Skip to main content

Environmental impact of A.I.

IT industry is the fastest growing consumer of electricity. Even if the industry is on the forefront of also using renewable energy the ever growing demand is using much of the available clean energy from other uses. And the advent of A.I. is just accelerating the demand of electricity.

Even though the use of the A.I. powered tools and services are rapidly growing the usage isn't really the problem. There are recent studies that show that using an A.I. to for example generate content produces less emissions than doing it the traditional way. One study even points out that human generating a prompt for a language model has bigger carbon footprint than the machine produces while generating the answer!

The real problem is the energy used to train those models. It requires massive amount of power to train an advanced LLM. Due to these power demands Microsoft for example missed it target to resuce their carbon footprint and Amazon is buying their own nuclear powerplant to power their efforts to keep up with the game.

Even if the work done with the help of an A.I. can reduce the emissions the training will make the situation even worse as we seek to produce ever more capable models. The only hope to offset those increasing emissions is the promise that with the help of the machine we could find new ways to optimize the power usage and reduce emissions of other industries.