How to build AI models that don’t hurt Mother Earth – one AI model at a time. – Birger Ober, Sustainable Digital Infrastructure (SDIA) & Christoph Buchli, Helio AG


If training a single AI model can emit as much as five cars during their lifetimes, how can I as a data analyst or scientist ensure that my AI models do not harm the environment? You will walk away from this session with a deeper understanding of how you and your organisation can minimise impact, without sacrificing performance.

Key Takeaways

  • Creating environmentally friendly Analytics and AI models is not just about using renewable energy. We’ll show you the bigger picture and where the opportunities are to realise cost advantages and reduce your environmental impact.
  • Learn how to package your AI models in delay-tolerant, interruptible work-packages that can be run on idling or spare computing capacity rather than requiring 100% reliable & available data centers. Go on to understand how to you can make your models transportable across data centers so that your model can travel to the data and not the other way around.
  • Learn the latest on how to assess and select data centers that use green power, maximise utilisation, re-deploy excess heat and much more. See what true sustainable AI infrastructure looks like as we showcase a new data center in Jordbro Stockholm – powered by the SDIA member Vattenfall.

Add comment