The EPFL AI Center spoke to Dr. Xavier Ouvrard about the EcoCloud Research Facility.
AI is not disconnected from materiality, so it requires energy, it requires cooling, it requires a lot of topics that need to be covered by an holistic approach in order to improve the overall sustainability (of AI computing).
What projects is EcoCloud pursuing to make AI datacenters more sustainable?
At the moment I am supervising the installation of an experimental facility where we will be able to play on different subject.
First, we will be able to fine tune the power consumption of the different servers, we will have all the instrumentation for doing some fine tuning at the scale of the second, and even at the level of the milliseconds, and also we will have some research on cooling and we can think of retrieving the heat very close by the CPU, and valorize this energy to reinject it in the central heating for instance of the campus.
AI does have a large energy footprint but can it also have a positive impact?
It helps to solve challenges in sustainability, where you really work on the sustainability issues, you can think to Urban Twin, which is one of the projects where EcoCloud is involved and where you are going to make a digital twin of a city, having all the facets of energy, water consumption, buildings and also other data of the city where you can think of making some models to have a better sustainability.