Tasks 🥇

We chose three tasks across three modalities, all of which are useful in the global fight against the climate crisis:

  1. 📝 Detecting climate disinformation 📝: based on text from news articles.
  2. 🔥 Classifying regions at risk of wildfires 🔥: based on image data gathered via satellite.
  3. 🌳🪓 Detecting illegal deforestation 🌳🪓 : from bio-acoustic data recorded in the jungle

The datasets for each task will be hosted on the Hugging Face Hub.

Why these tasks?

We want to make sure that our AI applications match real life needs in the fight against climate change and environmental destruction. We settled on wildfires, deforestation of primary forests, and climate disinformation.

With the planet heating quicker and quicker, dangerous wildfires have become a constant reality. New analyses by the World Resources Institute show that over the last 25 years alone, the area burned by wildfires every year has increased by more than 5%. That additional burned area amounts to 6 million hectares of land per year, the equivalent of an entire country like Croatia. To address this problem, one task will focus on the image-based classification of regions according to their wildfire risk.

Deforestation by humans is just as big a problem. In 2023 alone, 19.2 million hectares of forest were lost due to human activities such as logging and farming, according the Global Forest Watch, often illegally. While in many cases, this takes place out of sight, there are hints of sound that point to these destructive activities. In that spirit, the second task will be to identify illegal logging in forests based on bio-acoustic data.

A third major problem is the rise of climate-related disinformation. A recent scientific article estimated that 14.8% of Americans do not believe in anthropogenic climate change. Much of this is probably due to disinformation. With climate change being one of the central threats to the wellbeing of our societies, this is a problem that must urgently be addressed. Our third and last task will therefore challenge participants to detect climate disinformation in newspaper articles.

Of course, there are many other issues that must urgently be addressed. We hope that our challenge will kickstart the development of more frugal AI models that address the full spectrum of the environmental problems (and opportunities!) we are facing at the decisive moment for the future of our planet.

Task Partners

We will announce more details about the tasks and release the datasets in the coming weeks!