Human(e) Futures Colloquium - The High Environmental Cost of AI Use

The High Environmental Cost of AI Use 

New data centres are being built worldwide (including one planned locally, just outside Regina in RM Sherwood).  These use stunning amounts of electricity and water taken from local power grids and groundwater supplies to cool the huge computers required for LLMs ("genAI") like ChatGPT.  I"ll discuss how AI data centres are different from data centres used for more traditional computer cluster work, like the CANFAR/Compute Canada data centres I use for my own astronomy research (which still use a lot of energy, but more efficiently).  Companies like SpaceX are now pushing to send "AI data centres" into orbit, with SpaceX submitting a proposal for a million new, huge satellites.  They argue this is less environmentally damaging, which is only true if you ignore the huge environmental costs of launches, increasing collision risks in orbit, and the deposition of metal, plastic, batteries, and computers into Earth's atmosphere (and possibly onto Earth's surface as happened here in Saskatchewan in 2024) after these satellites reach their end of life after a few short years.  We'll end with a discussion of ways that governments need to carefully reprioritize how we make use of the limited orbital space around the Earth for satellites before we end up in the worst-case scenario of Kessler Syndrome, when no one would get to use satellites anymore. 

Presenter: Samantha Lawler (Dept. of Physics and Astronomy)