ESR success in Climate Crisis AI Hackathon

Two of our ESRs (Shahine Bouabid and Peter Manshausen) worked with Paula Harder in a team for the Climate Crisis AI Hackathon, a 48h hour challenge organized by ConcordAI & MTL NewTech. It brought together hundreds of international talents to tackle problems at the intersection of AI and climate change, ranging from Kaggle-like challenges to open artistic performances. Peter, Shahine, and Paula used machine learning to show just how much climate change is reshaping our world, and their resulting piece won two awards: 'Experience sharing and networking opportunities with experts' and 'AI R&D Program Invitation'.

Shahine shares his thoughts on the event:

Peter, Paula Harder and I decided to step out of our comfort zone and take on the AI-artist challenge, using AI to transform climate and environmental data into a unique piece of art.

Where should Oxford be on the map 50 years from now if we'd like it to have the same temperature it has today? How north should the Sahara have shifted to maintain its average temperature from 1950? Humans are reshaping the world. We made it visible. 

Using CMIP6 historical and projected surface temperature fields from 1950 to 2100, we computed the smallest yearly spatial shift each location on Earth should accomplish to maintain its past temperature. Global Warming becomes Plate Tectonics: deserts grow, forests retract. continents drift and change their shape. To further extrapolate, we trained a CNN to predict yearly temperature variations based on global trends from available data and used it to keep land masses in motion beyond 2100.

Stepping a bit aside from the usual prediction-oriented ML mindset was a great opportunity for us to freewheel, experiment with programming tools and summon our creativity to see the larger picture and where iMIRACLI fits in from a different angle.

Congratulations to Shahine, Peter, and Paula! You can read more about their entry at Devpost.