Thinking

Balancing the environmental cost of AI

It’s almost impossible to use the internet today without encountering artificial intelligence. The “new” field has actually been around since the 1950s, but bigger, more comprehensive models and technologies seem to pop up at an almost exponential rate. This rapid growth has spurred important conversations on AI’s implications not only on our life but on the environment as well.

Some interesting facts about AI’s use on the environment:

  • Electricity: ChatGPT’s energy consumption over a year could power the entirety of Ireland for over two days. In 2023, data centers, which are used to operate AI models, represented 2% of the global electricity demand, a number that has only increased over the years and will continue to do so as larger models requiring more energy are built.
  • Carbon: Studies have shown that training a single AI model can produce as much carbon as five cars over their lifetimes of use - Similarly, data centers are responsible for over 2% of the world’s carbon output!
  • Water: One data center can use up as much coolant water per day as a town of 50,000 people! These centers are often located in areas where power is inexpensive or drawn from renewables, but oftentimes these are water scarce regions, such as the Southwest US.
  • E-waste: The servers and technology used to run AI models, store information, and facilitate access have to evolve alongside the programs they’re running. This means that the technology quickly becomes obsolete and is disposed of frequently (and oftentimes improperly), contributing to issues of e-waste and resulting in over-mining of natural resources necessary to make components such as microchips, etc. 

At the same time, AI can have uses that benefit the environment. For example, chatbots such as ChatGPT can surface new ideas, and AI is currently powering robots that quickly sort through recycled materials at plants. Investing in more thoughtful applications of AI, rather than integrating it just to keep up with what’s new and shiny, can help reduce the following environmental impacts. Companies who are more open about their resource usage will in turn help policymakers navigate the new territory of standards-making and emissions tracking for AI initiatives. Considering open-source solutions and the level of results truly needed can also help reduce the environmental footprint of an AI solution.  

When developing AI strategies, it’s important to account not just for the technological and business needs, but also the environmental impact. At the heart of a well-rounded AI strategy is finding that balance between performance, sustainability, and scalability. 

We emphasize balance—ensuring that the strategies we design take into account the broader context of people, technology, and the environment— to create systems that not only meet today’s needs but do so in a way that is sustainable for the future. With proper consideration for its implications on local (and global) resources and communities, more efficient, thoughtful use of AI could be of help in our fight against climate change.

Next post: The rise of greenhushing: It’s not easy being green
Next story

The rise of greenhushing: It’s not easy being green

While the issue of corporate greenwashing has received significant attention, a new worrisome trend has emerged: a practice being referred to as “greenhushing”.
Continue reading