Electricity prices will increase for everyone as demand from new data centers (on top of the rising demand for green technologies such as electric vehicles and electric heat pumps) outpaces the rate at which new producers can expand the supply. Even if the companies operating the data centers buy renewable electricity, doing so bids up the cost for other users, delaying decarbonization overall. Ultimately, the AI boom will require lots of climate-wrecking fossil fuels, principally the methane that Governor Josh Shapiro is so eager to see burned.
You could understandably wonder if I am picking on the environmental impacts of AI because I am a worker particularly threatened by it, like a stable owner 100 years ago complaining about automobile emissions. But my self-interest notwithstanding, we all need to recognize that even if the services seem free, the costs are more than our planet can bear.
