The Environmental Cost of A.I. ⇥ theatlantic.com
Karen Hao, the Atlantic:
Microsoft isn’t a company that exists to fight climate change, and it doesn’t have to assume responsibility for saving our planet. Yet the company is trying to convince the public that by investing in a technology that is also being used to enrich fossil-fuel companies, society will be better equipped to resolve the environmental crisis. Some of the company’s own employees described this idea to me as ridiculous. To these workers, Microsoft’s energy contracts demonstrate only the unsavory reality of how the company’s AI investments are actually used. Driving sustainability forward? Maybe. Digging up fossil fuels? As Prapoo put it in that September conference call, it’s a “game changer.”
Josh Saul, Naureen S. Malik, and Mark Chediak, Bloomberg:
From Florida to Oregon, utilities are racing to meet a surge in demand from power-hungry AI data centers, manufacturing facilities and electric vehicles. The staying power of gas, which in 2016 overtook coal as the No. 1 US source of electricity, has surprised some experts who not so long ago had projected the era of frenzied domestic demand growth for the fuel might soon come to an end.
[…]
Still, the booming power needs are hard to ignore. Electricity usage by data centers is poised to surge as much as ten times current levels by 2030. In addition to building new plants, some power companies will retire gas plants at a slower rate than previously expected, said power market analyst Patrick Finn of energy consultancy Wood Mackenzie. “It makes clean energy goals that much more difficult to attain,” he said.
It’s also worth noting that data centers are at the moment a relatively small slice of total worldwide energy usage — currently something like 1%, dwarfed by cars, heavy industry, commercial buildings, and so on. That could change — a report from the Electric Power Research Institute projected that the electricity required by AI companies could rise to reach up to 9% of the United States’ energy mix, which would, quite frankly, be insane. (If you think the web is overrun with AI content now, imagine a world where one tenth of all the electricity we generate is going into pumping out more of the stuff.)
BUT. There’s another element at work here, and that, as energy analysts have pointed out, is that utilities have a self-interested reason to take AI companies’ energy projections at face value, or even to inflate them: It means they can raise rates and lobby to build more gas plants!
Between these reports and the last batch I linked to, the non-A.I.-generated summary seems to be: businesses are largely entrusted with setting their own environmental targets by the accounting measures of their own design, which they are bending in order to hurriedly launch A.I. features.
I am not an A.I. doomer; I am also not a cheerleader. That umbrella of technologies covers a bunch of exciting innovations, many of which are already finding their way into many of our lives, whether we have agreed to that or not. But it is alarming to see the flexibility of even mediocre environmental targets because of this admittedly interesting technology. Yet I am not convinced it is worth accelerating the damage we are inflicting upon our world.